Energy and AI最新文献

{"title":"Probabilistic ultra-short-term solar photovoltaic power forecasting using natural gradient boosting with attention-enhanced neural networks","authors":"Zhe Song ,&nbsp;Fu Xiao ,&nbsp;Zhe Chen ,&nbsp;Henrik Madsen","doi":"10.1016/j.egyai.2025.100496","DOIUrl":"10.1016/j.egyai.2025.100496","url":null,"abstract":"<div><div>Probabilistic forecasting provides insights in estimating the uncertainty of photovoltaic (PV) power forecasts. In this study, an innovative probabilistic ultra-short-term PV power forecasting framework that integrates natural gradient boosting (NGBoost) and deep neural networks is developed. Specifically, an attention-enhanced neural network combining convolutional neural networks (CNN) and bidirectional long short-term memory (BiLSTM) networks is employed for feature engineering to extract abstract features from time-series data. The extracted features are then fed into an optimized NGBoost model to yield probabilistic forecasts. In comparison to the benchmark models, i.e., the recently reported quantile regression (QR)-based deep learning methods and NGBoost, the proposed model demonstrates an enhanced ability to capture variation patterns in PV power output, further improving the forecast skill score by approximately 15–60 % in deterministic forecasting. In terms of probabilistic forecasting, the proposed model shows superior forecast reliability and sharpness compared to all benchmark methods. Its continuous ranked probability score (CRPS) ranges from 0.0710 kW to 0.0898 kW, achieving reductions of approximately 21–43 % over QR-based models and 29–40 % over NGBoost. Furthermore, within confidence intervals of 10–90 %, the proposed model consistently maintains higher coverage probabilities along with narrower average forecast intervals, as evidenced by a lower Winkler score (WS) than the benchmark models. The findings of this study provide insightful references for probabilistic PV power forecasting research, contributing to efficient solar power management and dispatch.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"20 ","pages":"Article 100496"},"PeriodicalIF":9.6,"publicationDate":"2025-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143629321","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A comparative framework for evaluating machine learning models in forecasting electricity demand for port microgrids","authors":"Alexander Micallef ,&nbsp;Maurice Apap ,&nbsp;John Licari ,&nbsp;Cyril Spiteri Staines ,&nbsp;Zhaoxia Xiao","doi":"10.1016/j.egyai.2025.100494","DOIUrl":"10.1016/j.egyai.2025.100494","url":null,"abstract":"<div><div>This study presents a framework for forecasting electricity demand in port microgrids using advanced machine learning models, including Random Forest, Least Squares Boosting Ensemble, and Gaussian Process Regression. These models were evaluated under different forecasting setups (fixed origin, expanding windows, and rolling windows) and compared against simpler baseline methods, such as Linear Regression and Naive models. The study assessed the effectiveness of machine learning models in handling dynamic electricity demand patterns in port environments and highlighted the advantages of data-driven models. Results indicate that the Random Forest (expanding window) model outperforms the other models, achieving a root mean square error of 1.1848 MW and a mean average percentage error of 7.2483 %. Gaussian Process Regression with Exponential kernel follows closely with a root mean square error of 1.1904 MW and a mean average percentage error of 7.5017 %. In contrast, the Naive Method (previous day) shows the poorest performance with a root mean square error of 4.5357 MW and a mean average percentage error of 18.1485 %. Partial Dependence Plots reveal that features such as weighted port calls play a significant role in improving prediction accuracy. These findings highlight the effectiveness of machine learning models in accurately forecasting port microgrid demand and optimizing energy management.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"20 ","pages":"Article 100494"},"PeriodicalIF":9.6,"publicationDate":"2025-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143619143","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Review of machine learning techniques for energy sharing and biomass waste gasification pathways in integrating solar greenhouses into smart energy systems","authors":"Navid Mahdavi,&nbsp;Animesh Dutta,&nbsp;Syeda Humaira Tasnim,&nbsp;Shohel Mahmud","doi":"10.1016/j.egyai.2025.100498","DOIUrl":"10.1016/j.egyai.2025.100498","url":null,"abstract":"<div><div>The integration of solar greenhouses into smart energy systems (SESs) remains largely unexplored, despite their potential to enhance energy sharing and hydrogen production. This review investigates the role of solar greenhouses as active energy contributors within SESs, emphasizing their biomass waste gasification for hydrogen production and their integration into district heating and cooling (DHC) networks. A structured classification of machine learning (ML) and deep learning (DL) techniques applied in forecasting and optimizing these processes is provided. Additionally, the evolution of DHC systems is analyzed, with a focus on fifth-generation DHC (5GDHC) networks, which facilitate bidirectional energy exchange at near-ambient temperatures. The review highlights that existing studies have predominantly addressed SES advancements and ML-driven energy management without considering the contributions of solar greenhouses. A novel framework is proposed, illustrating their role as prosumers capable of exchanging electricity, hydrogen, and thermal energy within SESs. Key findings reveal that integrating solar greenhouses with SESs can enhance energy efficiency, reduce carbon emissions, and improve system resilience. Furthermore, ML-driven predictive control strategies, particularly model predictive control (MPC), are identified as essential for optimizing real-time energy flows and biomass gasification processes. This study provides a foundation for future research on the technical, economic, and environmental feasibility of integrating greenhouses into SESs. The insights presented offer a pathway toward more sustainable, AI-driven energy-sharing networks, supporting policymakers and industry stakeholders in the transition toward low-carbon energy solutions.