Accepted Abstracts & Papers

Abstract: This study presents a simulation-calibrated digital twin framework for real-time energy management and fault detection in commercial buildings. Commercial buildings often consume more energy than expected because of HVAC faults, sensor drift, control problems, and differences between assumed and actual occupancy patterns. These issues are not always detected by conventional building management systems, which can lead to unnecessary energy use and delayed maintenance actions. The proposed framework is designed as a three-layer digital twin process that combines building simulation, machine learning, and anomaly detection. The first layer develops a high-fidelity EnergyPlus model calibrated using sub-hourly building management system data and assessed against ASHRAE Guideline 14 criteria. The second layer uses machine learning surrogate models, including XGBoost and Random Forest, to support fast energy prediction at the zone level. These surrogate models reduce the computational burden of running full simulation models while keeping useful prediction accuracy for operational use. The third layer compares live sensor readings with digital twin forecasts and applies Isolation Forest and autoencoder-based anomaly detection to identify possible faults in real time. The framework is tested using a UK commercial hotel archetype. The results show calibrated model performance with NMBE below 5% and CV-RMSE below 15%, surrogate model R² values above 0.85, and fault detection accuracy of 91% with a false-positive rate below 8%. The digital twin-based scheduling approach also estimates energy reduction of 12–18% compared with baseline operation. The study shows that combining simulation calibration, machine learning inference, and real-time fault detection can support building energy management, reduce avoidable energy use, and improve operational decision-making. The proposed framework can be used as a repeatable process for future smart building applications and can support the development of net-zero-ready commercial building operations through better monitoring, prediction, and control.

Keywords: Digital twin, building energy management, fault detection, EnergyPlus, machine learning, anomaly detection, HVAC optimisation, building calibration, smart buildings, net-zero

https://icsest.org/accepted-papers.html#CA0111

Abstract: This study proposes an AI-driven digital twin framework for transferring data, modelling methods, technical expertise, and infrastructure from the oil and gas sector to renewable and hybrid energy systems. The energy transition does not necessarily require the complete abandonment of existing petroleum-sector capabilities. Many methods developed for reservoir characterization, seismic interpretation, well-log analysis, hydrodynamic simulation, predictive maintenance, and production optimization can be adapted to support low-carbon energy assets. The proposed framework combines physical models, operational data, machine learning, real-time sensing, and scenario analysis within a digital twin architecture. It is designed to support both subsurface and surface energy systems, including geothermal assets, carbon storage sites, hydrogen infrastructure, offshore wind facilities, energy storage systems, and hybrid renewable networks. The framework includes four main layers: data acquisition and integration, physical and data-driven modelling, real-time state estimation and prediction, and decision support for operational optimization. Machine learning methods are used to identify patterns, estimate system behavior, predict equipment or asset performance, and evaluate alternative operating scenarios. The study also considers how existing oil and gas data platforms, engineering workflows, and workforce capabilities can be repurposed for renewable energy and smart-grid applications. A structured transfer methodology is proposed to identify which models, datasets, technical skills, and digital tools can be reused directly, which require adaptation, and which are unsuitable for cross-sector application. The framework may help energy companies reduce duplication, shorten development time, improve asset monitoring, and support workforce transition. However, its effectiveness depends on data quality, model validation, interoperability, cybersecurity, governance, and the physical differences between petroleum and renewable energy systems. The study concludes that digital twins can provide a practical bridge between traditional and low-carbon energy sectors when they are supported by clear validation procedures, domain-specific adaptation, and coordinated workforce development. Future work should apply the framework to a defined energy asset and evaluate its predictive accuracy, operational value, implementation cost, and scalability.

Keywords: Digital twin, artificial intelligence, energy transition, oil and gas, renewable energy, machine learning, petroleum geosciences, sustainable energy systems

https://icsest.org/accepted-papers.html#CA0125

Abstract: This study investigates the prediction of energy consumption in artificial intelligence training workloads using statistical, machine learning, ensemble, boosting, and neural network models. The increasing size and computational demand of AI models have raised concerns about electricity use, operating cost, and environmental impact. Reliable energy prediction can help researchers and organizations estimate resource requirements, compare model configurations, and plan more energy-efficient training processes. The study analyzes a dataset containing 1,059,206 observations from training workloads involving MobileNetV2, EfficientNetB0, VGG16, ResNet50, InceptionV3, and DenseNet121 architectures. Energy consumption measured in watt-hours is used as the target variable, while the predictors include power consumption, training duration, model architecture, training accuracy, loss, and temporal characteristics. The analysis follows a machine learning workflow that includes exploratory data analysis, feature engineering, multicollinearity assessment, data scaling, model training, and performance evaluation. Seven prediction methods are compared: Ordinary Least Squares Regression, Ridge Regression, Lasso Regression, Classification and Regression Trees, Random Forest, Extreme Gradient Boosting, and Artificial Neural Networks. Model performance is evaluated using the coefficient of determination, Mean Absolute Error, and Root Mean Squared Error. The results show that the linear regression model achieves an R² of 0.9382, while Random Forest reaches an R² of 0.9719. XGBoost produces an R² of 0.9994 and a Mean Absolute Error of 0.1517 Wh. The Artificial Neural Network reports the lowest prediction errors, with an R² close to 1.0000, a Mean Absolute Error of 0.0313 Wh, and a Root Mean Squared Error of 0.0471 Wh. Feature analysis identifies training duration and power consumption as the main predictors of energy use. The findings show that operational and workload characteristics can support energy forecasting for AI training. The proposed comparison may assist with workload planning, energy monitoring, model selection, and the development of sustainable computing practices.

Keywords: Sustainable AI, energy consumption, machine learning, artificial neural networks, XGBoost, green computing, AI sustainability, predictive analytics

https://icsest.org/accepted-papers.html#CA0115

Abstract: This study presents a hybrid diagnostic framework combining digital twin technology and machine learning for real-time monitoring and anomaly detection in Proton Exchange Membrane (PEM) water electrolyzers. PEM electrolyzers are essential for green hydrogen production, but their performance and lifetime can be affected by power supply fluctuations, operating temperature variations, and electrochemical degradation. To address these challenges, the proposed system integrates a MATLAB/Simulink-based PEM electrolyzer model with Azure Machine Learning and Azure Digital Twins. Simulated operating data, including electrolyzer current, voltage, and hydrogen flow rate, are used to train predictive models and visualize system behavior in a digital twin environment. Extreme Random Trees models are applied for current and voltage prediction, while an Elastic Net model is used for hydrogen flow rate estimation. The predicted values are compared with simulated values to identify deviations and support early anomaly detection. The results demonstrate strong prediction accuracy, with coefficients of determination of 0.99820, 0.99693, and 0.99665 for current, voltage, and hydrogen flow rate, respectively. The corresponding Normalized Root Mean Square Error values indicate low prediction error and confirm the reliability of the proposed approach. The findings show that digital twin and machine learning integration can improve PEM electrolyzer diagnostics, support failure anticipation, and contribute to more reliable and efficient green hydrogen production systems.

Keywords: Anomaly, Diagnostics, Digital Twin, PEM Electrolyzer, Machine Learning

https://icsest.org/accepted-papers.html#CA7113

Abstract: This study presents a framework for integrating smart technologies into sustainable building design within civil and environmental engineering. The proposed approach combines sensor networks, real-time data analytics, intelligent control systems, renewable energy integration, and adaptive energy management to improve building performance. Environmental conditions, occupancy patterns, indoor air quality, temperature, lighting demand, and energy use are continuously monitored to support dynamic control of heating, ventilation, air-conditioning, lighting, and other building systems. The framework also incorporates solar photovoltaic generation and demand forecasting to improve the use of locally produced renewable electricity and reduce dependence on conventional grid supply. Adaptive algorithms estimate short-term energy demand and renewable energy availability, allowing building systems to adjust operating schedules and equipment settings in advance. Simulation models and pilot applications are used to examine changes in energy consumption, operating costs, greenhouse gas emissions, and indoor environmental quality. The reported findings indicate that coordinated sensing, prediction, and automated control can reduce unnecessary energy use while maintaining occupant comfort. The framework also supports better regulation of temperature, lighting, and air quality, which may improve occupant well-being and productivity. The study highlights the value of combining civil engineering design principles with digital intelligence to create buildings that respond to changing environmental conditions, occupancy levels, and energy availability. The proposed approach may be applied to residential, commercial, institutional, and mixed-use buildings in different urban settings. However, its effectiveness depends on building type, climate, control strategy, sensor reliability, occupant behavior, data quality, and the availability of renewable energy resources. Future work should provide detailed numerical results, compare the smart framework with conventional building operation, evaluate implementation costs, and examine cybersecurity, maintenance, interoperability, and user acceptance.

Keywords: Sustainable building design, smart technologies, energy efficiency, civil engineering, environmental engineering

https://icsest.org/accepted-papers.html#CA0127

Abstract: This study examines the efficiency limitations of thermoradiative (TR) cells and explores the potential use of low-bandgap molecular materials for improving nighttime radiative energy conversion. Conventional photovoltaic (PV) cells generate electricity by absorbing solar radiation during daylight, whereas TR cells operate through the reverse principle by emitting thermal radiation from the Earth toward the cold night sky. Although TR cells offer the possibility of nighttime power generation, their practical output remains limited and is reported to reach only about 25% of the output of conventional PV cells. This limitation is mainly associated with the bandgap requirements of suitable semiconductor or molecular materials. The study discusses the physical operating differences between PV and TR devices, including radiative exchange between hot and cold reservoirs and the role of electrochemical potential in photon-mediated energy conversion. It further proposes the investigation of selected low-bandgap molecular structures, including halogen-based chemical compounds, as potential material candidates for TR cell design. The work highlights the need for improved material selection and molecular engineering to enhance TR cell efficiency. The study contributes a conceptual direction for future research in nighttime renewable energy generation and low-bandgap energy-conversion materials.

Keywords: Thermoradiative cell, photovoltaic cell, bandgap, chemical molecules, halogens, heat sink

https://icsest.org/accepted-papers.html#CA0603

Abstract: This study presents a Digital Twin-ready web interface for intelligent monitoring of concrete structures using automated non-destructive testing. The study addresses the need for accurate and organized field data in infrastructure monitoring, especially for concrete structures where condition assessment is often completed through manual testing, recording, and interpretation. Manual procedures in Ultrasonic Pulse Velocity and Rebound Hammer testing can increase the possibility of recording errors, delay data processing, and limit the use of inspection results for digital infrastructure management. To address these issues, this study develops a web-based platform that supports automated data capture, real-time processing, condition classification, and digital storage of concrete testing results. The platform is developed using the Flask framework and applies Optical Character Recognition to extract readings from non-destructive testing instruments through smartphone-based image acquisition. The collected data are processed to calculate pulse velocity, classify concrete quality based on accepted criteria, estimate compressive strength, and store inspection records in a centralized database. The system was evaluated using fifty concrete specimens tested through both Ultrasonic Pulse Velocity and Rebound Hammer methods. The validation process examined the ability of the platform to reduce manual data entry, support rapid condition assessment, and provide a user-friendly interface for inspection data management. The results show that the developed platform can automate data extraction and provide structured access to testing results for monitoring purposes. The interface also supports visualization, storage, and retrieval of inspection information, allowing the physical condition of concrete specimens to be represented in a digital format. This digital record can support future Digital Twin applications by providing a basic data acquisition layer for structural condition monitoring. The study shows that integrating automated non-destructive testing, web-based data management, and digital monitoring tools can improve inspection workflow and support smarter infrastructure maintenance. Future development may include direct IoT sensor integration, automated reporting, predictive analytics, and wider testing on field structures.

