Pub Date : 2025-07-26DOI: 10.1016/j.egyai.2025.100578
Ling Liu , Jihui Zhuang , Yuelei Wang , Pei Li , Dongping Guo , Xiaoming Cheng
Accurate prediction of transient nitrogen oxides (NOx) emissions from diesel vehicles is essential for precise emission inventories and effective pollution control but challenged by data nonlinearity and dynamic operating conditions. This study develops the Wavelet Decomposition (WD)-Parallel Spatiotemporal Attention-based Long Short-Term Memory (PSTALSTM) model, using real-world Portable Emission Measurement System (PEMS) and On-Board Diagnostics (OBD) data. WD preprocessing reduces emission data non-stationarity, generating more stable inputs. The PSTALSTM architecture, built upon Bidirectional Long Short-Term Memory (Bi-LSTM), incorporates a parallel attention mechanism to adaptively weight features and temporal segments, effectively capturing spatiotemporal correlations within the emission data. Validation with on-road test data demonstrates WD-PSTALSTM's superior performance over existing models. It achieves reductions exceeding 20 % in mean absolute error (MAE) and 15 % in root mean square error (RMSE), significantly enhancing prediction accuracy. These results establish WD-PSTALSTM as an effective approach for forecasting transient diesel engine NOx emissions. The research provides valuable methodologies for emission prediction based on vehicle operational data, contributing to environmental pollution mitigation efforts.
{"title":"WD-PSTALSTM: a data-driven hybrid model for prediction of diesel vehicle NOx emissions","authors":"Ling Liu , Jihui Zhuang , Yuelei Wang , Pei Li , Dongping Guo , Xiaoming Cheng","doi":"10.1016/j.egyai.2025.100578","DOIUrl":"10.1016/j.egyai.2025.100578","url":null,"abstract":"<div><div>Accurate prediction of transient nitrogen oxides (NOx) emissions from diesel vehicles is essential for precise emission inventories and effective pollution control but challenged by data nonlinearity and dynamic operating conditions. This study develops the Wavelet Decomposition (WD)-Parallel Spatiotemporal Attention-based Long Short-Term Memory (PSTALSTM) model, using real-world Portable Emission Measurement System (PEMS) and On-Board Diagnostics (OBD) data. WD preprocessing reduces emission data non-stationarity, generating more stable inputs. The PSTALSTM architecture, built upon Bidirectional Long Short-Term Memory (Bi-LSTM), incorporates a parallel attention mechanism to adaptively weight features and temporal segments, effectively capturing spatiotemporal correlations within the emission data. Validation with on-road test data demonstrates WD-PSTALSTM's superior performance over existing models. It achieves reductions exceeding 20 % in mean absolute error (MAE) and 15 % in root mean square error (RMSE), significantly enhancing prediction accuracy. These results establish WD-PSTALSTM as an effective approach for forecasting transient diesel engine NOx emissions. The research provides valuable methodologies for emission prediction based on vehicle operational data, contributing to environmental pollution mitigation efforts.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100578"},"PeriodicalIF":9.6,"publicationDate":"2025-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144757547","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-26DOI: 10.1016/j.egyai.2025.100569
Tobias Hackmann , Yunus Emir , Michael A. Danzer
Electrochemical impedance spectroscopy, a method for battery diagnostics, is used to estimate the internal temperature of a lithium-ion battery cell during highly dynamic load profiles. For the first time, a recurrent neural network is trained and evaluated with operando impedance data for temperature estimation. Furthermore, an approach is considered that guides the training process of the neural network by incorporating physical constraints. The model’s development based on an extensive series of measurements with different load profiles, tested under realistic conditions on large-format lithium-ion cells. The estimation accuracy of the data-driven approach is evaluated and compared against model-based methods, including the extended Kalman filter. An impedance correction model is proposed, which leads to a significant enhancement of the model-based estimation. The recurrent neural network under consideration achieves a mean square error of 1.07 for the investigated testing profiles in the temperature range up to 60 .
