Energy and AI最新文献

{"title":"Denseception network method for transient stability prediction of power systems","authors":"Dan Liu ,&nbsp;Xia Chen ,&nbsp;Kezheng Jiang ,&nbsp;Wei Ge ,&nbsp;Linfei Yin","doi":"10.1016/j.egyai.2025.100550","DOIUrl":"10.1016/j.egyai.2025.100550","url":null,"abstract":"<div><div>In the context of accelerated global energy transition, the high proportion of renewable energy grid connections and the proliferation of power control devices have significantly increased the tangled and haziness of the electromechanical transients in power grids, and the transient stability prediction has become an international forefront problem in the construction of smart grid security and defense system. However, existing methods face triple limitations: traditional physical models rely on ideal assumptions and are computationally inefficient; shallow data-driven models have insufficient feature extraction capabilities; and existing deep learning methods have poor generalization and lack interpretability. To manage the issues highlighted above, this study proposes a deep learning-based Denseception architecture and its accompanying data modeling method, which achieves a breakthrough in high-precision continuous numerical prediction of transient stability indicator (TSI) with engineering practicality. The heterogeneous multi-scale feature fusion network is constructed by integrating the DenseNet dense cross-layer connectivity, Xception deep separable convolution, and the dynamic weighting mechanism of the fully connected layers, which significantly improves the efficiency of the cross-scale dynamic feature extraction; and the three-channel two-dimensional spatial-temporal feature reconstruction method is innovatively designed, which reconstructs the temporal data of the whole fault process into an image-like structure, and combines with the adversarial training strategy to enhance the cross-topology generalization capability. The experiment reveals that the TSI prediction error of the Denseception model is prominently lower than that of the mainstream deep learning model in the IEEE 39–10 and 145–50 systems, which is the best performance. This study overcomes the contradiction between speed, accuracy, and generalizability of traditional methods, provides a full chain solution for the dynamic security defense of a high percentage new energy power grids, and provides a critical time window for emergency control.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100550"},"PeriodicalIF":9.6,"publicationDate":"2025-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144548851","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}

{"title":"Accelerating sodium-ion electrode material development through AI-driven optimization and predictive modeling","authors":"Sara Alzaabi ,&nbsp;Ali Elkamel ,&nbsp;Georgios N. Karanikolos ,&nbsp;Ali Alhammadi","doi":"10.1016/j.egyai.2025.100537","DOIUrl":"10.1016/j.egyai.2025.100537","url":null,"abstract":"<div><div>Sodium-ion batteries (SIBs) are gaining traction as a cost-effective and sustainable alternative to lithium-ion batteries for large-scale energy storage, due to sodium’s abundance, low cost, and safety advantages. However, the discovery of high-performance electrode materials for SIBs remains a significant challenge because of the complex interactions between compositional and structural features that govern key properties such as specific capacity, average voltage, and volume change. In this study, we present an artificial intelligence (AI)-driven framework that integrates machine learning and multi-objective optimization to accelerate the design of sodium-ion battery electrodes. Four predictive models, namely Decision Tree, Random Forest, Support Vector Machine (SVM), and Deep Neural Network (DNN), were trained on a feature-rich dataset derived from high-throughput computational databases. The DNN model achieved the highest predictive accuracy, with R<span><math><msup><mrow></mrow><mrow><mn>2</mn></mrow></msup></math></span> values up to 0.97 and mean absolute errors (MAE) below 0.11 for the target properties. To support material selection, the DNN was coupled with the Non-dominated Sorting Genetic Algorithm II (NSGA-II) to identify Pareto-optimal materials that maximize specific capacity while minimizing volume expansion. The resulting candidates exhibit balanced electrochemical performance and potential for practical SIB applications. This study demonstrates the power of combining deep learning and optimization to guide the discovery of next-generation energy storage materials with high efficiency and reduced experimental overhead.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100537"},"PeriodicalIF":9.6,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144556996","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}