Keywords: Digital Twin, Intelligent Monitoring, Non-Destructive Testing, Automation, UPV, Rebound Hammer, Optical Character Recognition, Smart Infrastructure

https://icsest.org/accepted-papers.html#CA0109

Abstract: This study proposes a privacy-preserving AI-based cyber-physical security framework for renewable smart grids and IoT-supported energy management systems. The framework addresses four connected needs: cyberattack detection, privacy protection, physical-system validation, and secure operational response. Renewable smart grids depend on distributed energy resources, smart meters, photovoltaic systems, wind generation, battery storage, electric vehicle charging, IoT sensors, edge devices, and digital control platforms. These connected systems improve energy management but also increase the risk of false data injection, device compromise, denial-of-service attacks, unauthorized control actions, replay attacks, privacy leakage, and model poisoning. The proposed framework is organized into five layers. The first layer collects renewable energy, device, and grid measurements. The second layer performs local data filtering, synchronization, cleaning, normalization, and feature extraction at the edge. The third layer applies federated learning, differential privacy, secure aggregation, and protected model updates so that distributed nodes can support collaborative learning without sending raw energy data to a central server. The fourth layer combines cyber indicators with physical measurements such as voltage, current, frequency, power, battery state of charge, renewable generation, and load behavior. A combined anomaly and risk score is used to distinguish possible cyberattacks from equipment faults, renewable intermittency, communication errors, and normal operational changes. The fifth layer supports alerting, command validation, device isolation, secure reconfiguration, and operator decision support. The paper also compares the framework with existing studies and reviews the suitability of power-system, IoT, smart-meter, renewable simulation, and hybrid cyber-physical datasets for future evaluation. The analysis shows that no single dataset can test all framework functions. A combined evaluation approach is therefore required. The proposed model provides a structured basis for future implementation and testing of secure, privacy-aware, and resilient renewable smart-grid systems.

Keywords: Renewable smart grids, cyber-physical security, privacy-preserving AI, federated learning, IoT energy management, distributed energy resources, false data injection, anomaly detection, risk scoring, secure energy management

https://icsest.org/accepted-papers.html#CA9436

Abstract: This study proposes a blockchain-enabled Internet of Things framework for decentralized energy management in smart urban infrastructure. The framework addresses problems found in centralized energy management systems, including single points of failure, limited settlement transparency, privacy risks from detailed meter data, and weak coordination among utilities, building operators, aggregators, municipalities, electric-vehicle operators, and prosumers. The proposed architecture combines urban energy assets, IoT sensing devices, edge gateways, decentralized identities, a permissioned blockchain, smart contracts, and energy management applications. Smart meters and controllers collect information from buildings, rooftop photovoltaic systems, battery storage, electric-vehicle chargers, HVAC systems, lighting, and flexible loads. Edge gateways validate, filter, aggregate, and sign the collected information before selected records are submitted to the blockchain. Detailed operational data remain in protected off-chain storage, while hashes, credentials, event metadata, commitments, and settlement proofs are recorded on-chain. Smart contracts manage participant registration, offer submission, rule validation, local market clearing, demand-response rewards, settlement, reputation updates, and dispute handling. Safety-related control remains at the device or edge level, while operational scheduling is performed at 15-minute intervals and settlement is completed through the permissioned ledger. The framework is evaluated using a simulation of an urban micro-district containing smart buildings, photovoltaic generation, battery storage, controllable loads, and electric-vehicle charging over a 24-hour period. Three operating cases are compared: an uncoordinated baseline, a centralized energy management system, and the proposed blockchain-IoT framework. The simulation reports a reduction in peak demand from 132 kW to 112 kW, an operating cost improvement of 18.7%, an emissions-proxy reduction of 14.9%, and a renewable self-consumption improvement of 16.8%. The modeled permissioned blockchain also maintains settlement latency suitable for 15-minute energy management intervals. The study shows that blockchain can support energy coordination when it is used for identity, verification, settlement, and audit rather than direct safety-critical grid control.

Keywords: Blockchain, Internet of Things, smart contracts, decentralized energy management, smart urban infrastructure, demand response, smart cities, renewable energy integration

https://icsest.org/accepted-papers.html#CA1131

Abstract: This study proposes a federated learning framework for privacy-preserving residential load forecasting in European smart cities. Smart meters and Internet of Things devices generate detailed household electricity data that can support demand-side management, peak reduction, tariff planning, and renewable energy integration. However, centralized machine learning systems require household data to be collected and stored on a common server, which may increase privacy risks and create concerns related to data minimization, purpose limitation, and privacy-by-design requirements. The proposed framework keeps residential smart-meter data on local household gateways or edge devices and shares only protected model updates with a central aggregation server. A hierarchical federated learning structure is used to coordinate model training across multiple households and local groups without transferring raw consumption records. Secure aggregation is applied to combine local model updates, while differential privacy is used to reduce the risk of information leakage from shared gradients. The framework is designed for short-term residential load forecasting and can support utilities and energy aggregators in demand-response scheduling, peak shaving, and time-varying tariff design. The proposed method is evaluated using a multi-year residential electricity dataset from a European smart-city setting. Its forecasting performance is compared with a centrally trained deep learning model using standard prediction measures such as Mean Absolute Error, Root Mean Squared Error, and coefficient of determination. The evaluation also considers communication cost, privacy protection, model convergence, and the effect of non-identical household consumption patterns. The study aims to determine whether federated learning can maintain useful forecasting accuracy while reducing the need to centralize sensitive household data. The framework may support the development of privacy-aware demand-side management systems that allow energy providers to use distributed intelligence while keeping detailed residential electricity records under local control. Further validation is required to examine privacy loss, communication overhead, household participation, and performance under different building types, weather conditions, and energy-use patterns.

Keywords: Federated learning, residential load forecasting, GDPR compliance, smart cities, demand-side management

https://icsest.org/accepted-papers.html#CA0118

Abstract: This study proposes HAISEM, a hybrid artificial intelligence framework for solar power forecasting and energy-management support in West African power systems. The region has high solar-resource potential but continues to face weak grid infrastructure, limited reserve margins, diesel dependence, and atmospheric conditions that are not well represented in many forecasting models developed for temperate climates. The proposed framework combines gradient-boosted decision trees, including XGBoost and LightGBM, with a long short-term memory network. The tree-based models use engineered predictors such as clear-sky index, aerosol optical depth proxies, ambient temperature, relative humidity, time variables, and seasonal indicators, while the recurrent network captures short-term temporal patterns in irradiance and solar generation. The outputs of the base models are combined through a regularized meta-learner, and model hyperparameters are selected through Bayesian optimization using a Tree-structured Parzen Estimator. Explainability is incorporated through SHAP-based feature attribution at two levels: during feature selection and as an operator-facing layer that identifies the main drivers of each forecast. The framework also proposes an evaluation method that weights forecast errors according to their operational consequences for reserve scheduling, storage use, curtailment, and hybrid diesel–solar dispatch. Implementation is planned in Python using scikit-learn, XGBoost, LightGBM, TensorFlow, Optuna, and SHAP, with meteorological inputs drawn from open reanalysis products and available ground-station observations. The study is presented as a proposed methodology, and no experimental validation has yet been completed. Its intended contribution is a regionally adapted and explainable forecasting architecture for dust-affected, data-scarce tropical power systems. Future work should validate the framework using measured solar and meteorological data from defined West African sites, compare it with persistence and single-model baselines, evaluate forecast uncertainty, and quantify the operational value of improved predictions for grid and storage management.

Keywords: Artificial intelligence, solar energy forecasting, explainable machine learning, renewable energy, energy management, explainable artificial intelligence

https://icsest.org/accepted-papers.html#CA0134

Abstract: This study examines the economic viability of smart grid infrastructure for supporting renewable energy integration. Conventional power networks often face difficulties when managing variable and distributed electricity generation from solar and wind sources. The proposed analysis evaluates whether investments in smart meters, IoT sensors, artificial intelligence-based load management, energy storage, and digital control systems can produce long-term economic benefits that justify their initial cost. A cost-benefit assessment is conducted over a 20-year period by comparing conventional grid infrastructure with modernized smart grid systems. The evaluation includes capital expenditure, maintenance cost, technical energy losses, peak-demand charges, outage-related losses, system reliability, and renewable energy integration benefits. Data are drawn from 15 smart grid pilot projects implemented in Europe, North America, and Asia. The analysis estimates that smart grid systems can reduce technical transmission and distribution losses by approximately 12–15% and lower peak-demand-related costs by about 18%. The expected payback period ranges from 7 to 11 years, depending on grid density, infrastructure condition, electricity demand, and the share of renewable generation. The findings also indicate that avoided outage costs account for a large part of the total economic benefit, showing that reliability improvement should be included as a financial value rather than treated only as a technical outcome. The study suggests that smart grid investment can provide positive long-term returns while improving system flexibility, supporting renewable energy use, and reducing disruption risk. A phased implementation strategy is proposed, beginning with dense urban areas and high-demand zones where early financial returns may be stronger. The study provides a cost-based framework that can support policymakers, utilities, investors, and infrastructure planners in evaluating grid modernization projects under different technical and economic conditions.

Keywords: Smart grids, renewable energy economics, infrastructure investment, cost-benefit analysis, grid modernization

https://icsest.org/accepted-papers.html#CA0120

Abstract: This study presents the optimization and performance evaluation of a hybrid renewable energy system combining photovoltaic (PV) generation, micro-hydropower, and pumped hydro energy storage for rural electrification. The objective is to enhance energy reliability and system efficiency in off-grid communities by integrating complementary renewable sources with storage. The proposed system configuration includes a 100 kW PV array, a 200 kW micro-hydropower plant, and a pumped storage system designed to balance fluctuations in water flow and solar irradiance. System optimization is carried out using HOMER Pro software to minimize net present cost (NPC), cost of energy (COE), and unmet load, while MATLAB/Simulink is used for dynamic validation. A modified incremental conductance algorithm is implemented for maximum power point tracking and system sizing. The case study is based on Mutobo village in Rwanda, incorporating real meteorological and hydrological data. Simulation results indicate that the hybrid system effectively compensates for seasonal variability, with solar generation supporting hydropower during dry periods and pumped storage ensuring continuous supply. The optimized configuration demonstrates improved reliability, efficient energy utilization, and suitability for rural electrification. The study confirms that hybrid PV–hydropower systems with pumped storage offer a scalable and cost-effective solution for sustainable energy deployment in remote regions.