{"title":"Operando impedance-based battery cell internal temperature estimation under non-stationarity and non-linearity conditions","authors":"Tobias Hackmann , Yunus Emir , Michael A. Danzer","doi":"10.1016/j.egyai.2025.100569","DOIUrl":"10.1016/j.egyai.2025.100569","url":null,"abstract":"<div><div>Electrochemical impedance spectroscopy, a method for battery diagnostics, is used to estimate the internal temperature of a lithium-ion battery cell during highly dynamic load profiles. For the first time, a recurrent neural network is trained and evaluated with operando impedance data for temperature estimation. Furthermore, an approach is considered that guides the training process of the neural network by incorporating physical constraints. The model’s development based on an extensive series of measurements with different load profiles, tested under realistic conditions on large-format lithium-ion cells. The estimation accuracy of the data-driven approach is evaluated and compared against model-based methods, including the extended Kalman filter. An impedance correction model is proposed, which leads to a significant enhancement of the model-based estimation. The recurrent neural network under consideration achieves a mean square error of 1.07 <span><math><mrow><mo>°</mo><mi>C</mi></mrow></math></span> for the investigated testing profiles in the temperature range up to 60 <span><math><mrow><mo>°</mo><mi>C</mi></mrow></math></span>.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100569"},"PeriodicalIF":9.6,"publicationDate":"2025-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144721375","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-26DOI: 10.1016/j.egyai.2025.100566
Dylan Wald , Olga Doronina , Kathryn Johnson , Ryan King , Michael Sinner , Kevin Griffin , Rohit Chintala , Deepthi Vaidhynathan , Jibonananda Sanyal , Marc Day
Management of the electrical grid is becoming more complex due to the increased penetration of alternative energy generation technologies and a broadening diversity of electric loads. This complexity creates challenges in balancing demand and generation that can increase the potential for grid instabilities. One effective way to address this issue is to leverage previously unexploited demand flexibility through advanced control strategies. In this work, we propose an advanced control method, called adaptive neural parameter-varying model predictive control (ANPV-MPC), to control the temperature and energy consumption of a building via its Heating, Ventilation, and Air Conditioning system. ANPV-MPC combines key ideas in parameter-varying control, adaptive control, and online learning strategies to bridge the gap between computationally efficient linear model predictive control and more accurate nonlinear model predictive control. The novelty in ANPV-MPC is the use of a physics-inspired Bayesian neural network to estimate the coefficients of the parameter-varying linear control model. The Bayesian neural network additionally provides uncertainty estimates, triggering online training to capture evolving building system conditions. We show that ANPV-MPC can approximate the building system dynamics with a 28.39% higher accuracy than traditional linear model predictive control, resulting in 36.23% better control performance without increasing complexity of the optimal control problem. ANPV-MPC also adapts in real time to previously unseen conditions using online learning, further improving its performance.
{"title":"A neural-network-enhanced parameter-varying framework for multi-objective model predictive control applied to buildings","authors":"Dylan Wald , Olga Doronina , Kathryn Johnson , Ryan King , Michael Sinner , Kevin Griffin , Rohit Chintala , Deepthi Vaidhynathan , Jibonananda Sanyal , Marc Day","doi":"10.1016/j.egyai.2025.100566","DOIUrl":"10.1016/j.egyai.2025.100566","url":null,"abstract":"<div><div>Management of the electrical grid is becoming more complex due to the increased penetration of alternative energy generation technologies and a broadening diversity of electric loads. This complexity creates challenges in balancing demand and generation that can increase the potential for grid instabilities. One effective way to address this issue is to leverage previously unexploited demand flexibility through advanced control strategies. In this work, we propose an advanced control method, called adaptive neural parameter-varying model predictive control (ANPV-MPC), to control the temperature and energy consumption of a building via its Heating, Ventilation, and Air Conditioning system. ANPV-MPC combines key ideas in parameter-varying control, adaptive control, and online learning strategies to bridge the gap between computationally efficient linear model predictive control and more accurate nonlinear model predictive control. The novelty in ANPV-MPC is the use of a physics-inspired Bayesian neural network to estimate the coefficients of the parameter-varying linear control model. The Bayesian neural network additionally provides uncertainty estimates, triggering online training to capture evolving building system conditions. We show that ANPV-MPC can approximate the building system dynamics with a 28.39% higher accuracy than traditional linear model predictive control, resulting in 36.23% better control performance without increasing complexity of the optimal control problem. ANPV-MPC also adapts in real time to previously unseen conditions using online learning, further improving its performance.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100566"},"PeriodicalIF":9.6,"publicationDate":"2025-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144766606","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-23DOI: 10.1016/j.egyai.2025.100572