{"title":"Physics-Guided Memory Network for building energy modeling","authors":"Muhammad Umair Danish ,&nbsp;Kashif Ali ,&nbsp;Kamran Siddiqui ,&nbsp;Katarina Grolinger","doi":"10.1016/j.egyai.2025.100538","DOIUrl":"10.1016/j.egyai.2025.100538","url":null,"abstract":"<div><div>Accurate energy consumption forecasting is essential for efficient resource management and sustainability in the building sector. Deep learning models are highly successful but struggle with limited historical data and become unusable when historical data are unavailable, such as in newly constructed buildings. On the other hand, physics-based models, such as EnergyPlus, simulate energy consumption without relying on historical data but require extensive building parameter specifications and considerable time to model a building. This paper introduces a Physics-Guided Memory Network (PgMN), a neural network that integrates predictions from deep learning and physics-based models to address their limitations. PgMN comprises a Parallel Projection Layers to process incomplete inputs, a Memory Unit to account for persistent biases, and a Memory Experience Module to optimally extend forecasts beyond their input range and produce output. Theoretical evaluation shows that components of PgMN are mathematically valid for performing their respective tasks. The PgMN was evaluated on short-term energy forecasting at an hourly resolution, critical for operational decision-making in smart grid and smart building systems. Experimental validation shows accuracy and applicability of PgMN in diverse scenarios such as newly constructed buildings, missing data, sparse historical data, and dynamic infrastructure changes. This paper provides a promising solution for energy consumption forecasting in dynamic building environments, enhancing model applicability in scenarios where historical data are limited or unavailable or when physics-based models are inadequate.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100538"},"PeriodicalIF":9.6,"publicationDate":"2025-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144597493","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}

{"title":"Graph-based two-level clustering for electric vehicle usage patterns","authors":"Dhanashree Balaram ,&nbsp;Brett Dufford ,&nbsp;Sonia Martin,&nbsp;Gianina Alina Negoita,&nbsp;Matthew Yen,&nbsp;William A. Paxton","doi":"10.1016/j.egyai.2025.100539","DOIUrl":"10.1016/j.egyai.2025.100539","url":null,"abstract":"<div><div>Electric vehicles (EVs) continue to gain popularity over internal combustion vehicles, but driving and charging an EV is a fundamentally different experience. EV usage patterns include features such as charging speed, battery state of charge, and battery depth of discharge. Understanding EV usage patterns from data is essential to optimizing the performance and management of EVs. However, extracting useful conclusions from individual EV data is computationally intensive. Traditional clustering methods offer a solution by grouping vehicles by behavior pattern, but they still pose computational challenges for long time-series data profiles. To efficiently extract EV behavior insights, we propose a scalable two-level clustering approach and test it on a dataset of 3,082 real EVs over the course of a year. We first use level one clustering to weekly data segments, revealing distinct patterns in state of charge utilization. Next, we apply level two graph-based clustering to group individual vehicles that operate similarly on a longer timescale. Our scalable and adaptable clustering approach can aid in battery lifecycle management, charge demand forecasting, and fleet energy management, all critical tasks to facilitate the continued growth of the EV market.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100539"},"PeriodicalIF":9.6,"publicationDate":"2025-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144713440","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}

{"title":"Multi-temporal PV power prediction using long short-term memory and wavelet packet decomposition","authors":"Amirhasan Sardarabadi ,&nbsp;Amirhossein Heydarian Ardakani ,&nbsp;Silvana Matrone ,&nbsp;Emanuele Ogliari ,&nbsp;Elham Shirazi","doi":"10.1016/j.egyai.2025.100540","DOIUrl":"10.1016/j.egyai.2025.100540","url":null,"abstract":"<div><div>The integration of photovoltaic (PV) systems into power grids presents operational challenges due to the inherent variability in solar power generation. Accurate PV power forecasting can help address these issues by enhancing grid reliability and energy management. This study introduces a novel hybrid deep learning approach that combines Wavelet Packet Decomposition (WPD) and Long Short-Term Memory (LSTM) networks to improve forecasting accuracy across multiple time horizons. The proposed model incorporates a dynamic weighting mechanism to optimally integrate the forecasts of decomposed subseries, effectively capturing both high- and low-frequency components of the power signal. Using real-world data from a solar parking site at the University of Twente, Netherlands, the proposed models are compared with standard LSTM, Linear Regression, and Persistence baselines across 15 min, 1-hour, and day-ahead horizons. The WPD-LSTM model with weight optimization reduces nRMSE by up to 72.5%, 52.9%, and 34.7% compared to Persistence, and by 68.6%, 36.1%, and 7.5% compared to standalone LSTM, respectively. These results highlight the effectiveness of the hybrid approach in delivering more accurate and robust PV power forecasts.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100540"},"PeriodicalIF":9.6,"publicationDate":"2025-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144535869","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}