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"20 ","pages":"Article 100498"},"PeriodicalIF":9.6,"publicationDate":"2025-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143685449","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-objective optimization of hybrid electric vehicles energy management using multi-agent deep reinforcement learning framework","authors":"Xiaoyu Li ,&nbsp;Zaihang Zhou ,&nbsp;Changyin Wei ,&nbsp;Xiao Gao ,&nbsp;Yibo Zhang","doi":"10.1016/j.egyai.2025.100491","DOIUrl":"10.1016/j.egyai.2025.100491","url":null,"abstract":"<div><div>Hybrid electric vehicles (HEVs) have the advantages of lower emissions and less noise pollution than traditional fuel vehicles. Developing reasonable energy management strategies (EMSs) can effectively reduce fuel consumption and improve the fuel economy of HEVs. However, current EMSs still have problems, such as complex multi-objective optimization and poor algorithm robustness. Herein, a multi-agent reinforcement learning (MADRL) framework is proposed based on Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm to solve such problems. Specifically, a vehicle model and dynamics model are established, and based on this, a multi-objective EMS is developed by considering fuel economy, maintaining the battery State of Charge (SOC), and reducing battery degradation. Secondly, the proposed strategy regards the engine and battery as two agents, and the agents cooperate with each other to realize optimal power distribution and achieve the optimal control strategy. Finally, the WLTC and HWFET driving cycles are employed to verify the performances of the proposed method, the fuel consumption decreases by 26.91 % and 8.41 % on average compared to the other strategies. The simulation results demonstrate that the proposed strategy has remarkable superiority in multi-objective optimization.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"20 ","pages":"Article 100491"},"PeriodicalIF":9.6,"publicationDate":"2025-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143552827","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Geospatial Mapping of Large-Scale Electric Power Grids: A Residual Graph Convolutional Network-Based Approach with Attention Mechanism","authors":"Razzaqul Ahshan ,&nbsp;Md. Shadman Abid ,&nbsp;Mohammed Al-Abri","doi":"10.1016/j.egyai.2025.100486","DOIUrl":"10.1016/j.egyai.2025.100486","url":null,"abstract":"<div><div>Precise geospatial mapping of grid infrastructure is essential for the effective development and administration of large-scale electrical infrastructure. The application of deep learning techniques in predicting regional energy network architecture utilizing extensive datasets of geographical information systems (GISs) has yet to be thoroughly investigated in previous research works. Moreover, although graph convolutional networks (GCNs) have been proven to be effective in capturing the complex linkages within graph-structured data, the computationally demanding nature of modern energy grids necessitates additional computational contributions. Hence, this research introduces a novel residual GCN with attention mechanism for mapping critical energy infrastructure components in geographic contexts. The proposed model accurately predicts the geographic locations and links of large-scale grid infrastructure, such as poles, electricity service points, and substations. The proposed framework is assessed on the Sultanate of Oman’s regional energy grid and further validated on Nigeria’s electricity transmission network database. The obtained findings showcase the model’s capacity to accurately predict infrastructure components and their spatial relationships. Results show that the proposed method achieves a link-prediction accuracy of 95.88% for the Omani network and 92.98% for the Nigerian dataset. Furthermore, the proposed model achieved <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> values of 0.99 for both datasets in terms of regression. Therefore, the proposed architecture facilitates multifaceted assessment and enhances the capacity to capture the inherent geospatial aspects of large-scale energy distribution networks.