Jintao He , Lingfeng Shi , Yonghao Zhang , Meiyan Zhang , Yu Yao , Hailang Sang , Hua Tian , Gequn Shu
The CO2-based combined cooling and power (CCP) system is regarded as a highly promising alternative for waste heat recovery in refrigerated trucks, owing to its environmental advantages and multienergy output. The CCP system implemented in refrigerated trucks is more intricate than conventional waste heat recovery systems. It not only produces energy to satisfy demand via waste heat recovery but also incorporates refrigeration capabilities, substituting the standalone refrigeration unit to sustain low temperatures in refrigerated trucks. This coupling of power and refrigeration subcycles significantly increases the complexity of system control and the requirements for stability. Current research primarily focuses on the steady-state performance of CCP systems, neglecting the impact of load variations on the system's dynamic response in real operating conditions, thereby limiting a comprehensive assessment of operational performance under complex scenarios. This study proposes a hybrid control strategy based on deep deterministic policy gradient deep reinforcement learning and conducts dynamic simulations to comprehensively evaluate the energy efficiency performance of the CCP system. The results show that under the China Heavy-Duty Commercial Vehicle Test Cycle conditions, this strategy reduces fuel consumption by 6.63 % per 100 km while ensuring that the CCP system remains within safety constraints throughout the entire operation. These findings provide important insights for the application of CCP systems in the cold chain transportation sector.
{"title":"Optimizing combined cooling and power systems in refrigerated trucks: a deep deterministic policy gradient approach","authors":"Jintao He , Lingfeng Shi , Yonghao Zhang , Meiyan Zhang , Yu Yao , Hailang Sang , Hua Tian , Gequn Shu","doi":"10.1016/j.egyai.2025.100572","DOIUrl":"10.1016/j.egyai.2025.100572","url":null,"abstract":"<div><div>The CO<sub>2</sub>-based combined cooling and power (CCP) system is regarded as a highly promising alternative for waste heat recovery in refrigerated trucks, owing to its environmental advantages and multienergy output. The CCP system implemented in refrigerated trucks is more intricate than conventional waste heat recovery systems. It not only produces energy to satisfy demand via waste heat recovery but also incorporates refrigeration capabilities, substituting the standalone refrigeration unit to sustain low temperatures in refrigerated trucks. This coupling of power and refrigeration subcycles significantly increases the complexity of system control and the requirements for stability. Current research primarily focuses on the steady-state performance of CCP systems, neglecting the impact of load variations on the system's dynamic response in real operating conditions, thereby limiting a comprehensive assessment of operational performance under complex scenarios. This study proposes a hybrid control strategy based on deep deterministic policy gradient deep reinforcement learning and conducts dynamic simulations to comprehensively evaluate the energy efficiency performance of the CCP system. The results show that under the China Heavy-Duty Commercial Vehicle Test Cycle conditions, this strategy reduces fuel consumption by 6.63 % per 100 km while ensuring that the CCP system remains within safety constraints throughout the entire operation. These findings provide important insights for the application of CCP systems in the cold chain transportation sector.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100572"},"PeriodicalIF":9.6,"publicationDate":"2025-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144724036","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-23DOI: 10.1016/j.egyai.2025.100562
Tianxiang Cui , Yujian Ye , Yiran Li , Nanjiang Du , Xingke Song , Yicheng Zhu , Xiaoying Yang , Goran Strbac
The energy market plays a fundamental role in the global economy, shaping energy prices, inflation, and financial stability across nations. As the world transitions toward low-carbon energy solutions, optimizing trading strategies in this complex and dynamic market has become increasingly critical for investors, polic-ymakers, and energy brokers. Traditional data-driven models often struggle to capture the multifaceted and interconnected factors influencing energy markets, such as macroeconomic conditions, investor sentiment, and the accelerating shift toward decarbonization. To address these challenges, a novel framework is proposed that combines reinforcement learning with methods for analyzing disagreement and connectedness, alongside advanced natural language processing techniques, to develop trading strategies for energy markets. The proposed method integrates structured time-series data with unstructured textual data to incorporate diverse factors, including the interplay between economic influences, green energy transitions, and investor sentiment. The proposed framework also employs a