{"title":"Flexible aggressiveness probabilistic optimisation (FAPO) bidding for peer-to-peer electricity trading","authors":"Ioanna Kalospyrou,&nbsp;Timothy Hutty,&nbsp;Robert Milton,&nbsp;Solomon Brown","doi":"10.1016/j.egyai.2025.100543","DOIUrl":"10.1016/j.egyai.2025.100543","url":null,"abstract":"<div><div>The maximisation of renewable energy generation is critical for net-zero aspiring countries around the globe. Local energy markets facilitate the seamless incorporation of energy from distributed renewable energy resources into the electricity network, serving as platforms for trading locally-generated renewable energy between prosumers in residential communities. However, local energy markets’ essential role in distributed energy resource integration is not enough to encourage participation. Prosumers are more likely to join a local energy market if financial incentives are offered. To address this, we present the Flexible Aggressiveness Probabilistic Optimisation (FAPO) bidding strategy for trading electricity within a local energy market aimed at maximising participation incentives. This is formulated as an optimisation problem targeting the maximisation of prosumers’ individual utilities. The FAPO methodology is applied in a simplified local energy market simulation environment, and its results are compared to two other well-established bidding strategies: Zero Intelligence-Constrained and Adaptive Aggressiveness. The results indicate that FAPO achieved a wider range of clearing prices than both Adaptive Aggressiveness and Zero Intelligence-Constrained, incentivising greater prosumer participation. Specifically, FAPO enabled the trading of 1.48 MWh of electricity, compared to 1.34 MWh with Adaptive Aggressiveness and 1.37 MWh with Zero Intelligence-Constrained. Furthermore, FAPO cleared 100% of all asks and 98% of all bids, while the other two strategies cleared approximately 90% of submitted orders. Consequently, FAPO is proven to be an engaging bidding methodology likely to attract more prosumers to local energy markets. This is critical for the successful acceptance, uptake, and widespread application of this financial market type, which is key for smooth distributed energy resource integration into the network.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100543"},"PeriodicalIF":9.6,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144522281","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}

{"title":"Decision-focused fine-tuning of time series foundation models for dispatchable feeder optimization","authors":"Maximilian Beichter ,&nbsp;Nils Friederich ,&nbsp;Janik Pinter ,&nbsp;Dorina Werling ,&nbsp;Kaleb Phipps ,&nbsp;Sebastian Beichter ,&nbsp;Oliver Neumann ,&nbsp;Ralf Mikut ,&nbsp;Veit Hagenmeyer ,&nbsp;Benedikt Heidrich","doi":"10.1016/j.egyai.2025.100533","DOIUrl":"10.1016/j.egyai.2025.100533","url":null,"abstract":"<div><div>Time series foundation models provide a universal solution for generating forecasts to support optimization problems in energy systems. Those foundation models are typically trained in a prediction-focused manner to maximize forecast quality. In contrast, decision-focused learning directly improves the resulting value of the forecast in downstream optimization rather than merely maximizing forecasting quality. The practical integration of forecast values into forecasting models is challenging, particularly when addressing complex applications with diverse instances, such as buildings. This becomes even more complicated when instances possess specific characteristics that require instance-specific, tailored predictions to increase the forecast value. To tackle this challenge, we use decision-focused fine-tuning within time series foundation models to offer a scalable and efficient solution for decision-focused learning applied to the dispatchable feeder optimization problem. To obtain more robust predictions for scarce building data, we use Moirai as a state-of-the-art foundation model, which offers robust and generalized results with few-shot parameter-efficient fine-tuning. Comparing the decision-focused fine-tuned Moirai with a state-of-the-art classical prediction-focused fine-tuning Moirai, we observe an improvement of 9.45% in Average Daily Total Costs.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100533"},"PeriodicalIF":9.6,"publicationDate":"2025-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144655153","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}

{"title":"Transformer-based forecasting for high-frequency natural gas production data","authors":"Saifei Ma ,&nbsp;Tiantian Zhang ,&nbsp;Haibo Wang ,&nbsp;Haoyu Wang ,&nbsp;Nan Li ,&nbsp;Haiwen Zhu ,&nbsp;Jianjun Zhu ,&nbsp;Jianli Wang","doi":"10.1016/j.egyai.2025.100535","DOIUrl":"10.1016/j.egyai.2025.100535","url":null,"abstract":"<div><div>Accurate prediction of natural gas well production data is crucial for effective resource management and innovation, particularly amid the global transition to sustainable energy. Traditional models struggle with high-frequency, high-dimensional datasets generated by digital transformation in the oil and gas industry. This study explores the application of Transformer-based models — Transformer, Informer, Autoformer, and Patch Time Series Transformer (PatchTST) — for forecasting high-frequency natural gas production data. These models utilize self-attention mechanisms to capture long-term dependencies and efficiently process large-scale datasets. Autoformer achieves predictive success through its Seasonal Decomposition Attention mechanism, which effectively extracts trend-seasonality patterns. However, our experiments show that Autoformer exhibits sensitivity to dataset changes, as performance declines when using old parameters compared to retrained models, highlighting its reliance on dataset-specific retraining. Experimental results demonstrate that increasing sampling frequency significantly enhances prediction accuracy, reducing MAPE from 0.556 to 0.239. Additionally, these models consistently track actual production trends across extended forecast horizons. Notably, PatchTST maintains stable performance using either pretrained or retrained parameters, showcasing superior adaptability and generalization. This makes it particularly suitable for real-world applications where frequent retraining may not be feasible. Overall, the findings validate the applicability of Transformer-based models, particularly PatchTST, in dynamic and precise natural gas production forecasting. This study provides valuable insights for advancing adaptive, data-driven resource management strategies.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100535"},"PeriodicalIF":9.6,"publicationDate":"2025-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144480982","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}

