China's dual-carbon targets necessitate a transition toward a greener, safer, and more efficient energy system; however, substantial disparities persist across provinces. This study evaluates high-quality energy development across 30 Chinese provinces (2011-2022) under the dual-carbon agenda and identifies differentiated transition pathways. Using a PCA-TOPSIS framework with regional pattern classification, we find an "east-high, west-low, central-dip" spatial structure and a nationwide improvement trend over time. Beijing and Guangdong remain persistent leaders, whereas the central region is the primary weak link. Green energy and energy innovation are the strongest contributors to provincial performance, highlighting the importance of clean supply and technological capability. Policy implications emphasize differentiated approaches: strengthen innovation leadership in the east, accelerate structural upgrading and clean substitution in central and resource-dependent provinces, and improve infrastructure and market integration to unlock renewable advantages in the west.

The increasing integration of renewable energy requires wind-solar-storage systems to actively participate in spot markets, but traditional trading methods struggle with the dual uncertainties of generation and price volatility. To address this, we propose a novel day-ahead trading optimization method using a Behavior Cloning-enhanced Multi-Agent Proximal Policy Optimization (BC-MAPPO) algorithm. First, we establish a comprehensive bi-level model for bidding and market clearing that balances operational constraints with practical market rules. Next, by utilizing historical high-quality trading trajectories to initialize the policy network, our BC-MAPPO framework resolves the cold-start problem, accelerates training convergence, and prevents agents from falling into local optima. We then internalize this model as a multi-agent Markov game to seamlessly optimize high-dimensional bidding actions. Extensive validations on a modified IEEE 30-bus system confirm that the integrated algorithm adaptively captures price signals to execute superior arbitrage strategies, significantly enhancing economic returns even under extreme market volatility.

Data elements, as a new quality of productivity, can create significant value for society. In the era of big data, equitable pricing of data elements is crucial for realizing value. Given their unique characteristics, distinct from traditional production factors, traditional pricing methods are no longer viable. Therefore, we provide an overview of the relevant contents of data elements and pricing methods. Secondly, we consider the special characteristics of electric power data in the context of China’s ‘dual-carbon’ and summarize the progress of data element pricing research in the energy and electric power industry. Finally, we put forward the development trend of the energy and electric power industry’s data elements from the perspective of the problems and the development direction.

Strategic bidding for wind–battery hybrid systems is increasingly critical as electricity spot markets transition toward market-oriented mechanisms, particularly in Chinese pilot regions. However, dual uncertainties—wind generation variability and volatile locational marginal prices (LMPs)—expose market participants to significant financial tail risk. This study develops a risk-constrained reinforcement learning framework for optimal bidding of wind–storage hybrid systems. We employ soft actor–critic (SAC) for continuous action control and integrate conditional value-at-risk (CVaR) into reward design to explicitly penalize low-probability, high-loss outcomes. The framework incorporates realistic operational constraints, including linearized battery degradation costs and a market-compatible single-bid abstraction for hourly settlement. Using one-year historical operational data from a 150 MW wind farm (with a 91-day test period), we find that storage integration increases annual profit by 108.4–114.2% relative to wind-only operation. Critically, the SAC–CVaR policy (η = 0.35) preserves 97.3% of risk-neutral profit ($7.71 M vs. $7.93 M) while substantially mitigating downside risk: CVaR@95% improves by 42.4% (−$549 vs. −$952) and VaR@95% improves by 30.1% (−$275 vs. −$393). The trained policy achieves sub-millisecond inference (0.262 ms per decision, ~3820 decisions/s), corresponding to a 3.8 × 104–5.7 × 104× speedup over optimization-based solvers (10–15 s per decision), enabling real-time deployment. Behavioral analysis reveals that the agent learns adaptive, forecast-normalized bidding strategies with more conservative reporting in high-price regimes and counter-cyclical battery dispatch patterns, demonstrating effective coordination between profitability and risk control under volatile market conditions.