chain-of-reasoning technique to classify investor types, distinguishing between sentiment-driven disagreement and cross-disagreement, and utilizes a connectedness-based method to model the interrelationships among market variables, providing a comprehensive understanding of market dynamics. As a showcase, this framework is applied to the West Texas Intermediate crude oil market, demonstrating its ability to outperform traditional price-prediction-based trading strategies. Experimental results highlight that the proposed framework delivers superior investment returns while addressing key limitations of existing models in terms of data integration and flexibility. This study underscores the potential of the proposed framework as a robust and adaptable solution for optimizing trading strategies across the broader energy market, with particular relevance to the global transition toward sustainable energy systems.
{"title":"Toward profitable energy futures trading strategies using reinforcement learning incorporating disagreement and connectedness methods enabled by large language models","authors":"Tianxiang Cui , Yujian Ye , Yiran Li , Nanjiang Du , Xingke Song , Yicheng Zhu , Xiaoying Yang , Goran Strbac","doi":"10.1016/j.egyai.2025.100562","DOIUrl":"10.1016/j.egyai.2025.100562","url":null,"abstract":"<div><div>The energy market plays a fundamental role in the global economy, shaping energy prices, inflation, and financial stability across nations. As the world transitions toward low-carbon energy solutions, optimizing trading strategies in this complex and dynamic market has become increasingly critical for investors, polic-ymakers, and energy brokers. Traditional data-driven models often struggle to capture the multifaceted and interconnected factors influencing energy markets, such as macroeconomic conditions, investor sentiment, and the accelerating shift toward decarbonization. To address these challenges, a novel framework is proposed that combines reinforcement learning with methods for analyzing disagreement and connectedness, alongside advanced natural language processing techniques, to develop trading strategies for energy markets. The proposed method integrates structured time-series data with unstructured textual data to incorporate diverse factors, including the interplay between economic influences, green energy transitions, and investor sentiment. The proposed framework also employs a chain-of-reasoning technique to classify investor types, distinguishing between sentiment-driven disagreement and cross-disagreement, and utilizes a connectedness-based method to model the interrelationships among market variables, providing a comprehensive understanding of market dynamics. As a showcase, this framework is applied to the West Texas Intermediate crude oil market, demonstrating its ability to outperform traditional price-prediction-based trading strategies. Experimental results highlight that the proposed framework delivers superior investment returns while addressing key limitations of existing models in terms of data integration and flexibility. This study underscores the potential of the proposed framework as a robust and adaptable solution for optimizing trading strategies across the broader energy market, with particular relevance to the global transition toward sustainable energy systems.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100562"},"PeriodicalIF":9.6,"publicationDate":"2025-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144704729","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-20DOI: 10.1016/j.egyai.2025.100563
Xiaodan Yu , Ruijia Jiang , Xiaolong Jin , Hongjie Jia , Yunfei Mu , Wei Wei , Wanxin Tang
Modern distribution networks with high penetration of distributed energy resources (DERs) are undergoing continuous expansion in scale. However, the increasing complexity of network structure and the high installation cost of measurement equipment introduce operational challenges including state variability and measurement data incompleteness. Substantial data loss significantly compromises fault detection accuracy and network performance, creating obstacles for distributed energy management and posing critical challenges to distribution network state estimation. To address these issues, this paper proposes a hybrid state estimation framework (MC-VMD-ARIMA-LSTM) that integrates alternating-minimization matrix completion (MC) with variational mode decomposition (VMD), autoregressive integrated moving average (ARIMA) modeling, and long short-term memory (LSTM) neural networks for enhanced power flow analysis in low-observability distribution networks. The methodology features a dual-timescale approach: (1) At individual time intervals, an alternating-minimization matrix completion model is formulated, incorporating linearized power flow constraints; (2) For multi-timescale analysis, the measurement dataset undergoes VMD-based decomposition, with subsequent specialized processing where ARIMA handles low-frequency components and LSTM manage high-frequency residuals. The results of state estimation are obtained through systematic component reconstruction. Comprehensive evaluations using IEEE 33-bus distribution network and actual distribution system measurement datasets demonstrate the framework's effectiveness in both multi-timescale data assimilation and state estimation accuracy under limited observability conditions.