{"title":"Enhancing electric vehicle charging load prediction in data-scarce scenarios: A hybrid deep learning-based approach integrating clustering analysis and transfer learning","authors":"Rehman Zafar ,&nbsp;Pei Huang ,&nbsp;Yongjun Sun","doi":"10.1016/j.egyai.2025.100545","DOIUrl":"10.1016/j.egyai.2025.100545","url":null,"abstract":"<div><div>Accurate electric vehicle (EV) load forecasting is crucial for efficient grid operations and demand-side management, yet it is challenging in data-scarce scenarios. Transfer learning (TL) offers a solution by transferring knowledge from data-rich to data-limited scenarios. However, when the knowledge domain exhibits highly diverse behaviors, applying TL alone could introduce large biases, reducing accuracy and limiting its effectiveness. To address this problem, this study proposes a hybrid deep learning-based framework that integrates TL and K-means clustering. The proposed approach consists of two phases. In the source domain phase, a deep-learning-based model is trained using the full dataset and then fine-tuned using clustered user behaviors. In the target domain phase with limited data, TL is applied to transfer knowledge from the source-domain fine-tuned cluster models. For validation, the developed prediction method has been tested using real-world datasets and compared with two other cases: one with applying TL from the source-domain base model trained from full dataset, and one without applying TL. Results show the hybrid method improves forecasting accuracy, reducing the normalized root mean squared error by 3.99 % and 8.22 %, respectively. This study establishes a structured approach for targeted knowledge transfer, enhancing prediction accuracy in data-scarce settings. The framework is scalable and adaptable to other energy forecasting applications, supporting sustainable and resilient energy management.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100545"},"PeriodicalIF":9.6,"publicationDate":"2025-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144501782","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}

{"title":"Enhancing visual feature constraints in segmentation models for photovoltaic panel recognition","authors":"Zhiyu Zhao ,&nbsp;Kangning Li ,&nbsp;Yunhao Chen ,&nbsp;Jinyang Wang","doi":"10.1016/j.egyai.2025.100544","DOIUrl":"10.1016/j.egyai.2025.100544","url":null,"abstract":"<div><div>The integration of remote sensing and artificial intelligence technologies into photovoltaic (PV) power generation has significantly enhanced the efficiency and precision of monitoring and evaluating PV station construction. However, most semantic segmentation models are primarily developed for natural scenes, often neglecting the distinctive visual attributes of PV panels. We introduce a visual feature constraint method designed to tailor the segmentation network to the unique aspects of PV panels, including their texture, color, and shape. The method incorporates a constraint module, comprised of three adversarial autoencoders, into a conventional segmentation model. This technique represents a versatile training framework that can be seamlessly integrated with state-of-the-art models, providing clear insights into the learning process. Experimental results with UperNet, SegFormer, DeepLabV3+, TransUNet, CorrMatch, SCSM and UKAN as baseline models show a maximum IoU improvement of 2.16 %. Notably, UperNet attains the superior segmentation outcomes, whereas DeepLabV3+ exhibits the greatest benefit from the imposed constraints. Furthermore, our findings reveal that various models exhibit distinct sensitivities to different visual features, and employing multiple constraints typically yields better results than relying on single-feature constraints. Collectively, our proposed method showcases its potential to advance PV panel segmentation in remote sensing applications, presenting a scalable and effective solution.</div></div>","PeriodicalId":34138,"journal":{"name":"Energy and AI","volume":"21 ","pages":"Article 100544"},"PeriodicalIF":9.6,"publicationDate":"2025-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144522282","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}