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"20 ","pages":"Article 100486"},"PeriodicalIF":9.6,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143570547","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Hybrid forecasting of demand flexibility: A top-down approach for thermostatically controlled loads","authors":"Luca Massidda,&nbsp;Marino Marrocu","doi":"10.1016/j.egyai.2025.100487","DOIUrl":"10.1016/j.egyai.2025.100487","url":null,"abstract":"<div><div>Demand-side flexibility is crucial to balancing supply and demand, as renewable energy sources are increasingly integrated into the energy mix, and heating and transport systems are becoming more and more electrified. Historically, this balancing has been managed from the supply side. However, the shift towards renewable energy sources limits the controllability of traditional fossil fuel plants, increasing the importance of demand response (DR) techniques to achieve the required flexibility. Aggregators participating in flexibility markets need to accurately forecast the adaptability they can offer, a task complicated by numerous influencing variables. Based on a top-down approach, this study addresses the problem of forecasting electricity demand in the presence of flexibility from thermostatically controlled loads. We propose a hybrid model that combines data-driven techniques for probabilistic estimation of electricity consumption with a disaggregation of electricity consumption to identify the fraction of thermal loads, subject to flexibility, which is simulated by a virtual battery model. The technique is applied to a synthetic dataset that simulates the response of a European neighborhood to demand response interventions. The results demonstrate the model’s ability to accurately predict both the reduction in electricity demand during DR events and the subsequent rebound in consumption. The model achieves a mean absolute percentage error (MAPE) lower than 17.0%, comparable to the accuracy without flexibility. The results obtained are compared with a direct data-driven approach, demonstrating the validity and effectiveness of our model.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"20 ","pages":"Article 100487"},"PeriodicalIF":9.6,"publicationDate":"2025-02-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143511454","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Advancing oil and gas emissions assessment through large language model data extraction","authors":"Zhenlin Chen ,&nbsp;Roujia Zhong ,&nbsp;Wennan Long ,&nbsp;Haoyu Tang ,&nbsp;Anjing Wang ,&nbsp;Zemin Liu ,&nbsp;Xuelin Yang ,&nbsp;Bo Ren ,&nbsp;James Littlefield ,&nbsp;Sanmi Koyejo ,&nbsp;Mohammad S. Masnadi ,&nbsp;Adam R. Brandt","doi":"10.1016/j.egyai.2025.100481","DOIUrl":"10.1016/j.egyai.2025.100481","url":null,"abstract":"<div><div>The oil and gas industry strives to improve environmental stewardship and reduce its carbon footprint, but lacks comprehensive global operational data for accurate environmental assessment and decision-making. This challenge is compounded by dispersed information sources and the high costs of accessing proprietary databases. This paper presents an innovative framework using Large Language Models (LLMs) – specifically GPT-4 and GPT-4o – to extract critical oil and gas asset information from diverse literature sources.</div><div>Our framework employs iterative comparisons between GPT-4’s output and a dataset of 129 ground truth documents labeled by domain experts. Through 11 training and testing iterations, we fine-tuned prompts to optimize information extraction. The evaluation process assessed performance using true positive rate, precision, and F1 score metrics. The framework achieved strong results, with a true positive rate of 83.74% and an F1 score of 78.16% on the testing dataset.</div><div>The system demonstrated remarkable efficiency, processing 32 documents in 61.41 min with GPT-4o, averaging 7.09 s per extraction - a substantial improvement over the manual method. Cost-effectiveness was also achieved, with GPT-4o reducing extraction costs by a factor of 10 compared to GPT-4.</div><div>This research has significant implications for the oil and gas industry. By creating an organized, transparent, and accessible database, we aim to democratize access to critical information. The framework supports more accurate climate modeling efforts, enhances decision-making processes for operations and investments, and contributes to the sector’s ability to meet environmental commitments. These improvements particularly impact emissions reduction and energy transition strategies, potentially transforming how data is extracted and utilized in this field and beyond.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"20 ","pages":"Article 100481"},"PeriodicalIF":9.6,"publicationDate":"2025-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143520623","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Metaheuristic multi-objective optimization with artificial neural networks surrogate modeling for optimal energy-economic performance for CSP technology","authors":"A. Allouhi ,&nbsp;M. Benzakour Amine ,&nbsp;K.A. Tabet Aoul","doi":"10.1016/j.egyai.2025.100488","DOIUrl":"10.1016/j.egyai.2025.100488","url":null,"abstract":"<div><div>Among CSP technologies, the linear Fresnel reflector (LFR) can provide reliable carbon-neutral electricity for large-scale applications. In this study, the performance of a large solar LFR power plant under varying climatic conditions and the dependency of the performance on major plant design specifications, such as solar multiple and full-load thermal storage hours, were examined. Next, artificial neural network (ANN) surrogate models were introduced to predict the annual capacity factor of 100 MWe power plants operating with LFR technology. Single-hidden-layer ANN models with different numbers of neurons in the hidden layer were used and the Levenberg–Marquardt training algorithm was adopted. To overcome overfitting, validation and Bayesian Regularization approaches were compared. As training and testing data, 36 geographical sites with various combinations of design parameters were used. Through multi-objective optimization techniques, including the Multi-Objective Particle-Swarm Optimizer and Multi-Objective Grey Wolf Optimizer coupled with ANN surrogate modeling, this study navigates the trade-offs to identify Pareto-optimal solutions for large-scale LFR-based CSP integration based on the energy and cost criteria. The study also identified Site 4 (S4) as a promising candidate for optimal balance between the capacity factor (51.05%) and specific cost (5246.71$/kW), showcasing the practical implications of the research for sustainable and efficient CSP plant implementation.