Keywords: Hybrid Power System, Micro-hydropower, PV systems, Pumped storage, HOMER optimization

https://icsest.org/accepted-papers.html#CA7436

Abstract: This study presents the design and evaluation of a smart trumpet-assisted Archimedes wind turbine for improving energy harvesting under low-speed and multidirectional wind conditions. Conventional horizontal-axis wind turbines may perform poorly in urban environments because of turbulent, changing, and low-velocity airflow. The proposed system combines a multidirectional trumpet-shaped wind concentrator, guide vanes, a convergent section, a honeycomb flow conditioner, an Archimedes spiral rotor with a 60° blade opening angle, a permanent magnet generator, a divergent section, and an HMI–PLC-based monitoring and control platform. The complete geometry is developed in SolidWorks, while the aerodynamic behavior is analyzed through three-dimensional computational fluid dynamics simulations in ANSYS Fluent 2025 R1 using the SST k–ω turbulence model. Simulations are performed at inlet wind speeds of 3, 5, 10, and 15 m/s. The numerical results show progressive airflow acceleration through the trumpet-shaped concentrator, improved pressure distribution, and stable aerodynamic loading across the investigated operating range. The turbine behavior changes from primarily drag-assisted starting at lower wind speeds to combined lift-and-drag energy extraction at higher wind speeds. The reported drag coefficient ranges from 0.6705 to 0.6972, while the lift coefficient ranges from 0.0311 to 0.0566. Electrical power output increases from 4.33 W at an inlet speed of 3 m/s to 597.72 W at 15 m/s. The corresponding turbine efficiency increases from 33.31% to 36.83%. The main contribution is the integration of airflow concentration, an Archimedes rotor, and intelligent monitoring within a single wind energy system. The proposed design may support decentralized renewable energy generation in urban and other low-wind environments where conventional turbine configurations are less effective. Further work should include experimental validation, comparison with an unconcentrated Archimedes turbine, uncertainty analysis, acoustic testing, structural assessment, and long-term outdoor performance evaluation.

Keywords: Archimedes wind turbine, sustainable energy systems, low-wind-speed energy harvesting, computational fluid dynamics, smart wind energy, renewable energy

https://icsest.org/accepted-papers.html#CA0126

Abstract: This study develops a coupled numerical and machine-learning framework for recovering low-grade heat from urban detention ponds, reducing summer thermal pollution in receiving streams, and storing captured heat underground for winter use. Urban runoff from paved surfaces can raise detention-pond and receiving-stream temperatures, creating ecological stress in temperature-sensitive aquatic systems. The proposed system extracts heat from pond water through a submerged high-density polyethylene heat exchanger connected to a water-glycol circulation loop, a water-source heat pump, and a borehole thermal-energy storage field. Heat extraction is constrained by a stream-protection requirement that limits pond discharge temperature to 21 °C. The modeled system represents a 1-hectare impervious catchment with a runoff coefficient of 0.9 draining to a 300 m² pond with a depth of 2 m. Design storms with 2-, 10-, and 50-year return periods are simulated using EPA SWMM and a pavement surface-energy balance. The resulting runoff hydrographs and temperature profiles are used as boundary conditions in a COMSOL Multiphysics model that couples turbulent non-isothermal flow, buoyancy-driven stratification, heat transfer, and groundwater movement. Simulations cover individual storm events and annual charging and discharging cycles. Mesh and time-step independence are assessed, and outputs are compared with monitored pond-temperature records and published borehole-storage data. A machine-learning surrogate model is trained on the multiphysics simulation results to support rapid design-space exploration and constrained multi-objective optimization. The model predicts warm-season heat capture of approximately 12–20 MWh thermal per hectare for a 5 °C temperature reduction. Seasonal heat recovery is projected at 40–50% during the first cycle and 60–75% after three to five cycles, corresponding to approximately 8–14 MWh of recoverable winter heat. Heat extraction reduces peak effluent temperature by 3–6 °C and thermal loading by 50–65%. The framework provides a basis for evaluating detention ponds as combined stormwater-cooling and seasonal energy-recovery infrastructure.

Keywords: Stormwater thermal pollution, low-grade heat recovery, seasonal thermal energy storage, detention pond, water-source heat pump, machine-learning surrogate, multi-objective optimization, water–energy nexus

https://icsest.org/accepted-papers.html#CA0132

Abstract: This study reports the synthesis and electrochemical evaluation of a CuS/CuO heterostructure produced through post-synthesis calcination for use in energy-storage electrodes. Copper sulfide is first prepared using a hydrothermal method and then calcined in air at 300 °C for one hour to induce partial oxidation and form an interface-engineered CuS/CuO structure. X-ray diffraction is used to identify the crystalline phases and confirm heterostructure formation. Field-emission scanning electron microscopy shows a rough and textured surface morphology, while X-ray photoelectron spectroscopy provides information about changes in copper and sulfur oxidation states after calcination. These structural and surface changes indicate that thermal treatment partially transforms CuS into CuO while preserving the underlying nanostructured framework. The electrochemical properties of the prepared material are evaluated in a three-electrode configuration using cyclic voltammetry, galvanostatic charge-discharge, and electrochemical impedance spectroscopy. The calcined CuS/CuO electrode delivers a reported specific capacitance of 866 F/g at a current density of 1.0 A/g. During long-term cycling, the electrode retains approximately 79% of its initial capacitance after 2,300 charge-discharge cycles and maintains a coulombic efficiency of 95.9%. The improved charge-storage behavior is attributed to the combined redox activity of CuS and CuO, increased surface roughness, additional electroactive sites, and enhanced ion-transport pathways across the heterostructure interface. The results indicate that controlled post-synthesis calcination can be used to tune the phase composition and electrochemical properties of copper-based electrode materials. The developed CuS/CuO heterostructure may have potential for supercapacitor and hybrid energy-storage applications. Further work should include comparison with uncalcined CuS and pure CuO electrodes, detailed impedance analysis, higher cycle counts, rate-capability testing, mass-loading information, and evaluation in a practical two-electrode device.

Keywords: Energy storage, copper sulfide, heterostructure, electrochemical performance, specific capacitance

https://icsest.org/accepted-papers.html#CA0130

Abstract: This study proposes a Blockchain-IoT architecture for peer-to-peer energy exchange in decentralized microgrids in Ghana. The study addresses challenges linked to grid instability, energy losses, limited renewable energy use, and the absence of a trusted local market for energy trading among prosumers and consumers. As Ghana expands mini-grid and rooftop solar systems, a practical mechanism is needed to manage local energy production, consumption, pricing, and settlement. The proposed framework combines IoT-enabled smart meters, an off-chain communication layer, and a permissioned blockchain system with smart contracts. The IoT layer records real-time energy generation and consumption data from connected users, while the off-chain layer aggregates and verifies meter readings before sending trusted transaction data to the blockchain. The blockchain layer stores transaction records, manages digital wallets, and executes smart contracts for bid matching, energy allocation, pricing, and payment settlement. A double-auction market model is used to match seller offers with buyer bids during fixed trading intervals. The smart contract prioritizes low-cost local energy supply while meeting consumer demand and applying a market clearing price for completed transactions. The study also presents implementation considerations, including platform selection, consensus mechanisms, smart contract deployment, communication protocols, scalability, cybersecurity, privacy, and regulatory governance. A simulation based on a theoretical Ghanaian microgrid scenario is used to test the proposed trading mechanism. The simulation shows that the model can execute automated peer-to-peer energy transactions, support partial order fulfillment, calculate clearing prices, and record settlement data through the blockchain ledger. The results indicate that a Blockchain-IoT framework can support transparent local energy trading, improve coordination between prosumers and consumers, reduce dependence on centralized intermediaries, and help manage surplus renewable energy within community microgrids. The study provides a practical architecture that can be further tested through pilot deployment in Ghana’s developing decentralized energy systems.

Keywords: Blockchain, Peer-to-Peer Energy Trading, Internet of Things, Microgrids, Smart Contracts, Energy, Decentralized System

https://icsest.org/accepted-papers.html#CA9998

Abstract: This study investigates the process air dew point temperature in a solar-powered atmospheric water harvesting system using a desiccant wheel. Atmospheric water harvesting is considered as a possible method for producing potable water in areas affected by freshwater scarcity, irregular rainfall, and limited access to conventional water sources. Existing water harvesting systems often face limitations because desiccant-bed systems have low water production rates, while vapor-compression systems require high electrical energy. This study examines a solar-regenerated desiccant wheel system that adsorbs moisture from ambient air and transfers it to a regeneration air stream, where the moisture can be condensed and collected as water. A numerical model of the rotary desiccant wheel is developed using heat and mass transfer equations and solved through a MATLAB-based finite difference approach. The model is validated using experimental data collected under real atmospheric conditions at Sardar Vallabhbhai National Institute of Technology, Surat. The study analyzes the effects of main operating and design parameters, including wheel rotational speed, wheel length, facial area ratio, adsorption air velocity, and regeneration air velocity, on the dew point temperature of the process air. The results show that the facial area ratio has a stronger effect on moisture transfer than wheel length. A facial area ratio of 3 improves adsorption performance by increasing the adsorption sector and reducing the regeneration sector, which increases the humidity ratio and dew point temperature of the outlet process air. The maximum dew point temperature of 42.4 °C is achieved with a 20 cm wheel length, facial area ratio of 3, wheel speed of 15 rev/h, regeneration air velocity of 2 m/s, and adsorption air velocity of 8 m/s. A 40 cm wheel length produces a similar maximum dew point temperature of 42.6 °C at lower rotational speed, while also reducing the dry-bulb temperature of the regenerated air. The findings show that a solar-regenerated desiccant wheel can improve the feasibility of atmospheric water harvesting by increasing the process air dew point temperature and reducing the need for high-energy cooling systems.

Keywords: Atmospheric water harvesting, Desiccant wheel, Solar air heater

https://icsest.org/accepted-papers.html#CA4386

Abstract: This study examines how Indonesia’s Universal Service Obligation program provides mobile connectivity in frontier, outermost, and underdeveloped regions where reliable electricity and terrestrial communication networks are not available. Conventional infrastructure planning usually assumes that electricity and fiber connections are established before mobile communication services are introduced. In many Indonesian islands, however, mobile internet infrastructure has been deployed before the national electricity grid or terrestrial backbone reaches the area. The study defines the 4G base station used in these locations as a self-provisioning node with two forms of operational autonomy. On the energy side, the site uses solar photovoltaic generation and battery storage, with hybrid support where needed, to operate without waiting for grid extension. On the communication side, satellite backhaul is used when fiber or microwave connections are unavailable. This combination allows connectivity services to be delivered in locations where separate energy and communication market failures would otherwise delay investment. The study uses a single-case research design based on primary operational and technical data from the Telecommunication and Information Accessibility Agency under Indonesia’s Ministry of Communication and Digital Affairs. The analysis considers site power arrangements, battery autonomy, backhaul options, operating requirements, and the financial role of the Universal Service Obligation mechanism. The findings indicate that most remote sites rely on solar or solar-hybrid systems, while fully off-grid locations use battery storage designed for several days of operation. The model can also reduce operating costs compared with diesel-based power supply, especially in areas where fuel transport is difficult. The study argues that a single public-service funding instrument can address both the energy and connectivity gaps by including local power generation within the communication infrastructure project. The Indonesian approach is compared with operator-led tower solarization in Africa and electrification-first planning in the Philippines. The framework may support future planning for remote digital infrastructure in countries where conventional grid and backbone expansion is slow or commercially unviable.