{"title":"Method for compensating multi-time measurement data of distribution network based on alternating minimization matrix completion combined with VMD-ARIMA-LSTM","authors":"Xiaodan Yu , Ruijia Jiang , Xiaolong Jin , Hongjie Jia , Yunfei Mu , Wei Wei , Wanxin Tang","doi":"10.1016/j.egyai.2025.100563","DOIUrl":"10.1016/j.egyai.2025.100563","url":null,"abstract":"<div><div>Modern distribution networks with high penetration of distributed energy resources (DERs) are undergoing continuous expansion in scale. However, the increasing complexity of network structure and the high installation cost of measurement equipment introduce operational challenges including state variability and measurement data incompleteness. Substantial data loss significantly compromises fault detection accuracy and network performance, creating obstacles for distributed energy management and posing critical challenges to distribution network state estimation. To address these issues, this paper proposes a hybrid state estimation framework (MC-VMD-ARIMA-LSTM) that integrates alternating-minimization matrix completion (MC) with variational mode decomposition (VMD), autoregressive integrated moving average (ARIMA) modeling, and long short-term memory (LSTM) neural networks for enhanced power flow analysis in low-observability distribution networks. The methodology features a dual-timescale approach: (1) At individual time intervals, an alternating-minimization matrix completion model is formulated, incorporating linearized power flow constraints; (2) For multi-timescale analysis, the measurement dataset undergoes VMD-based decomposition, with subsequent specialized processing where ARIMA handles low-frequency components and LSTM manage high-frequency residuals. The results of state estimation are obtained through systematic component reconstruction. Comprehensive evaluations using IEEE 33-bus distribution network and actual distribution system measurement datasets demonstrate the framework's effectiveness in both multi-timescale data assimilation and state estimation accuracy under limited observability conditions.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100563"},"PeriodicalIF":9.6,"publicationDate":"2025-07-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144723945","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-16DOI: 10.1016/j.egyai.2025.100568
Chengyin Shi , Cong Yin , Weilong Luo , Hailong Liu , Hao Tang
The Proton Exchange Membrane Fuel Cell (PEMFC) converts the chemical energy of hydrogen fuel directly into electrical energy with broad application prospects. Understanding how current density is distributed in the PEMFC systems is crucial as it is a key factor influencing system performance. However, direct modeling for current distribution may encounter the challenge of dimensional catastrophe owing to the high dimensionality of the data. This paper uses a high-resolution segmented measurement device with 396 points to conduct experimental tests on the current distribution of a PEMFC with reactive area of 406 cm2 during a stepwise increase in load current. The current distribution is modeled based on the test results to learn the mapping relationship between the experimental parameters and the current distribution. The proposed model utilizes a Conditional Variational Auto-Encoder (CVAE) to generate current distributions. The MSE (Mean-Square Error) of the trained CVAE model reaches 9.2 × 10–5, and the comparison results show that the 222.9A current distribution error has the largest MSE of 6.36 × 10–4 and a KL Divergence (Kullback-Leibler Divergence) of 9.55 × 10–4, both of which are at a low level. This model enables the direct determination of the current distribution based on the experimental parameters, thereby establishing a technical foundation for investigating the impact of experimental conditions on fuel cells. This model is also of great significance for research on fuel cell system control strategies and fault diagnosis.