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"20 ","pages":"Article 100488"},"PeriodicalIF":9.6,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143527504","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Hybrid transformer model with liquid neural networks and learnable encodings for buildings’ energy forecasting","authors":"Gabriel Antonesi ,&nbsp;Tudor Cioara ,&nbsp;Ionut Anghel ,&nbsp;Ioannis Papias ,&nbsp;Vasilis Michalakopoulos ,&nbsp;Elissaios Sarmas","doi":"10.1016/j.egyai.2025.100489","DOIUrl":"10.1016/j.egyai.2025.100489","url":null,"abstract":"<div><div>Accurate forecasting of buildings' energy demand is essential for building operators to manage loads and resources efficiently, and for grid operators to balance local production with demand. However, nowadays models still struggle to capture nonlinear relationships influenced by external factors like weather and consumer behavior, assume constant variance in energy data over time, and often fail to model sequential data. To address these limitations, we propose a hybrid Transformer-based model with Liquid Neural Networks and learnable encodings for building energy forecasting. The model leverages Dense Layers to learn non-linear mappings to create embeddings that capture underlying patterns in time series energy data. Additionally, a Convolutional Neural Network encoder is integrated to enhance the model's ability to understand temporal dynamics through spatial mappings. To address the limitations of classic attention mechanisms, we implement a reservoir processing module using Liquid Neural Networks which introduces a controlled non-linearity through dynamic reservoir computing, enabling the model to capture complex patterns in the data. For model evaluation, we utilized both pilot data and state-of-the-art datasets to determine the model's performance across various building contexts, including large apartment and commercial buildings and small households, with and without on-site energy production. The proposed transformer model demonstrates good predictive accuracy and training time efficiency across various types of buildings and testing configurations. Specifically, SMAPE scores indicate a reduction in prediction error, with improvements ranging from 1.5 % to 50 % over basic transformer, LSTM and ANN models while the higher R² values further confirm the model's reliability in capturing energy time series variance. The 8 % improvement in training time over the basic transformer model, highlights the hybrid model computational efficiency without compromising accuracy.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"20 ","pages":"Article 100489"},"PeriodicalIF":9.6,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143464875","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Municipal heat planning within The World Avatar","authors":"Yi-Kai Tsai ,&nbsp;Markus Hofmeister ,&nbsp;Srishti Ganguly ,&nbsp;Kushagar Rustagi ,&nbsp;Yong Ren Tan ,&nbsp;Sebastian Mosbach ,&nbsp;Jethro Akroyd ,&nbsp;Markus Kraft","doi":"10.1016/j.egyai.2025.100479","DOIUrl":"10.1016/j.egyai.2025.100479","url":null,"abstract":"<div><div>This paper presents a novel integration of building energy simulation with The World Avatar (TWA), a dynamic knowledge graph and agent-based framework designed for comprehensive and interoperable digital representation of the world. The study addresses the imperative for accurate and granular building energy data in energy planning scenarios. By leveraging knowledge graph, agents within TWA replace default assumptions in simulation tools with real-time and location-specific input data, such as building geometry, usage, weather, and terrain elevation. This integrated approach automates the simulation process, enabling agents to retrieve input data, execute simulations, and update the knowledge graph with results in a consistent format. To demonstrate this approach, we developed a simulation agent using the City Energy Analyst. Validation against external datasets from Germany and Singapore shows that the agent significantly improves simulation accuracy. The study also highlights the challenges in data acquisition and processing for municipal heat planning, aligning with the requirements of the German Heat Planning Act. Using Pirmasens, a mid-sized city in Germany, as an example, we demonstrate the practical applicability of the agent in municipal heat planning by providing highly granular data on the heating demands and the solar potentials for heat generation. An accompanying economic analysis further evaluates the cost implications and energy storage requirements associated with the installation of solar collectors, and identifies zones in the city with high solar suitability. These insights enable data-driven decision-making, showcasing the potential of this integrated approach to support municipal heat planning.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"20 ","pages":"Article 100479"},"PeriodicalIF":9.6,"publicationDate":"2025-02-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143429796","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}