Keywords: Dual-axis autonomy, self-provisioning infrastructure, solar microgrid, satellite backhaul, universal service obligation

https://icsest.org/accepted-papers.html#CA0119

Abstract: This study proposes the integration of machine learning (ML) models with Computational Fluid Dynamics (CFD) to improve the efficiency of thermal system design optimization. Thermal systems, including heat exchangers, cooling systems, and energy conversion equipment, are commonly evaluated using energy, exergy, economic analysis, and CFD-based simulation. Although CFD provides high-fidelity performance prediction, it is computationally expensive when repeated simulations are required across wide design parameter ranges. To address this limitation, the study develops a hybrid ML–CFD framework in which high-fidelity CFD simulations are first used to generate a dataset linking design variables, flow parameters, and thermal performance outputs. Surrogate ML models, including regression-based ensemble methods and artificial neural networks, are then trained to estimate CFD results with reduced computational cost. The trained models are integrated into an optimization framework to replace repeated CFD simulations while maintaining acceptable prediction accuracy. Model performance is evaluated using mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R²). The findings indicate that ML-enabled CFD can significantly reduce computation time while preserving reliable prediction accuracy. The proposed framework offers a scalable approach for thermal system optimization and has potential applications in real-time design, engineering decision support, and future digital-twin development for energy and thermal systems.

Keywords: Machine Learning, surrogate modelling, heat transfer, Computational Fluid Dynamics, thermal systems, design optimization

https://icsest.org/accepted-papers.html#CA0106

Abstract: This study introduces Algorithmic Carbon Bias as a quantitative framework for measuring the greenhouse gas emissions caused by machine decision systems. Automated systems used in logistics, e-commerce, energy management, and resource allocation are often designed to improve cost, speed, or efficiency without directly considering the emissions created by their decisions. As a result, some algorithms may increase physical-world carbon emissions even when the computing process itself appears efficient. The proposed framework measures the difference between the emissions produced by an algorithm’s selected decision and the emissions of a sustainability-weighted alternative operating under the same practical constraints. This difference is expressed through an Algorithmic Carbon Bias coefficient that represents the average carbon effect per decision. The coefficient can be estimated using observable decision records and public emission-intensity data without requiring access to proprietary source code or internal model parameters. The framework is also included in a multi-objective optimization model with a carbon penalty term, allowing the environmental cost of algorithmic decisions to be expressed as a shadow carbon price. The study evaluates the method through agent-based simulations of urban last-mile delivery networks and counterfactual ranking analysis of e-commerce recommendation systems. Product lifecycle emissions, grid carbon intensity, traffic conditions, and demand uncertainty are included in the analysis. Monte Carlo simulation is used to represent uncertainty, while the Generalized Method of Moments is applied to estimate structural parameters. The results show that optimization based only on economic performance can produce an emission premium of approximately 12–22% compared with carbon-weighted alternatives. Reducing this bias is associated with an operational efficiency cost below 2%, indicating that emission reductions may be achieved with limited loss of operational performance. The study also suggests that including carbon constraints during algorithm design can reduce cumulative emissions more effectively than applying corrective measures after deployment. The framework can support future work in carbon-aware logistics, recommendation systems, smart grid demand response, and renewable energy dispatch.

Keywords: Algorithmic carbon bias, emission externality, AI governance, carbon-aware computing, last-mile delivery, multi-objective optimization, GHG Protocol, sustainable energy

https://icsest.org/accepted-papers.html#CA0117

Abstract: This study evaluates the implementation of an ISO 50001:2018 energy management system in a corporate building located in Asunción, Paraguay, where increasing commercial electricity demand poses potential challenges for national energy supply despite the country’s abundant hydroelectric resources. The research aims to assess energy performance, identify significant energy uses, and propose practical efficiency improvements using a structured methodology aligned with ISO 50001 principles. The approach combines on-site energy audits, including equipment inventory, luxometric measurements, and staff interviews with historical electricity billing data correlated with monthly temperature records to establish a climate-sensitive energy baseline. The analysis identifies air conditioning and lighting as the major contributors to energy consumption, representing approximately 63% and 17% of total usage, respectively. Several improvement measures are evaluated, including LED lighting retrofits, occupancy sensor installation, HVAC temperature setpoint optimization at 24 °C, and tariff restructuring. Financial analysis indicates that LED lighting upgrades alone achieve an internal rate of return (IRR) of 20.18%. The implementation of all recommended measures could reduce annual electricity consumption by approximately 51.95%, generating estimated yearly savings of 12.72 million Paraguayan Guaraníes. Additionally, the building’s thermal sensitivity is projected to decrease by 27.3%, with the energy performance indicator improving from 267.26 to 194.38 kWh/°C. The findings demonstrate that structured energy management systems can significantly enhance operational efficiency and climate resilience in commercial buildings within emerging economies.

Keywords: ISO 50001, Energy Management, Corporate Building, Energy Efficiency, Paraguay

https://icsest.org/accepted-papers.html#CA7342

Abstract: This study examines how green finance, trade openness, carbon pricing, technological innovation, and digital readiness influence sustainable energy transitions and business model development in emerging and transitioning economies. The analysis covers 62 economies from 2010 to 2024 and uses country-level information from international energy, finance, development, and policy databases. A composite Green Finance Energy Business Model Index is developed to measure the combined progress of green finance, renewable energy use, innovation, institutional quality, and sustainable business practices. The study applies panel regression methods, including a two-step system generalized method of moments estimator, to address possible endogeneity and differences across countries. It also uses threshold analysis to examine whether the effect of carbon pricing changes at different price levels. The findings indicate that greater green finance depth, measured through green bond issuance relative to gross domestic product, is associated with higher index scores and stronger sustainable energy outcomes. Trade openness strengthens this relationship by supporting technology transfer and access to low-carbon equipment. Digital readiness also improves the effect of green finance by supporting financial technology, data access, and new energy-service models. The results further suggest that carbon pricing has a non-linear effect, with positive outcomes increasing up to an estimated threshold before the marginal effect begins to weaken. Income-group analysis shows that green finance has a larger effect in upper-middle-income economies than in lower-income economies, which indicates that institutional capacity and financial market development influence policy outcomes. Case examples from China, the European Union, and Sub-Saharan Africa are used to explain how green finance policies, disclosure rules, mobile payment systems, and off-grid solar models operate under different institutional conditions. The study concludes that green finance is more effective when it is supported by credible standards, trade access, digital infrastructure, carbon-pricing rules, and institutional capacity.

Keywords: Green finance, energy markets, sustainable business models, emerging economies, carbon pricing

https://icsest.org/accepted-papers.html#CA9270

Abstract: This study investigates the optimization of bio-ethanol supply chains and commercial distribution networks in Sub-Saharan Africa, with a focus on localized logistics and regulatory compliance in the Democratic Republic of the Congo. The distribution of industrial-grade bio-ethanol in emerging markets faces several operational challenges, including transport limitations, regulatory requirements, market coordination, and supply continuity. This study examines the case of ETS Zola Business, a localized enterprise operating in Kinshasa, to understand how formal business registration, supply agreements, and regional logistics practices can support the distribution of bio-based commodities. The study uses field information from a 20,000-liter bulk acquisition and distribution model of 96° industrial ethanol sourced through national biotechnology partnerships. It analyzes how corporate identifiers, including Id. Nat, NIF, and RCCM registrations, support formal business-to-business market participation, reduce transaction barriers, and improve supply-line control. The study also considers the logistical conditions of urban transport hubs in Kinshasa, especially in the N’djili municipality, where infrastructure and last-mile delivery issues affect the movement of industrial bio-based chemicals. The proposed analysis focuses on the relationship between regulatory alignment, supply-chain organization, commercial distribution, and cost management. The study suggests that stronger formal documentation, structured supply agreements, digital fiscal records, and improved logistics coordination can help local distributors reduce supply shortages and improve market access. It also discusses the potential role of smart logistics tracing tools and demand-side analytics in improving decentralized distribution systems in areas with unstable infrastructure. The study contributes to the discussion of sustainable energy and bio-based commodity markets by showing how local enterprises can support bio-economy development through organized logistics, regulatory compliance, and practical distribution planning.

Keywords: Bio-ethanol Distribution, Supply Chain Optimization, Sub-Saharan Logistics, Regulatory Compliance, Emerging Bio-economies, Commercial Retail Management

https://icsest.org/accepted-papers.html#CA6452

Abstract: This study examines the social, economic, and policy factors influencing the adoption of solar-powered farming systems among smallholder farmers in Southeast Nigeria. Rural agricultural communities in the region continue to face unreliable electricity supply, high energy costs, limited irrigation capacity, and post-harvest losses that reduce productivity and household income. Solar technologies may provide decentralized energy for irrigation pumps, crop processing, cold storage, lighting, water supply, and other agricultural activities. The study is guided by the Diffusion of Innovation Theory and the Sustainable Livelihoods Framework and adopts a quantitative survey design involving smallholder farmers across selected states in Southeast Nigeria. The analysis focuses on the relationship between solar energy adoption and farm productivity, production costs, household income, food security, and livelihood stability. It also examines the influence of initial investment cost, access to credit, technical knowledge, perceived usefulness, extension support, maintenance services, and government policy on farmers' adoption decisions. The study argues that solar-powered agricultural systems can reduce dependence on diesel generators and unreliable grid electricity while supporting more efficient irrigation, processing, and storage. However, widespread adoption remains constrained by high purchase costs, limited financing options, inadequate technical support, weak awareness, and inconsistent renewable energy policies. The study recommends targeted credit programs, equipment subsidies, cooperative ownership models, farmer training, stronger extension services, local maintenance capacity, and clearer policy support for agricultural solar applications. These measures may improve affordability and reduce the technical and financial risks faced by rural households. The findings are expected to support policymakers, agricultural agencies, financial institutions, renewable energy providers, and farming organizations in designing programs that connect clean energy access with agricultural development. Further research should provide detailed survey results, compare adopters and non-adopters, estimate economic returns, and examine differences across farm types, income groups, gender, and location.