{"title":"Study of current distribution generation in PEMFC based on conditional variational auto-encoder","authors":"Chengyin Shi , Cong Yin , Weilong Luo , Hailong Liu , Hao Tang","doi":"10.1016/j.egyai.2025.100568","DOIUrl":"10.1016/j.egyai.2025.100568","url":null,"abstract":"<div><div>The Proton Exchange Membrane Fuel Cell (PEMFC) converts the chemical energy of hydrogen fuel directly into electrical energy with broad application prospects. Understanding how current density is distributed in the PEMFC systems is crucial as it is a key factor influencing system performance. However, direct modeling for current distribution may encounter the challenge of dimensional catastrophe owing to the high dimensionality of the data. This paper uses a high-resolution segmented measurement device with 396 points to conduct experimental tests on the current distribution of a PEMFC with reactive area of 406 cm<sup>2</sup> during a stepwise increase in load current. The current distribution is modeled based on the test results to learn the mapping relationship between the experimental parameters and the current distribution. The proposed model utilizes a Conditional Variational Auto-Encoder (CVAE) to generate current distributions. The MSE (Mean-Square Error) of the trained CVAE model reaches 9.2 × 10<sup>–5</sup>, and the comparison results show that the 222.9A current distribution error has the largest MSE of 6.36 × 10<sup>–4</sup> and a KL Divergence (Kullback-Leibler Divergence) of 9.55 × 10<sup>–4</sup>, both of which are at a low level. This model enables the direct determination of the current distribution based on the experimental parameters, thereby establishing a technical foundation for investigating the impact of experimental conditions on fuel cells. This model is also of great significance for research on fuel cell system control strategies and fault diagnosis.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100568"},"PeriodicalIF":9.6,"publicationDate":"2025-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144713439","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper introduces a novel hybrid optimization algorithm, Adaptive Hybrid PSO-Embedded GA (AHPEGA), which dynamically adapts to optimization performance by integrating Particle Swarm Optimization (PSO) and Genetic Algorithms (GA). The primary objective is to enhance the neuroevolutionary training of multilayer perceptron-based controllers (MLPCs) through the joint optimization of model parameters and structural hyperparameters. Traditional training methods frequently encounter issues such as premature convergence and limited generalization. AHPEGA addresses these limitations through an adaptive training strategy that dynamically adjusts parameters during the evolutionary process, thereby improving convergence speed and solution quality. By effectively reducing entrapment in local minima and balancing exploration and exploitation, AHPEGA improves the quality of neural controller design. The algorithm’s performance is evaluated against conventional optimization methods, demonstrating significant improvements in accuracy, convergence speed, and consistency across multiple runs. The practical applicability of the proposed method is demonstrated through simulation in the context of a VSC-based islanded microgrid (MG), where ensuring reliable and effective control under variable operating conditions is critical. This highlights AHPEGA’s capability to optimize intelligent control strategies in MG systems, particularly under dynamic and uncertain conditions, reinforcing its practical value in real-world energy environments.