With the proliferation of electric vehicles (EVs), accurate charging demand and station occupancy forecasting are critical for optimizing urban energy and the profit of EV aggregators. Existing approaches in this field usually struggle to capture the complex spatio-temporal dependencies in EV charging behaviors, and their limited model parameters hinder their ability to learn complex data distribution representations from large datasets. To this end, we propose a novel EV spatio-temporal large language model (EV-STLLM) for accurate prediction. Our proposed framework is divided into two modules. In the data processing module, we utilize variational mode decomposition (VMD) for data denoising, improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) for data multi-frequency decomposition, and fuzzy information granulation (FIG) for extracting multi-scale information. Additionally, ReliefF is used for feature selection to mitigate redundancy. In the forecasting module, the core architecture is implemented based on a GPT-2 (117M parameters) backbone. By leveraging this Large Language Model (LLM) as the foundational learner, the model benefits from extensive pre-trained knowledge and massive parameter expansion, which allow it to capture complex patterns more effectively than traditional neural networks. Firstly, we fully capture the intrinsic spatio-temporal characteristics of the data by integrating adjacency matrices derived from the regional stations network and spatiotemporal-frequency embedding information. Then, the partially frozen graph attention (PFGA) module is utilized to maintain the sequential feature modeling capabilities of the pre-trained large model while incorporating EV domain knowledge. Extensive experiments using real-world data from Shenzhen, China, verify the model's effectiveness for short-term forecasting at 3, 6, and 9-step horizons. The proposed EV-STLLM demonstrates superior performance compared to state-of-the-art benchmarks, consistently maintaining the lowest MAPE for EV charging demand below 0.20 across two datasets.

With the continuous growth of renewable installation capacity, power system operation is shifting from single-station forecasting to cluster-level coordinated forecasting. To address the limitation that conventional temporal models cannot fully exploit inter-station spatiotemporal coupling, this paper proposes a spatio-temporal graph neural network (STGNN)-based forecasting method for hybrid wind-solar clusters. Multiple stations are modeled as graph nodes. A fused adjacency matrix is constructed by integrating statistical correlation and station-type prior knowledge. Spatial dependencies are extracted by graph diffusion, while temporal dynamics are modeled by a GRU encoder, enabling multi-node multihorizon joint forecasting. Using the public GEFCom2014 dataset, a 13-node hybrid cluster (10 wind nodes and 3 PV nodes) is built, and case studies are conducted in MATLAB R2024a. Results show that for forecasting the next 4 hours using the previous 24 hours, the proposed method outperforms Persistence, LSTM, and GRU baselines on MAE, RMSE, and R2. Specifically, MAE and RMSE are 0.1054 and 0.1482, reduced by 20.41% and 26.14% compared with Persistence. Ablation analysis further confirms the contribution of spatial graph modeling. The method provides a reproducible technical route for rolling renewable forecasting and dispatch support.

ObjectivesTo address challenges of short-term electricity price prediction arising from high penetration of renewable energy into the grid, this study proposes a multi-modal, multi-task short-term electricity price prediction model, EP-VLM, based on a vision language model (VLM).MethodsThe model integrates three modules: a patch memory augmenter, a frequency-domain convolutional encoder, and a structured text embedder. These modules process temporal, visual, and textual multi-modal features, respectively, and convert them into tokens that can be embedded into the VLM. By extracting multi-modal fused features, the model can comprehensively capture the complex factors influencing electricity prices.ResultsExperiments based on real data from a provincial electricity market in northern China demonstrate that EP-VLM significantly outperforms existing time-series prediction models across different prediction horizons and exhibits strong few-shot learning capability.ConclusionsThe model effectively improves electricity price prediction precision, validating its accuracy and robustness in ultra-short-term and day-ahead electricity price prediction.