Keywords: Renewable energy, sustainable solar energy, agricultural productivity, energy policy, sustainable livelihoods

https://icsest.org/accepted-papers.html#CA0128

Abstract: This study investigates the role of carbon footprint reduction, social acceptance, and climate resilience in supporting sustainable energy transitions. Energy production and consumption remain major sources of greenhouse gas emissions, and many countries are working to move from carbon-intensive systems toward renewable and low-carbon energy models. However, the shift to sustainable energy systems depends not only on technology but also on public acceptance, institutional planning, cultural conditions, and fair access to energy resources. This study examines how social, environmental, institutional, and cultural factors influence the acceptance and implementation of low-carbon energy systems. It focuses on public perceptions of renewable and low-carbon technologies, equity and inclusion in energy access, environmental effects of carbon-intensive development, and the role of climate resilience in long-term energy planning. The study is based on Sustainable Development Theory, Social Acceptance Theory, and Socio-Technical Transition Theory to explain the relationship between technological change, policy systems, and social behavior. A mixed-method approach is proposed, including policy analysis, stakeholder interviews, and secondary data review. Through this approach, the study aims to identify how inclusive policies, community participation, and responsible governance can support carbon reduction and improve the use of sustainable energy systems. The study also examines the challenges that may limit public acceptance, such as unequal access, lack of institutional readiness, weak public involvement, and resistance to changes in existing energy systems. The expected findings suggest that low-carbon energy transitions require a balanced approach that combines technological innovation with social inclusion, environmental responsibility, and participatory decision-making. The study contributes to the discussion on sustainable energy transitions by showing that climate resilience and carbon reduction should be planned together with public needs, social justice, and institutional capacity.

Keywords: Carbon Footprint, Sustainable Energy Transitions, Climate Resilience, Social Acceptance, Environmental Sustainability

https://icsest.org/accepted-papers.html#CA5322

Abstract: This study investigates the use of the PlanetSustain Mobile Application among Indian students as a smart technology tool for supporting sustainable energy practices, environmental awareness, and climate action education. Sustainable energy behavior and responsible lifestyles have become important issues because of climate change, ecological damage, and increasing energy demand. In India, the large number of internet users and the growing use of mobile applications provide an opportunity to use digital platforms for sustainability education among students. This study focuses on how mobile-based sustainability tools can help students understand carbon emissions, energy-saving actions, recycling practices, and responsible consumption. The PlanetSustain Mobile Application allows users to track Scope 1, Scope 2, and Scope 3 carbon emissions and take part in climate-related activities through a digital interface. The study examines the awareness, usage, and perceived usefulness of this application among students in selected colleges and universities in India. It is based on the Technology Acceptance Model and the Theory of Planned Behavior, which explain how users accept technology through perceived usefulness, behavioral intention, and personal attitudes. The study also connects with Sustainable Development Goal 7 by addressing clean energy awareness and smart digital participation. A quantitative survey design is used, with structured questionnaires collected from an estimated sample of 250 students through purposive sampling. Descriptive statistics, correlation analysis, and regression analysis are proposed to examine the relationship between mobile application usage, environmental awareness, and sustainable energy behavior. The study expects to show how smart mobile applications can support student engagement in sustainability practices when they are easy to use, useful, and connected to daily behavior. The findings may help educators, policymakers, sustainability practitioners, and application developers design digital tools that support clean energy education, carbon awareness, and climate-responsible behavior among young adults.

Keywords: Sustainable Energy, Smart Technologies, Mobile Application Usage, Environmental Awareness, Climate Action Education

https://icsest.org/accepted-papers.html#CA0372

Abstract: This study examines whether Canada’s adoption of the United Nations Declaration on the Rights of Indigenous Peoples Act in 2021 has led to a meaningful shift from consultation toward Indigenous co-governance in federal renewable energy policy. The analysis focuses on whether legal commitments to Free, Prior and Informed Consent, self-determination, and partnership have been translated into binding decision-making authority. A qualitative policy analysis is conducted using nine purposively selected federal renewable energy policy instruments issued between 2016 and 2025. The documents are coded through framework analysis using a five-level Governance Authority Spectrum adapted from Arnstein’s ladder of citizen participation. The spectrum distinguishes between information sharing, consultation, advisory participation, shared decision-making, and binding co-governance. Policies adopted before and after 2021 are compared to identify changes in how Indigenous communities are categorized, how authority is allocated, and whether consent or co-decision mechanisms are formally established. The findings indicate that post-2021 policy instruments contain more frequent references to engagement, partnership, co-development, and financial inclusion. However, decision-making authority remains primarily state-controlled. Indigenous communities are commonly described as stakeholders, partners, or interested parties, but their formal role generally remains consultative or advisory. None of the reviewed instruments clearly establishes binding co-decision authority across the policy cycle. The analysis therefore identifies a gap between the language of inclusion and the institutional transfer of governance power. The study argues that financial participation, project ownership opportunities, and expanded consultation are important but do not by themselves constitute co-governance. Meaningful implementation of UNDRIP in the renewable energy sector requires clearly defined consent procedures, shared regulatory authority, enforceable dispute-resolution mechanisms, and Indigenous participation in agenda setting, approval, implementation, and monitoring. The findings may support policymakers, regulators, Indigenous governments, energy developers, and researchers working to strengthen rights-based governance in Canada’s clean energy transition.

Keywords: UNDRIP, Indigenous participation, Free, Prior and Informed Consent, Canada, clean energy

https://icsest.org/accepted-papers.html#CA0129

Abstract: This review develops an integrated carbon-farming framework for Punjab’s rice–wheat production system by combining Direct-Seeded Rice, Alternate Wetting and Drying, and biochar production from paddy straw. Punjab’s agricultural sector faces interconnected challenges that include groundwater depletion, methane emissions from continuously flooded rice fields, open-field residue burning, declining soil quality, and limited access by small and medium-sized farmers to carbon markets. The proposed framework examines how the three practices can be implemented together to reduce greenhouse gas emissions, improve water-use efficiency, increase stable soil carbon storage, and generate additional farm income. Direct-Seeded Rice can reduce irrigation demand and field-preparation costs, while Alternate Wetting and Drying can lower methane emissions by periodically allowing rice fields to drain. Converting paddy straw into biochar provides an alternative to open burning and may support long-term carbon sequestration, improved soil properties, and nutrient retention. The review synthesizes evidence concerning methane reduction, soil carbon storage, water conservation, production costs, crop performance, carbon-credit methodologies, and Monitoring, Reporting, and Verification requirements. It also proposes a digital MRV structure using geotagged farm records, remote sensing, water-level monitoring, mobile reporting, and periodic field verification. Farmer aggregation through cooperatives, producer organizations, or project developers is identified as necessary to reduce transaction costs and enable participation in voluntary carbon markets. The framework further emphasizes financial incentives, extension support, equipment access, biochar quality standards, third-party verification, and safeguards against double counting. The combined approach may provide greater environmental and economic value than isolated adoption because it addresses field emissions, residue management, soil carbon, groundwater use, and market access within one coordinated program. However, the scale of benefits will depend on local soil conditions, irrigation practices, biochar production methods, carbon prices, verification costs, farmer participation, and policy stability. Pilot projects are recommended to quantify net emissions, crop yields, water savings, implementation costs, carbon-credit revenue, and farmer income under Punjab conditions.

Keywords: Alternate Wetting and Drying, biochar, carbon credits, carbon sequestration, climate-smart agriculture, Direct-Seeded Rice

https://icsest.org/accepted-papers.html#CA0135

Abstract: This study focuses on the design and development of a solar-powered automated waste segregation robot with a backup battery for Lorma Colleges Basic and Inclusive Education Schools. The study addresses the need for a simple and practical system that can support proper waste segregation in a school setting while promoting the use of renewable energy and automation. The proposed robot is designed to classify waste into plastic, paper, and organic categories using sensor-based detection, servo motor actuation, and an Arduino Uno microcontroller. The system is powered by a 6V solar panel and supported by a rechargeable lithium-ion battery to allow continued operation during low-sunlight conditions. A design and development research approach was used, supported by the waterfall model to guide the stages of planning, design, fabrication, testing, and evaluation. The prototype was developed to perform automatic waste detection, classification, and sorting with reduced need for manual handling. Its performance was evaluated based on waste detection accuracy, classification accuracy, sorting accuracy, energy use, and operational reliability. The results showed that the robot achieved 92% waste detection accuracy, 84% classification accuracy, and 76% sorting accuracy. These results indicate that the prototype can identify and separate common school waste materials with acceptable performance for a small-scale educational environment. The integration of solar power and battery backup also supports the system’s use as a sustainable and low-energy solution for school waste management. The study shows that automated waste segregation can help reduce manual sorting, improve waste handling practices, and support environmental awareness among students and school personnel. The proposed system can serve as a working prototype for future improvement, including better sensor calibration, larger waste capacity, improved sorting accuracy, and expanded testing under different school waste conditions.

Keywords: Arduino Uno, automated waste segregation, design and development research, prototype evaluation, renewable energy, solar-powered robot

https://icsest.org/accepted-papers.html#CA0107

Abstract: This study proposes the development of bio-intelligent concrete panels for self-sustaining air purification through the integration of microalgae, porous concrete surfaces, water circulation, and environmental monitoring sensors. The study responds to the need for construction materials that can support cleaner urban environments while reducing the environmental effects of buildings. Conventional concrete is mainly used as a structural material and does not actively improve air quality. In contrast, the proposed panel is designed as a living construction element that provides a surface for microalgae growth and supports carbon dioxide absorption through photosynthesis. The panel includes a bio-receptive porous concrete layer that allows microalgae to attach and grow under suitable moisture and light conditions. A simple water and nutrient circulation system is proposed to maintain algae activity and support continuous biological function. The microalgae absorb carbon dioxide from the surrounding air and use sunlight and water to produce oxygen and biomass. This process can help reduce local carbon dioxide concentration and improve air quality near building façades, public walls, and urban infrastructure. The proposed system also includes smart sensors to monitor carbon dioxide levels, temperature, humidity, surface moisture, and light intensity. These sensor readings can be used to observe the condition of the algae layer and support maintenance decisions. The study describes the main components of the panel, including bio-receptive concrete, the microalgae layer, the water circulation system, the nutrient delivery system, smart sensors, and a data processing unit. The concept can be applied to building envelopes, smart city infrastructure, and green construction systems where environmental performance is required in addition to structural function. The study suggests that algae-based concrete panels can be developed as active environmental materials for urban air purification and carbon sequestration. Further experimental work is needed to test algae growth, carbon dioxide absorption rate, oxygen generation, durability, water use, maintenance needs, and structural performance under real outdoor conditions.