{"title":"Adaptive Hybrid PSO-Embedded GA for neuroevolutionary training of multilayer perceptron controllers in VSC-based islanded microgrids","authors":"Yared Bekele Beyene , Getachew Biru Worku , Lina Bertling Tjernberg","doi":"10.1016/j.egyai.2025.100551","DOIUrl":"10.1016/j.egyai.2025.100551","url":null,"abstract":"<div><div>This paper introduces a novel hybrid optimization algorithm, Adaptive Hybrid PSO-Embedded GA (AHPEGA), which dynamically adapts to optimization performance by integrating Particle Swarm Optimization (PSO) and Genetic Algorithms (GA). The primary objective is to enhance the neuroevolutionary training of multilayer perceptron-based controllers (MLPCs) through the joint optimization of model parameters and structural hyperparameters. Traditional training methods frequently encounter issues such as premature convergence and limited generalization. AHPEGA addresses these limitations through an adaptive training strategy that dynamically adjusts parameters during the evolutionary process, thereby improving convergence speed and solution quality. By effectively reducing entrapment in local minima and balancing exploration and exploitation, AHPEGA improves the quality of neural controller design. The algorithm’s performance is evaluated against conventional optimization methods, demonstrating significant improvements in accuracy, convergence speed, and consistency across multiple runs. The practical applicability of the proposed method is demonstrated through simulation in the context of a VSC-based islanded microgrid (MG), where ensuring reliable and effective control under variable operating conditions is critical. This highlights AHPEGA’s capability to optimize intelligent control strategies in MG systems, particularly under dynamic and uncertain conditions, reinforcing its practical value in real-world energy environments.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100551"},"PeriodicalIF":9.6,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144655152","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-14DOI: 10.1016/j.egyai.2025.100560
David Hamlyn, Sunny Chaudhary, Tasmiat Rahman
Accurate estimation of diffuse horizontal irradiance (DHI) is critical for optimising photovoltaic system performance and energy forecasting yet remains challenging in regions lacking comprehensive ground-based instrumentation. Recent advancements using Vision Transformers (ViTs) trained on extensive sky image datasets have shown promise in replacing costly irradiance measurement equipment, but the scarcity of long-term, high-quality sky imagery significantly restricts practical implementation. Addressing this critical gap, this study proposes a novel dual-framework approach designed for data-scarce scenarios. First, calculated atmospheric parameters, including extraterrestrial irradiance and cyclic time encodings, are integrated to represent sky conditions without utilising any instrumentation. Next, a sequential pipeline initially predicts synthetic global horizontal irradiance (GHI) and uses it as a feature, to refine DHI estimation. Finally, a dual-parallel architecture simultaneously processes raw and overlay-enhanced fisheye sky images. Overlays are generated through unsupervised, physics-informed cloud segmentation to highlight dynamic sky features. Empirical validation is performed using data from the Chilbolton Observatory, chosen for its temperate climate and frequent cloud variability. To simulate data-scarce conditions, models are trained on a single month (e.g., January) and evaluated across a temporally disjoint, full-year test set. Under this setup, the sequential and dual-parallel frameworks achieve RMSE values within 2–3 W/m² and 1–6 W/m², respectively, of a state-of-the-art ViT trained on the complete dataset. By combining physics-informed modelling with unsupervised segmentation, the proposed method provides a scalable and cost-effective solution for DHI estimation, advancing solar resource assessment in data-constrained environments.
{"title":"Vision transformers for estimating irradiance using data scarce sky images","authors":"David Hamlyn, Sunny Chaudhary, Tasmiat Rahman","doi":"10.1016/j.egyai.2025.100560","DOIUrl":"10.1016/j.egyai.2025.100560","url":null,"abstract":"<div><div>Accurate estimation of diffuse horizontal irradiance (DHI) is critical for optimising photovoltaic system performance and energy forecasting yet remains challenging in regions lacking comprehensive ground-based instrumentation. Recent advancements using Vision Transformers (ViTs) trained on extensive sky image datasets have shown promise in replacing costly irradiance measurement equipment, but the scarcity of long-term, high-quality sky imagery significantly restricts practical implementation. Addressing this critical gap, this study proposes a novel dual-framework approach designed for data-scarce scenarios. First, calculated atmospheric parameters, including extraterrestrial irradiance and cyclic time encodings, are integrated to represent sky conditions without utilising any instrumentation. Next, a sequential pipeline initially predicts synthetic global horizontal irradiance (GHI) and uses it as a feature, to refine DHI estimation. Finally, a dual-parallel architecture simultaneously processes