Accurate photovoltaic (PV) power forecasting is essential for reliable grid dispatch and renewable energy integration, yet it remains challenging because PV generation is jointly shaped by weather variability, day-night transitions, regime-dependent dynamics, and strict physical constraints. We propose PARA-PV, a Physics-Aware Retrieval-Augmented framework that embeds physical knowledge throughout the forecasting process. The framework first encodes multivariate PV observations into patch-level representations and, through a physics-aware retrieval-augmented learner, retrieves historical patches and analog trajectories that are consistent with the current window in temporal shape, power level, PV operating state, and intra-day period; this yields a physically grounded base forecast. To supplement local memory with broader temporal knowledge, the base forecast is then calibrated against a frozen Chronos time-series foundation-model prior through a lightweight residual adapter, so that general temporal regularities are adapted to PV-specific dynamics without overriding the physically grounded prediction. Because residual conditional distribution shifts persist when weather and diurnal regimes change, a physics-aware distribution shift correction module subsequently adjusts the preliminary forecast using power, weather, timestamp, and day/night conditions, applying gated mean-shift and scale corrections selectively. Finally, a physics-constrained loss function partitions the samples into peak, ramping, night-time, and regular regimes and adaptively reweights their error contributions, preventing the dominant regular regime from suppressing learning of operationally critical states. Our code is available at https://github.com/weican1103/PARA-PV.

With the continuous advancement of the construction of new power systems with new energy as the main body, the demand for power grid regulation has increased significantly, and energy storage plays a vital role in the system balance process. In this context, aiming at the problem that the current spot time-of-use bidding trading mode has low adaptability to energy storage and the energy storage regulation potential is difficult to stimulate fully, this paper designs an energy storage participation spot market trading mode based on flexible energy block trading theory. It constructs a two-stage clearing algorithm of flexible energy block day-ahead clearing-intra-day flexible adjustment. By declaring more flexible energy block targets, the market clearing and flexible adjustment of energy storage regulation ability are realized. By ensuring the income of energy storage subjects, the energy storage regulation potential is further stimulated, and the overall market-clearing benefit is improved. Furthermore, the advantages of the proposed transaction mode in improving the flexibility of energy storage regulation and ensuring the income of energy storage are verified by simulation, which provides a practical reference for constructing a power market mechanism adapted to energy storage participation.

As a key carbon-emitting industry, the steel sector urgently requires a carbon emission factor model that can reflect the dynamic operation characteristics of its multi-energy system and the structure of the electricity market for carbon emission optimization. To address these limitations, this paper proposes a time-of-use carbon emission factor modeling method for integrated electricity-heat-hydrogen systems based on energy-carbon flow coupling relationships. The model establishes carbon flow allocation mechanisms for various energy production and storage devices within the industrial park and incorporates the carbon emission attributes of electricity purchased from external markets. By disaggregating the power output into three pathways—electricity, heat, and hydrogen—the number of decision variables on the supply side increases from 120 to 360 under an hourly resolution. Building on this, a bi-level optimization scheduling model with source-load coordination is developed. The upper level minimizes the operational cost of the park, while the lower level minimizes carbon emissions. By leveraging time-varying carbon emission factors to guide the temporal response of multi-energy loads, the model achieves coordinated optimization of economic efficiency and carbon reduction. Case study results demonstrate that the time-of-use carbon emission factor curve can accurately characterize the time-varying characteristics of system carbon intensity. After introducing time-of-use carbon emission factors, the system operation tends to avoid peak and high-carbon periods, with electricity, heat, and hydrogen loads achieving time-shifted adjustments, and the overall carbon emissions significantly reduced. The cumulative carbon emissions of the system are reduced by 10.8%. Meanwhile, the inclusion of purchased electricity carbon emissions can effectively enhance the integrity of carbon optimization, avoiding the underestimation of carbon responsibility and scheduling deviation, thereby verifying the effectiveness and adaptability of the proposed model.