Keywords: Carbon sequestration, sustainable smart material, algae-based construction, urban air purification, environmental sustainability

https://icsest.org/accepted-papers.html#CA0112

Abstract: This study develops a multi-view similarity framework for identifying cities whose air-quality models and data practices may be transferred to Monterrey, Mexico. Air pollution research and machine-learning applications remain unevenly distributed across regions, and Latin American cities are often underrepresented in the development of predictive air-quality tools. The proposed method addresses this gap by identifying structurally and environmentally comparable metropolitan areas with stronger open-data and modeling resources. A globally filtered cohort of more than 600 subtropical cities is assembled using Google Earth Engine and other geospatial data sources. The feature set includes three-dimensional urban morphology, topographic roughness, long-term climate conditions, and sector-specific emissions indicators. Variables are examined for correlation, scaled according to their statistical distributions, and combined in a custom weighted K-nearest-neighbors framework centered on Monterrey. To reduce dependence on a single weighting scheme, a multi-view sensitivity analysis is performed across seven weight configurations that assign different priorities to urban form, topography, climate, and emissions. Rank-ablation analysis is also used to determine which feature groups are most important for stable city matching. The results identify Santiago, Chile, as the most consistent structural analog, with an average rank of 1.6 and the top similarity position in five of the seven weighting scenarios. The analysis shows that mountain-basin topography is a major contributor to this result and that removing it causes substantial instability in the ranking. A practical viability filter is then applied to the broader similarity matrix, requiring open meteorological and pollution data, published reproducible machine-learning methods, and ranking stability across weighted and unweighted analyses. This process identifies Santiago, Los Angeles, Baoji, Dehradun, and Chandigarh as the five most suitable transfer-learning candidates. The framework offers a transparent way to support cross-city model transfer in data-limited settings, although future work should test actual transfer performance using Monterrey observations and compare transferred models with locally trained baselines.

Keywords: KNN similarity framework, air quality analysis, weather and pollution analysis, transfer learning, feature selection metrics

https://icsest.org/accepted-papers.html#CA0133

Abstract: This study examines flood-related environmental and infrastructure impacts in Pakistan's Indus River Basin and evaluates gaps in flood forecasting and early warning systems. The analysis focuses on the extent to which technical forecasting capacity is converted into timely and effective community-level action. A mixed-method approach combines secondary information from national and provincial disaster-management authorities with household surveys and focus group discussions conducted in selected flood-prone communities. Geographic information system mapping is used to integrate flood extent, population exposure, infrastructure vulnerability, and energy-system impacts. The study also evaluates radar and early warning performance across different geographic settings using indicators related to coverage, forecast accuracy, warning lead time, communication, and public response. The findings indicate that repeated flooding has damaged hydropower facilities, electricity transmission systems, housing, transportation networks, agricultural land, and community services. Although radar and forecasting infrastructure has improved in some urban and accessible areas, remote communities in northern and mountainous regions continue to experience limited coverage and inadequate warning access. Institutional fragmentation among federal, provincial, and district authorities further weakens the translation of forecasts into evacuation, preparedness, and emergency response. The results suggest that climate resilience depends not only on forecasting technology but also on communication systems, community trust, local preparedness, institutional coordination, and inclusive planning. The study recommends expanding radar and monitoring coverage, strengthening mobile and community-based warning channels, developing clear interagency responsibilities, conducting regular preparedness exercises, and integrating local knowledge into emergency planning. It also identifies mangrove restoration in the Indus Delta as a complementary nature-based approach for reducing coastal vulnerability and strengthening ecosystem resilience. The proposed framework may assist disaster-management agencies, infrastructure planners, energy authorities, and local governments in improving flood-risk reduction. Further research should provide detailed survey results, location-specific performance measures, infrastructure-loss estimates, and comparisons among urban, rural, mountainous, and delta communities.

Keywords: Environmental impact, climate resilience, Pakistan floods, early warning systems, Indus River Basin, nature-based solutions

https://icsest.org/accepted-papers.html#CA0138

Abstract: This study presents AI-KneeInsight, a cloud-based explainable artificial intelligence prototype designed to support preliminary assessment of knee osteoarthritis severity from radiographic images. Knee osteoarthritis is a common musculoskeletal condition associated with pain, reduced mobility, and disability. Conventional radiographic assessment can require specialist expertise and may be affected by differences in interpretation. The proposed system is intended to provide a structured decision-support workflow in which users upload knee radiographs and receive an automated severity classification, confidence estimate, probability display, and plain-language summary. The prototype is developed using Python, Gradio, and Hugging Face Spaces and includes image upload, preprocessing, model inference, confidence visualization, and report generation components. The system architecture is designed to support future integration of a validated deep learning model trained on labelled knee radiographs and evaluated against expert clinical assessments. Explainability features are included to help users understand the basis and uncertainty of the predicted severity category rather than receiving only a single classification label. The present work demonstrates the technical feasibility of combining medical image processing, machine learning, cloud deployment, and structured reporting within an accessible interface. However, the current prototype should be regarded as a research and demonstration tool rather than a clinically validated diagnostic system. Further evaluation is required using a clearly described radiographic dataset, an established osteoarthritis grading standard, independent test data, and appropriate performance measures such as sensitivity, specificity, precision, F1-score, area under the receiver operating characteristic curve, and inter-rater agreement with clinical experts. Future work should also assess model fairness, image-quality effects, privacy protection, cybersecurity, usability, and the risk of overreliance on automated outputs. With appropriate validation and governance, the proposed framework may support research on explainable AI, musculoskeletal imaging, and preliminary clinical decision support.

Keywords: Artificial intelligence, knee osteoarthritis, medical imaging, explainable AI, clinical decision support, deep learning, digital health, musculoskeletal imaging

https://icsest.org/accepted-papers.html#CA9203

Abstract: This study examines the use of machine learning to predict the post-listing performance of Canadian initial public offerings and proposes an AI-enabled decision-support system for entrepreneurial finance. Predicting IPO performance is difficult because newly listed firms have limited market histories and may experience rapid changes in price, trading volume, and investor attention. The study uses historical IPO and market data collected from publicly available financial sources, including Yahoo Finance and SwingTradeBot.com. The dataset covers Canadian IPOs listed during 2024 and 2025 and includes closing price, trading volume, percentage price change, volume surge, and the number of days since listing. The data preparation process includes cleaning, normalization, feature selection, and the construction of performance categories. A K-Nearest Neighbours classification model is developed in Python to identify IPO performance patterns from the selected market indicators. Power BI is used for exploratory analysis and data visualization, while a Streamlit prototype is developed to present model outputs through an accessible web interface. The reported results indicate that the K-Nearest Neighbours model achieved an accuracy of approximately 95.24% on the evaluated dataset. The prototype shows how predictive results can be presented to investors, analysts, and entrepreneurs through a simple decision-support environment. The proposed framework also allows future integration with financial data APIs to support automated data collection, continuous monitoring, and periodic model updating. The study suggests that machine learning can assist with the analysis of IPO market behavior when the prediction target, observation period, and validation method are clearly defined. However, the model should be tested using a larger sample, time-based validation, class-level performance measures, and comparisons with other statistical and machine learning methods. The resulting system should be treated as an analytical support tool rather than a substitute for financial judgment, risk assessment, or professional investment advice.

Keywords: Initial public offerings, machine learning, K-Nearest Neighbours, entrepreneurial finance, financial forecasting, decision-support system

https://icsest.org/accepted-papers.html#CA0123

Abstract: This study investigates the impact of structural breaks on forecasting performance in high-dimensional factor models commonly used in macroeconomic analysis. Factor models are widely employed to extract latent common factors from large datasets; however, their predictive reliability may deteriorate when structural changes occur in factor loadings. The research examines how detecting and addressing structural breaks can improve forecasting accuracy in such models. The study employs econometric techniques including principal component analysis for factor estimation and bootstrap-based structural break tests to identify instability in factor loadings. Using a dataset of 169 quarterly macroeconomic indicators from Germany and the United States covering the period 1993–2013, several model specifications are evaluated through pseudo out-of-sample forecasting exercises. Forecast accuracy is measured using root mean squared error (RMSE). The results show that removing variables exhibiting structural breaks significantly improves forecasting performance, reducing RMSE from 0.47 in baseline dynamic factor models to approximately 0.40 after break adjustment. The findings demonstrate that structural break detection combined with variable preselection provides a practical strategy for improving forecasting robustness in high-dimensional macroeconomic models. The study contributes to econometric forecasting literature by highlighting the importance of parameter stability when constructing predictive models based on large datasets.

Keywords: Factor models, structural breaks, forecasting, high-dimensional data, variable selection, macroeconomic prediction

https://icsest.org/accepted-papers.html#CA4190

Abstract: This study develops a supervised learning framework for transforming textual information from corporate 10-K filings into quantitative financial signals capable of predicting equity returns and volatility. The research focuses on Item 1A risk factor disclosures, which contain forward-looking information about firm-specific risks and operational uncertainties. An end-to-end processing pipeline is designed to acquire filings from the SEC EDGAR database, extract relevant risk sections, preprocess financial text, and construct document-level sentiment scores through a supervised lexicon learning model. Unlike dictionary-based approaches, the proposed model learns token weights directly from historical relationships between textual features, equity returns, and realized volatility. The methodology incorporates signal stabilization using Kalman filtering to reduce noise and enhance temporal consistency of the generated factors. Empirical analysis is conducted on more than 15,000 firm-year observations from S&P 500 companies covering the period 2000–2023. The results demonstrate that the supervised risk-lex factors provide significant predictive power for both cross-sectional stock returns and future volatility. Firms with high factor scores outperform those with low scores, while the signals also predict firm-specific volatility increases. The study shows that supervised financial text mining can generate interpretable and computationally efficient investment signals, offering practical applications for portfolio construction, risk monitoring, and systematic investment strategies.

Keywords: Text mining, financial sentiment, 10-K filings, risk factors, equity returns, volatility forecasting, supervised learning

https://icsest.org/accepted-papers.html#CA4831

Abstract: This study examines election forecasting through the lens of behavioral finance by comparing prediction market prices with probabilities generated by quantitative forecasting models. The research analyzes daily probability estimates from the PredictIt prediction market and the FiveThirtyEight forecasting model across 115 congressional races during the final 90 days of the 2018 United States midterm election cycle. A panel dataset of matched daily observations is constructed to evaluate forecasting performance using multiple metrics, including binary classification accuracy, Brier scores for probabilistic forecasting skill, and calibration diagnostics. The empirical results indicate that prediction markets exhibit higher binary classification accuracy, correctly identifying winners in approximately 86% of cases compared to about 84% for the forecasting model. However, when evaluated using proper probabilistic scoring rules, both methods demonstrate statistically indistinguishable forecasting skill. Calibration analysis further reveals systematic biases in prediction market prices, including underestimation of Democratic candidates and incumbents, consistent with behavioral finance theories related to trader composition and limits to arbitrage. These findings suggest that prediction markets function as behaviorally mediated information aggregates rather than perfectly efficient probability estimators. The study highlights the potential benefits of integrating prediction market signals with quantitative forecasting models to improve decision-making and forecasting robustness.