raw and overlay-enhanced fisheye sky images. Overlays are generated through unsupervised, physics-informed cloud segmentation to highlight dynamic sky features. Empirical validation is performed using data from the Chilbolton Observatory, chosen for its temperate climate and frequent cloud variability. To simulate data-scarce conditions, models are trained on a single month (e.g., January) and evaluated across a temporally disjoint, full-year test set. Under this setup, the sequential and dual-parallel frameworks achieve RMSE values within 2–3 W/m² and 1–6 W/m², respectively, of a state-of-the-art ViT trained on the complete dataset. By combining physics-informed modelling with unsupervised segmentation, the proposed method provides a scalable and cost-effective solution for DHI estimation, advancing solar resource assessment in data-constrained environments.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100560"},"PeriodicalIF":9.6,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144721374","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-14DOI: 10.1016/j.egyai.2025.100554
Muhammad Azam Hafeez , Alberto Procacci , Axel Coussement , Alessandro Parente
Reduced-order models offer a cost-effective and accurate approach to analyzing high-dimensional combustion problems. These surrogate models are built in a data-driven manner by combining computational fluid dynamics simulations with Proper Orthogonal Decomposition (POD) for dimensionality reduction and Gaussian Process Regression (GPR) for nonlinear regression. However, these models can yield physically inconsistent results, such as negative mass fractions. As a linear decomposition method, POD complicates the enforcement of constraints in the reduced space, while GPR lacks inherent provisions to ensure physical consistency. To address these challenges, this study proposes a novel constrained reduced-order model framework that enforces physical consistency in predictions. Dimensionality reduction is achieved by downsampling the dataset through low-cost Singular Value Decomposition (lcSVD) using optimal sensor placement, ensuring that the retained data points preserve physical information in the reduced space. We integrate finite-support parametric distribution functions, such as truncated Gaussian and beta distribution scaled to the interval , into the GPR framework. These bounded likelihood functions explicitly model the observational noise in the bounded space and use variational inference to approximate analytically intractable posterior distributions, producing GP estimations that satisfy physical constraints by construction. We validate the proposed methods using a synthetic dataset and a benchmark case of one-dimensional laminar NH3/H2 flames. The results show that the thermo-chemical state predictions comply with physical constraints while maintaining the high accuracy of unconstrained reduced-order models.
{"title":"Constrained reduced-order modeling using bounded Gaussian processes for physically consistent reacting flow predictions","authors":"Muhammad Azam Hafeez , Alberto Procacci , Axel Coussement , Alessandro Parente","doi":"10.1016/j.egyai.2025.100554","DOIUrl":"10.1016/j.egyai.2025.100554","url":null,"abstract":"<div><div>Reduced-order models offer a cost-effective and accurate approach to analyzing high-dimensional combustion problems. These surrogate models are built in a data-driven manner by combining computational fluid dynamics simulations with Proper Orthogonal Decomposition (POD) for dimensionality reduction and Gaussian Process Regression (GPR) for nonlinear regression. However, these models can yield physically inconsistent results, such as negative mass fractions. As a linear decomposition method, POD complicates the enforcement of constraints in the reduced space, while GPR lacks inherent provisions to ensure physical consistency. To address these challenges, this study proposes a novel constrained reduced-order model framework that enforces physical consistency in predictions. Dimensionality reduction is achieved by downsampling the dataset through low-cost Singular Value Decomposition (lcSVD) using optimal sensor placement, ensuring that the retained data points preserve physical information in the reduced space. We integrate finite-support parametric distribution functions, such as truncated Gaussian and beta distribution scaled to the interval <span><math><mrow><mo>[</mo><mi>a</mi><mo>,</mo><mi>b</mi><mo>]</mo></mrow></math></span>, into the GPR framework. These bounded likelihood functions explicitly model the observational noise in the bounded space and use variational inference to approximate analytically intractable posterior distributions, producing GP estimations that satisfy physical constraints by construction. We validate the proposed methods using a synthetic dataset and a benchmark case of one-dimensional laminar NH<sub>3</sub>/H<sub>2</sub> flames. The results show that the thermo-chemical state predictions comply with physical constraints while maintaining the high accuracy of unconstrained reduced-order models.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100554"},"PeriodicalIF":9.6,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144655154","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号