Accurate day-ahead price forecasting is critical for power market participants in the electricity market. However, the increasing penetration of renewable energy introduces greater complexity and intermittency into electricity price patterns, making accurate day-ahead forecasting a significant challenge. To this end, we propose a novel framework for day-ahead electricity price forecasting, EPformer. Specifically, the proposed framework begins with a two-stage data preprocessing module. Then, the proposed framework adopts an encoder-decoder architecture. It first employs Bidirectional Long Short-Term Memory (BiLSTM) as the temporal encoder to capture sequential dependencies and utilizes Temporal Convolutional Networks (TCN) as the feature encoder to extract data features, respectively. The learned representations and additional data features are then fused and fed into the decoder to generate the final prediction. In addition, the proposed framework is trained using a customized loss function that integrates time-frequency domain features. This strategy replaces the conventional MSELoss training paradigm, which enables the model to effectively capture peak and valley features in electricity price, while also mitigating the inherent autocorrelation in the label sequences under the direct forecasting (DF) paradigm. Finally, we conduct a comprehensive evaluation and validation of the proposed model on two electricity price datasets from Shanxi in China.

Accurate day-ahead electricity price forecasting is vital amidst high renewable energy penetration, which significantly intensifies price volatility and complexity. However, current methods struggle to capture the intricate nonlinear dynamics and multi-factor interactions inherent in modern markets. We propose the Electricity Price Multi-Factor Learning (EPMFL) framework, integrating a two-stage data preprocessing module with a multi-factor collaborative learning strategy. Specifically, the framework employs CEEMDAN for signal decomposition and Mutual Information (MI) for adaptive feature selection to reduce redundancy. For prediction, a collaborative module utilizes MLP, CNN, and LSTM to extract spatio-temporal features, while an MLP shared layer processes exogenous factors to facilitate information exchange. Furthermore, EPMFL incorporates a novel Dynamic Weighted Loss Function (DWLF) that adaptively assigns weights and applies targeted penalties to improve model interpretability and robustness. Empirical validation on datasets from Shaanxi, China, demonstrates that EPMFL consistently achieves superior forecasting accuracy and strong generalization over state-of-the-art models.

Guided by energy security strategies and the objectives of building a new power system, the grid integration of a high proportion of renewable energy and a vast number of diverse demand-side entities has significantly increased the randomness and uncertainty of the power system, leading to profound changes in the balance between supply and demand and in the system's interactive characteristics. This not only results in increasingly tight peak supply capacity but also places higher demands on the flexible response capabilities of various capacity types, such as frequency ragulating capacity, ramping capacity, and reserve capacity. Currently, China's power system remains constrained by the absence of a capacity market mechanism in ensuring capacity response flexibility and the optimal allocation of capacity resources. There is an urgent need to conduct research and design a regulating capacity market aimed at enhancing system flexibility, taking into account the current state of China's power market development. In light of this, this paper investigates the cascading substitution characteristics of multiple capacity types including electric energy capacity, frequency ragulating capacity, ramping capacity, and reserve capacity and proposes calculation methods for the cascading substitution coefficients and demand adjustments. Based on this, the paper designs key mechanisms for a regulating capacity market that accounts for the demand of multiple capacity types, constructs a coordinated clearing model for the regulating capacity market, and proposes a clearing solution method for the regulating capacity market. Furthermore, through case studies, this paper verifies that the designed regulating capacity market by enabling unified bidding and coordinated clearing of electric energy, frequency ragulating capacity, ramping capacity, and reserve capacity can simultaneously ensure system capacity adequacy and flexibility objectives. The research findings presented in this paper provide a reference framework for designing capacity market mechanisms that supplement and optimize the flexibility requirements of various types of capacity response under China's new power system.