Keywords: Prediction markets, election forecasting, behavioral finance, FiveThirtyEight, PredictIt, market efficiency

https://icsest.org/accepted-papers.html#CA6329

Abstract: This study develops a smart cane that combines ultrasonic sensing, water-level detection, computer vision, and location-tracking technologies to support safer and more independent mobility for people with visual impairments. Conventional smart canes commonly rely on distance sensors that can detect nearby obstacles but may provide limited information about the type, position, or characteristics of an object. The proposed system integrates ultrasonic sensors for detecting obstacles at different heights, a water-level sensor for identifying wet surfaces, and a computer vision module for recognizing objects in real time. A pretrained YOLOv5 object-detection model is implemented on a Raspberry Pi 4 installed within the cane. The system is designed to identify common objects encountered during walking and provide immediate information that may assist the user in selecting a safer path. A Neo-6M GPS module and GSM/SIM900A communication module are also incorporated to support location tracking and allow text messages containing the user’s location to be sent to relatives or caregivers during emergencies. The electronic circuit is developed using Proteus software, while selected mechanical components are modeled in SolidWorks 2021 and manufactured using three-dimensional printing. The reported object-detection results include precision values of 0.87, 0.91, and 0.91 and recall values of 0.91, 0.69, and 0.91 at different confidence levels. The system also achieves a mean average precision of 0.96 at an intersection-over-union threshold of 0.5 and 0.83 across thresholds from 0.5 to 0.95. These results indicate that combining distance sensing and computer vision can improve the amount of environmental information available to smart-cane users. The proposed device may support obstacle detection, wet-surface identification, emergency communication, and outdoor navigation. Further work should include testing with visually impaired participants, evaluation under different lighting and weather conditions, measurement of detection latency and battery life, and assessment of comfort, reliability, and user acceptance.

Keywords: Sensors, actuators, object detection, accuracy, computer vision integration, smart cane, YOLOv5, safety, visually impaired

https://icsest.org/accepted-papers.html#CA7344

Abstract: This study develops a fuzzy logic–based proportional–integral–derivative controller for improving the trajectory tracking performance of a quadrotor unmanned aerial vehicle. The proposed control structure combines a conventional PID controller with a Mamdani fuzzy inference system that adjusts the proportional, integral, and derivative gains in response to the current tracking error and the rate of change of error. A nonlinear quadrotor model based on Newton–Euler equations is used to represent translational and rotational dynamics. The control system is organized in a cascaded form with an outer position loop and an inner attitude loop. The position controller generates the desired roll and pitch commands, while the attitude controller regulates roll, pitch, yaw, and altitude. The fuzzy system uses seven linguistic terms for each input and output and applies 49 rules to tune the PID gains during operation. Simulation experiments are carried out in MATLAB/Simulink using constant and repeating reference trajectories under undisturbed and disturbed conditions. The proposed fuzzy PID controller is compared with a manually tuned conventional PID controller using tracking plots and performance indices, including the Integral of Absolute Error, Integral of Squared Error, and Root Mean Squared Error. The reported results indicate that the fuzzy PID controller improves several position and attitude responses and reduces selected tracking-error measures under external disturbance. The largest reported improvements occur in altitude, roll, and pitch tracking, while some performance measures for the horizontal position and yaw channels show mixed results. These findings indicate that adaptive gain adjustment can improve the response of a nonlinear quadrotor system, but the controller should be evaluated more systematically across different trajectories, disturbance levels, parameter changes, and actuator constraints. The study provides a simulation-based control framework that may support future work on robust quadrotor navigation, intelligent flight control, and real-time experimental validation.

Keywords: Quadrotor, trajectory tracking, fuzzy logic controller, fuzzy PID, UAV control

https://icsest.org/accepted-papers.html#CA7003

Abstract: This study develops an adaptive fuzzy proportional–integral–derivative controller for improving trajectory tracking, motion smoothness, and energy efficiency in a nonlinear quadrotor unmanned aerial vehicle operating under external disturbances and parameter uncertainty. Quadrotor systems are nonlinear, underactuated, and strongly coupled, which makes their performance sensitive to changes in mass, inertia, mechanical vibration, and environmental disturbances. A six-degree-of-freedom nonlinear model based on the Newton–Euler formulation is used to describe the translational and rotational dynamics of the vehicle. The proposed controller combines a conventional PID structure with an online fuzzy inference mechanism that adjusts the proportional, integral, and derivative gains according to the instantaneous tracking error and change in error. This adaptive structure is intended to overcome the limitations of fixed-gain controllers when operating conditions vary. The controller is evaluated through simulation under nominal conditions and under parameter variations involving changes in mass and inertia. External disturbances are also introduced to examine robustness during trajectory tracking. Performance is compared with a conventional PID controller using trajectory error, transient overshoot, acceleration variation, mechanical jerk, and estimated power consumption. The reported results indicate that the adaptive fuzzy PID controller reduces transient overshoot from approximately 77% to 0.02% under the stated test conditions. It also improves motion smoothness by reducing acceleration fluctuations and jerk during manoeuvring. The controller produces an estimated energy-efficiency improvement of approximately 3.07% during nominal flight. These findings suggest that online fuzzy gain adjustment can improve tracking accuracy and reduce mechanical stress under changing operating conditions. The proposed approach may support UAV applications that require stable flight, precise trajectory tracking, reduced actuator activity, and longer operational endurance. Further work should include hardware-in-the-loop testing, experimental flight validation, actuator saturation analysis, wind-disturbance testing, and comparison with other adaptive and robust control methods.

Keywords: Nonlinear quadrotor, trajectory tracking, acceleration, jerk, power consumption

https://icsest.org/accepted-papers.html#CA4829

Abstract: This study presents a location-aware disaster preparedness guide that provides personalized planning support for households and communities exposed to floods, cyclones, earthquakes, heatwaves, and landslides in India. Existing disaster systems mainly focus on warnings and emergency response, while users often receive limited support for preparing before an event. The proposed system addresses this gap through a web-based application that identifies local hazards using browser geolocation or manually entered location data. It combines geographic risk information with seasonal conditions and household details, including dwelling type, family composition, mobility needs, and available resources. The system uses Python, GeoPandas, Shapely, PostgreSQL with PostGIS, Streamlit, and Leaflet.js to process spatial data and generate location-specific guidance. Hazard layers include flood-prone areas, seismic zones, cyclone paths, landslide susceptibility, population information, health facilities, shelters, and evacuation routes. A rule-based engine ranks the main hazards for each location and creates preparedness guidance divided into before, during, and after phases. The application provides risk descriptions, interactive maps, household checklists, nearby shelter information, and downloadable reports for offline use. The system was evaluated through simulated geographic scenarios and user testing under low-bandwidth conditions. The reported evaluation considers spatial accuracy, precision, recall, F1-score, response time, personalization relevance, and usability. The system achieved an overall spatial accuracy of 94.3% and an F1-score of 0.94 in the reported tests. A usability study involving 50 participants produced a System Usability Scale score of 82.4, while 92% of users completed preparedness tasks using smartphones. The findings suggest that combining geospatial risk analysis with household information can produce more useful preparedness plans than general disaster guidance. The system can support households, schools, community workers, and local authorities, although further field testing with verified hazard data and real disaster conditions is required.

Keywords: Location-based checklists, personalized emergency plans, dynamic evacuation routes, real-time disaster alerts, family communication planning, household readiness scoring, region-specific preparedness

https://icsest.org/accepted-papers.html#CA7655

Abstract: This study compares supervised machine learning methods for detecting fraudulent credit card transactions in a highly imbalanced dataset. Credit card fraud detection remains difficult because fraudulent transactions represent only a small proportion of total activity, fraud patterns change over time, and excessive false alarms can reduce the practical usefulness of automated detection systems. The analysis uses the publicly available European credit card transaction dataset containing 284,807 transactions recorded over two days in September 2013, including 492 transactions labelled as fraudulent. Six classification methods are evaluated: Logistic Regression, Decision Tree, Random Forest, Naive Bayes, K-Nearest Neighbours, and XGBoost. The study applies data preprocessing, feature scaling where required, training and test separation, and class-balancing methods to reduce the effect of the unequal class distribution. Model performance is assessed using precision, recall, F1-score, accuracy, and the area under the receiver operating characteristic curve. Particular attention is given to recall and precision because overall accuracy can provide a misleading assessment when fraudulent observations are rare. The reported results indicate that tree-based and ensemble models perform better than several conventional classifiers after class balancing. Random Forest produces the highest reported overall performance, followed closely by XGBoost, Decision Tree, and K-Nearest Neighbours. However, the unusually high reported scores require verification using a clearly separated test set and resampling procedures applied only to the training data. The study also discusses the practical limitations of machine learning fraud detection, including computational cost, model interpretability, changing fraud behaviour, false-positive risk, and possible overfitting. The findings suggest that ensemble methods can support credit card fraud screening, but reliable deployment requires leakage-free validation, probability calibration, cost-sensitive evaluation, and testing on later or external transaction datasets. Future work should investigate real-time transaction streams, concept drift, explainable predictions, deep learning methods, and decision thresholds based on the different financial costs of false negatives and false positives.

Keywords: Credit card fraud detection, machine learning, supervised learning, comparative analysis, class imbalance, ethical issues

https://icsest.org/accepted-papers.html#CA3750

Abstract: This study presents an Android-based AI investment advisor and personal expense tracking system that combines automated transaction recording, budget management, financial visualization, and conversational guidance within one application. The system is designed to reduce manual financial record keeping and help users understand their spending patterns, savings, and investment options. Financial transaction information is collected from SMS notifications and processed using pattern-matching rules to identify the transaction amount, date, merchant, and transaction type. The extracted records are then grouped into expense categories such as food, travel, entertainment, and utilities. Users can create monthly or category-based budgets, receive spending alerts, and review their financial activity through charts and summary reports. The application uses Kotlin for the Android interface, Spring Boot for backend services, MongoDB for data storage, and REST APIs for communication between system components. A natural-language chatbot is included to answer financial questions and generate saving and investment suggestions based on user income, expenses, financial goals, and recorded behavior. The system was evaluated through functional, integration, performance, security, and user-interface testing on multiple Android devices. The reported results show an SMS transaction detection accuracy of 94.6%, an expense categorization accuracy above 92%, and a user-rated chatbot recommendation relevance of approximately 90%. The chatbot produced responses in less than two seconds on average during the reported tests. The system also applies authentication and encrypted communication to protect user information. The study shows that combining automatic expense extraction with budgeting tools and conversational assistance can simplify personal finance management and reduce the effort required to maintain financial records. However, investment suggestions should be presented as general educational guidance rather than guaranteed financial advice. Future development should include larger user studies, comparison with existing finance applications, evaluation of recommendation quality using financial experts, direct integration with regulated banking services, and clearer controls for privacy, risk, and user consent.

Keywords: AI Investment Advisor, Expense Tracking, Budget Management, Personal Finance, Financial Analytics, Intelligent Chatbot

https://icsest.org/accepted-papers.html#CA7446

Abstract: Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) with external knowledge, but fixed retrieval strategies increase latency and unnecessary computation in long conversations. This paper proposes RL-ARCS, an RL-based adaptive retrieval framework that dynamically selects retrieval actions based on the conversational state. By combining HNSW retrieval with a GPU-resident retrieval cache, RL-ARCS reduces redundant searches while balancing retrieval accuracy, latency, and evidence alignment. Experiments demonstrate 0.82 retrieval precision at k = 5, 0.80 refusal F1 score, and about 20% lower query latency than the strongest baseline, while maintaining stable performance over extended conversations.

Keywords: Retrieval Augmented Generation; Reinforcement Learning; Adaptive Retrieval; HNSW; GPU Cache; Latency Reduction; Retrieval Precision; Refusal Mechanism; Conversational AI; Approximate Nearest Neighbour Search

https://icsest.org/accepted-papers.html#CA8288

Abstract: Artificial intelligence (AI) is the capability of computational systems to perform tasks typically associated with human intelligence, such as learning, reasoning, problem-solving, perception, and decision-making. It is a field of research in engineering, mathematics and computer science that develops and studies methods and software that enable machines to perceive their environment and use learning and intelligence to take actions that maximize their chances of achieving defined goals. High-profile applications of AI include advanced web search engines, chatbots, virtual assistants, autonomous vehicles, and play and analysis in strategy games such as chess and Go. Since the 2020s, generative AI has become widely available to generate images, audio, and videos from text prompts. The traditional goals of AI research include learning, reasoning, knowledge representation, planning, natural language processing, and perception, as well as support for robotics. To reach these goals, AI researchers have used techniques including state space search and mathematical optimization, formal logic, artificial neural networks, and methods based on statistics, operations research, and economics. AI also draws upon psychology, linguistics, philosophy, neuroscience, and other fields. Some companies, such as OpenAI, Google DeepMind and Meta, aim to create artificial general intelligence (AGI) — AI that can complete virtually any cognitive task at least as well as a human.

Keywords: Not provided

https://icsest.org/accepted-papers.html#CA0113

Abstract: This study investigates the effect of road maintenance levy failures on road quality and housing prices using a regression discontinuity analysis of local referendum outcomes in Ohio. Local governments often rely on voter-approved property tax levies to fund road maintenance, and the failure of a renewal levy can reduce the resources available for routine repairs. This study examines whether these local fiscal decisions affect the condition of public roads and whether the effects are reflected in housing markets over time. The analysis uses more than 3,000 road maintenance tax renewal referendums from Ohio between 1991 and 2021. Close referendum outcomes are used to compare jurisdictions where road levies narrowly failed with jurisdictions where similar levies narrowly passed. This approach supports causal estimation because communities near the voting threshold are expected to be similar before the vote. The study combines referendum records, road maintenance funding data, housing transaction data, local demographic controls, and image-based road quality assessment. A dynamic regression discontinuity model is used to estimate changes in median house prices over the years following the referendum. Road quality is also assessed using satellite imagery and an AI-based classification model. The results show that levy failure reduces road maintenance funding by about 11% and is followed by a decline in assessed road quality. The housing price effects do not appear immediately but become visible after several years, which is consistent with the gradual deterioration of roads when maintenance is reduced. The study estimates that median house prices decline by about 9% over the post-vote period, with larger effects in urban areas and higher-priced housing markets. The findings suggest that short-term tax savings from rejecting road maintenance levies may be smaller than the later loss in property values. The study provides evidence that maintenance of existing infrastructure can affect local asset values and shows how voting records, housing market data, and image-based infrastructure measurement can be used together to study public finance and infrastructure outcomes.

Keywords: Road maintenance, local taxation, regression discontinuity, house prices, infrastructure disinvestment, property tax capitalization

https://icsest.org/accepted-papers.html#CA4029

Abstract: This study examines fiscal sustainability and debt-servicing pressure in Sub-Saharan Africa across short- and long-run periods, with attention to the fiscal effects of the COVID-19 period. The analysis uses an unbalanced panel of 783 observations covering 32 countries between 1993 and 2024. Fiscal reaction functions are estimated to determine whether governments adjust their primary fiscal balances in response to changes in public debt and whether this response remains stable when debt-servicing costs increase. Fixed-effects and mean-group estimators are used as benchmark approaches, while a Bayesian implied long-run framework is applied to account for uncertainty and differences in fiscal behavior across countries and sub-regions. The results indicate persistent fiscal dynamics and positive average responses to rising debt in both the short and long run. This suggests that fiscal adjustment mechanisms remain present across the region, although their strength and duration vary considerably. Higher debt-servicing burdens weaken the size and persistence of fiscal adjustment by reducing the resources available for public spending, investment, and countercyclical policy. The Bayesian estimates show that long-run fiscal sustainability is not uniform across countries and that regional averages can conceal different national debt paths and adjustment capacities. The COVID-19 period also affected fiscal responses through lower revenue, increased public expenditure, higher borrowing requirements, and changing financing conditions. These findings indicate that fiscal sustainability in Sub-Saharan Africa should not be assessed through a single regional coefficient or a uniform debt threshold. Country-specific debt composition, interest costs, revenue capacity, access to financing, and institutional conditions must also be considered. The study contributes to fiscal-policy analysis by combining conventional panel estimators with a Bayesian framework that represents parameter uncertainty and national heterogeneity. The findings may support governments, development institutions, and lenders in designing debt-management strategies that protect essential expenditure while improving the ability of countries to meet their debt obligations over time.

Keywords: Fiscal sustainability, debt servicing, heterogeneity, Bayesian analysis, fiscal reaction

https://icsest.org/accepted-papers.html#CA0122

Abstract: This study examines the sociology of decision-making within Pakistan’s Integrated Rural Development Program and investigates the social, administrative, and political factors that limited the program’s ability to achieve its objectives. The analysis focuses on how decisions were designed, communicated, implemented, and evaluated at the local level, where rural communities were expected to receive direct benefits. The study considers the relationship between government departments, local officials, political representatives, program managers, and rural residents in shaping development outcomes. It pays particular attention to community participation, institutional coordination, bureaucratic procedures, local leadership, service delivery, and the role of self-help in rural development. A case-study approach is used to examine program records, policy documents, development reports, and examples from selected rural projects. Interviews with program officials, local representatives, and community members may also be used to understand how decisions were made and why some actions were delayed, redirected, or poorly implemented. The analysis indicates that weak community participation reduced local ownership of development activities. Limited coordination between government agencies also created delays, duplication of work, and unclear responsibility. In some cases, political interference influenced project selection and resource distribution, while rigid administrative procedures restricted local flexibility. The study further finds that limited motivation among implementing staff and weak support for village-level self-help reduced the long-term effect of program activities. These problems show that rural development outcomes are shaped not only by funding and technical planning but also by social relationships, institutional behavior, and the distribution of decision-making power. The study concludes that rural development programs require participatory planning, clearer institutional roles, transparent local decision processes, regular monitoring, and stronger cooperation between public agencies and communities. The findings may support policymakers, program managers, and rural development practitioners in designing development programs that are more responsive to local needs and better able to produce lasting social and economic outcomes.

Keywords: Integrated rural development, decision-making, development programs

https://icsest.org/accepted-papers.html#CA0121

Abstract: This study examines the policy implementation challenges that affect the integration of oral health care into primary health care systems, with particular attention to low- and middle-income countries and comparative lessons from higher-income settings. Oral health services have historically operated separately from general health services, creating gaps in access, financing, workforce planning, prevention, referral, and continuity of care. The analysis considers barriers across governance, policy design, health workforce, financing, infrastructure, service delivery, and intersectoral coordination. At the governance level, weak political commitment, fragmented regulation, and the limited inclusion of oral health in national health strategies and universal health coverage packages reduce the mandate for integration. Workforce challenges include shortages of oral health professionals, limited task-shifting arrangements, and inadequate oral health training for physicians, nurses, and community health workers. Financing barriers arise from low public investment, dependence on out-of-pocket payment, and weak reimbursement mechanisms for preventive and basic oral health services. Infrastructure and service-delivery limitations include shortages of equipment, medicines, referral pathways, and information systems. The review applies the Consolidated Framework for Implementation Research and the World Health Organization health systems building blocks to examine how institutional context influences implementation. Comparative examples from Brazil, Thailand, and Australia are used to identify enabling conditions such as political commitment, inclusion in benefit packages, workforce development, community-based delivery, referral integration, and digital health support. The findings indicate that isolated policy measures are unlikely to achieve sustainable integration. Effective implementation requires coordinated action across governance, workforce, financing, service delivery, information systems, and community engagement. The study concludes that oral health should be treated as a core component of primary health care and universal health coverage rather than as a separate specialist service. Future research should evaluate implementation fidelity, cost-effectiveness, workforce models, patient outcomes, and the long-term effects of integration at community and health-system levels.

Keywords: Oral health integration, primary health care, policy implementation, universal health coverage, health systems strengthening, health workforce, oral health equity

https://icsest.org/accepted-papers.html#CA0124

Abstract: This study investigates the effectiveness of the OMAY Framework (Observe the Text, Meaning-Making, Analyze and Summarize, Yield Understanding) as an instructional intervention to improve reading comprehension among Grade 10 students identified as frustrated readers. The research is motivated by the 2025 Philippine Informal Reading Inventory (Phil-IRI) results, which indicated that 30% of learners in a selected class exhibited frustration-level comprehension. A one-group pre-test–post-test quantitative research design was employed, involving twelve purposively selected students who participated in eight structured remedial sessions over one month. Each session incorporated OMAY-based instructional strategies and formative assessments to monitor learner progress. Data were analyzed using mean scores, Mean Percentage Score (MPS), gain scores, and standard deviation. Results showed a significant improvement in reading comprehension, with MPS increasing from 67.8% to 91.7%. Formative assessment results demonstrated a steady upward trend, while reduced standard deviation indicated improved consistency and reduced performance gaps among learners. The findings suggest that the OMAY Framework is an effective and practical approach for enhancing reading comprehension skills in secondary-level students requiring targeted intervention.

Keywords: Reading comprehension, OMAY Framework, frustrated readers, remedial intervention, action research, Phil-IRI

https://icsest.org/accepted-papers.html#CA0105

Abstract: An In-Depth Analysis of How Party Leadership Facilitates Ideational Change in Congress Reveals Key Patterns of Change Among Leaders and Non-Leaders. This study examines the ideological paths of members of Congress using the Nokken-Poole version of DW-NOMINATE, with special emphasis on Democratic leaders during key political epochs such as the New Deal. The evidence shows that most groups are not ideologically changed significantly, but party leaders move more liberally during intense political change. Based on theories of party stability, the extent to which leaders impact legislative outcomes, and changing political conditions, they explain how the position of leaders shapes the political landscape. It is essential that further studies be conducted on case studies and the advancement of statistics that could regulate the connections between ideological change and leadership roles.

Keywords: Congressional leadership; ideological evolution; DW-NOMINATE; political polarization; New Deal era; party dynamics; legislative behavior; roll-call voting; Nokken-Poole dataset; American politics

https://icsest.org/accepted-papers.html#CA7378

Abstract: This paper explores compliance with International Financial Reporting Standards (IFRS) by various actors. The research focuses on the analysis of IFRS compliance in the year 2005 to 2014. It also examines the research paradigms of IFRS conformity researchers during that period. Furthermore, a comparative analysis of the research designs (methodologies) of the IFRS compliance studies is also presented. There are several reasons why IFRS is adopted and complied with. For instance, there is great diversity in reporting standards because of the voluntary as against mandatory adoption of IFRS (corporate governance, political economic).

Keywords: IFRS compliance; research paradigms; research design; comparative analysis; voluntary adoption; mandatory adoption; corporate governance; political economy

https://icsest.org/accepted-papers.html#CA0919