Precise electricity price prediction is crucial for power market trading strategies. Unlike traditional point forecasting, probabilistic forecasting better captures price volatility, offering more robust predictive information. However, existing models often overlook multiscale data information, hindering intrinsic feature extraction. Furthermore, non-end-to-end models rely on independent submodels without collaborative optimization, limiting performance. To address this, we present an innovative end-to-end probabilistic forecasting framework for day-ahead electricity prices (EP-Net) that improves forecasting accuracy and robustness. The architecture consists of two sub-modules. In the data decomposition and feature selection module, the data is decomposed into multiple frequency components using improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) to capture features across multiple time scales. Then, extreme gradient boosting (XGBoost) is employed for feature selection to mitigate redundant computations in model training. In the end-to-end probabilistic forecasting module, EP-Net directly achieves probabilistic forecasting of electricity prices. Specifically, we first design a multi-channel information fusion block based on ensemble learning to independently and fully extract time series information. Then, through a gating mechanism, weights learned from the original time series are used to adaptively fuse the outputs of each channel, thereby achieving global representation learning. Finally, a quantile regression loss function with penalty weighting is introduced during training to boost the model’s stability and interpretability towards extreme values. Experimental findings illustrate that EP-Net notably surpasses a variety of baseline models across the two datasets.

To accurately assess the transformation process of China’s power sector, this study addresses the limitations of existing literature, which primarily focuses on macro-level or single-dimensional evaluations while neglecting regional disparities. This study constructs a power transition development index based on four dimensions: clean, low carbon, security, and efficiency. By integrating the Entropy Weight Method with the TOPSIS model, it quantitatively evaluates and compares the power transition development levels of 30 Chinese provinces from 2011 to 2022. This reveals their spatiotemporal characteristics and regional disparities, providing a reference for formulating regional energy policies. Results indicate: (1) The security and clean dimensions collectively account for 72.1% of the total weighting, forming the core pillars of power transition. (2) The synthesis transition scores of 30 provinces range from 0.155 (Anhui) to 0.505 (Sichuan), revealing a spatial pattern where resource-rich western regions lead, while central provinces reliant on energy imports and traditional industrial provinces face greater transition pressures. (3) Provincial power transition capabilities steadily improved during the study period, with the synthesis score rising from 0.213 to 0.295—a 38.5% increase—though significant inter-provincial ranking disparities persist. This research proposes policy recommendations to advance China’s regional power transition, including strengthening policy coordination, promoting differentiated transition pathways, and accelerating technological innovation. These measures will help provinces more effectively implement emission reduction targets.

The rapid growth of EVs and the subsequent increase in charging demand pose significant challenges for load grid scheduling and the operation of EV charging stations. Effectively harnessing the spatiotemporal correlations among EV charging stations to improve forecasting accuracy is complex. To tackle these challenges, we propose EV-LLM for EV charging loads based on LLMs in this paper. EV-LLM integrates the strengths of Graph Convolutional Networks (GCNs) in spatiotemporal feature extraction with the generalization capabilities of fine-tuned generative LLMs. Also, EV-LLM enables effective data mining and feature extraction across multimodal and multidimensional datasets, incorporating historical charging data, weather information, and relevant textual descriptions to enhance forecasting accuracy for multiple charging stations. We validate the effectiveness of EV-LLM by using charging data from 10 stations in California, demonstrating its superiority over the other traditional deep learning methods and potential to optimize load grid scheduling and support vehicle-to-grid interactions.

The integration of wind energy into power grids necessitates accurate ultra-short-term wind power forecasting to ensure grid stability and optimize resource allocation. This study introduces M2WLLM, an innovative model that leverages the capabilities of Large Language Models (LLMs) for predicting wind power output at granular time intervals. M2WLLM overcomes the limitations of traditional and deep learning methods by seamlessly integrating textual information and temporal numerical data, significantly improving wind power forecasting accuracy through multi-modal data. Its architecture features a Prompt Embedder and a Data Embedder, enabling an effective fusion of textual prompts and numerical inputs within the LLMs framework. The Semantic Augmenter within the Data Embedder translates temporal data into a format that the LLMs can comprehend, enabling it to extract latent features and improve prediction accuracy. The empirical evaluations conducted on wind farm data from three Chinese provinces demonstrate that M2WLLM consistently outperforms existing methods, such as Generative Pre-trained Transforme for Time Series (GPT4TS), across various datasets and prediction horizons. The results highlight LLMs' ability to enhance accuracy and robustness in ultra-short-term forecasting and showcase their strong few-shot learning capabilities.

With the increasing penetration of renewable energy and the deepening of power market reforms, building loads, as crucial demand - side resources, play an increasingly vital role in power system operations through their flexible regulation capabilities. This paper proposes an optimization framework for aggregated building loads to participate in electricity markets, utilizing deep reinforcement learning. First, a load aggregation model that incorporates thermal dynamics and user comfort is established. Second, the decision - making process for aggregated building load participants is modeled as a Markov decision process and solved using the Deep Deterministic Policy Gradients (DDPG) algorithm. Third, a multi - agent reinforcement learning architecture with self - attention mechanisms is designed to achieve coordinated optimization control across building clusters. Finally, case studies validate the effectiveness of the proposed approach. Results demonstrate that, compared to traditional methods, this framework enhances aggregator profits while reducing peak - valley load variations, all achieved under maintained user comfort levels.

With the rapid expansion of renewable energy and the formal operation of provincial electricity spot markets in China, smart aggregators that coordinate flexible loads, storage resources, demand-response portfolios, and distributed energy resources increasingly face imbalance-settlement risk across the day-ahead and real-time segments of the electricity spot market. Existing studies often focus on average prices or point forecasts, which may overlook regime persistence, negative-price clustering, and tail exposure in high-frequency price spreads. This paper develops a regime-switching and tail-risk signal framework to characterize and forecast day-ahead–real-time price spreads in the Shandong electricity spot market and to translate these forecasts into risk-aware trading signals for representative smart aggregators. Using 35,136 non-public observations at 15 min resolution provided by State Grid Shandong Electric Power Company for 2024, the spread is analyzed using descriptive statistics, Markov regime-switching models, quantile regression, out-of-sample forecasting, trading-signal backtesting, component ablation, and robustness checks. The spread, defined as real-time price minus day-ahead price, has a mean of −7.50 Chinese yuan per megawatt-hour (CNY/MWh), a median of −0.005 CNY/MWh, 5% and 95% quantiles of −196.84 and 137.65 CNY/MWh, and 1% and 99% quantiles of −372.68 and 338.04 CNY/MWh, respectively. A three-state Markov model identifies negative-deviation high-volatility, near-zero low-volatility, and positive-deviation regimes with multi-hour persistence. In the December out-of-sample test, the upper- and lower-tail quantile signals achieve recall rates of 0.872 and 0.841, respectively, and removing lagged spreads increases mean absolute error (MAE) from 24.015 to 54.278 CNY/MWh. The framework provides risk-warning signals rather than causal identification or realized-profit evaluation.

Accurate probabilistic electricity price forecasting can characterize the uncertainty of electricity price information and provide more comprehensive forecasting information for market traders. However, existing methods struggle to effectively identify and extract factors related to electricity prices, resulting in limited prediction accuracy. To this end, we propose a novel probabilistic electricity price forecasting framework. Specifically, the framework can be divided into two parts. In the two-stage data preprocessing module, we utilize the Isolation Forest algorithm for outlier correction and ReliefF for feature selection to reduce feature redundancy. In the probabilistic electricity price forecasting module, we utilize the Multilayer Perceptron (MLP) to learn trend features from historical price data, while the customized feature convolutional network (FConv) extracts the feature patterns of factors related to electricity prices. Then, the cross-correlation embedding technique fuses these extracted features for a more comprehensive representation. In addition, the patch processing technique partitions the fused data into multiple patches to capture local features. Moreover, the incorporation of positional encoding helps the model to better capture temporal dependencies in the data. Then, we utilize the MLP to generate the final forecasts. Additionally, the proposed framework is trained using an improved loss function with penalty weights, aiming to enhance the model’s capability to predict outliers. Finally, we validated the effectiveness of our proposed model using two electricity price datasets from China.