This paper introduces a novel distributed data-driven control strategy to achieve robust voltage regulation in active distribution networks characterized by switching topologies. The proposed control scheme mitigates voltage violations by coordinating photovoltaic inverters and battery energy storage systems, accounting for uncertainties in the system model and photovoltaic generation/load variations. In our control design, a set-membership identification approach is firstly developed to estimate the parameter uncertainty set through recursive polytopic updating and adaptive disturbance bounding. Based on the obtained uncertainty set, the voltage regulation problem is formulated as a robust optimization problem, considering worst-case parameter deviations and prediction errors. To facilitate online implementation, this robust optimization problem is transformed into a tractable quadratic programming problem using convex relaxation and duality theory. To achieve better scalability and privacy, the entire control approach is implemented in a distributed manner using a consensus-based alternating direction method of multipliers, enabling decentralized computation with limited information exchange. Different from existing distributed data-driven voltage control methods, the proposed approach simultaneously handles topology-induced model uncertainty, PV/load prediction uncertainty, and bounded disturbances within a unified distributed framework. Case studies on a modified IEEE 34-bus distribution feeder show that the proposed method maintains all bus voltages within the prescribed range under topology switching and uncertainty of photovoltaic and loads, achieving zero voltage violations, whereas benchmark methods lead to 25, 61, and 151 violations, respectively. The results also demonstrate comparable voltage regulation performance to the centralized version while offering improved scalability and privacy preservation.

To address the challenges of low control accuracy and profitability in renewable energy charging stations participating in auxiliary services, caused by the uncertainties of electric vehicles (EVs) and renewable energy, this paper proposes a learning to predictive scheduling for orderly charging of EVs. To reduce the scheduling cost for charging stations engaging in auxiliary services and enhance participation enthusiasm among different types of EVs, a price incentive mechanism of multi-charge mode is introduced into the day-ahead optimization model. This mechanism is formulated based on charge and discharge energy boundaries and incentive response analysis. Furthermore, a dynamic tracking control method based on PD-iterative learning model predictive control (PDILMPC) is proposed to achieve precise tracking of the day-ahead optimal targets. To improve stability and mitigate power fluctuations during the EV power control process, a control priority coefficient is incorporated into the optimization function of the PD-ILMPC. Finally, a wind-photovoltaic-energy storage charging station is used as a case study to demonstrate the effectiveness and optimality of the proposed method. Simulation results show that the proposed method achieves accurate control and ensures the economic operation of the charging station.

The existing generative coding of distribution grids modeling and optimization face several issues, like complicated usage or high auto-codes error rates. This paper proposes the OptDisPro, a novel LLM-based multi-agent framework, enabling automatic optimal power flow (OPF) script modeling and solving. Driven by interactive linguistic instruction, it realizes automatic coding for customized requirements and flexibly adaptive heuristic optimization. Specifically, domain expertise and example scripts are encoded into structured prompt sequences to guide OptDisPro and enhance reasoning via Chain-of-Thought (COT). To mitigate LLM hallucinations, a contextual feedback mechanism is introduced, which collects error messages from the run-time environment for self-correction. Furthermore, Adaptive Selection of Multiple Algorithms (ASMA) is applied in the solving process, flexibly selecting heuristic algorithms to decrease the possibility of local optima. According to cases verification in multiple scenarios, the simulation results demonstrate the effectiveness and stability of OptDisPro in OPF problem of distribution network. The results also encourage further exploration of LLM applications in script online self-updating, autonomous OPF problem-solving and intelligent operation within distribution grid.

With the surge of distributed photovoltaics (PV) and electric vehicles (EV), distribution networks face severe bidirectional voltage challenges. Although distributed energy storage systems (ESS) are widely used to enhance grid resilience, extreme failure conditions still occur. To ascertain the true hosting limits, this paper proposes a graphical hosting capability assessment method for PV-EV-ESS integration. First, we analyze the voltage-energy deadlock of ESS under extreme scenarios and establish criteria for quantifying various conditions. We then introduce an improved adaptive quadtree algorithm, breaking the computational bottleneck of the point-bypoint method to achieve efficient and precise tracing of physical hosting boundaries. Second, the graphical approach intuitively quantifies the limitations of single ESS regulation, revealing three major evolution laws: the 'boundary translation' of initial SOC, the 'bottleneck switching' of capacity/power, and the 'spatial attenuation' of integration locations. Finally, introducing external coordination is demonstrated as an effective path to break ESS physical bottlenecks and enhance ultimate hosting capability. The proposed method overcomes both the lack of global perspective in traditional numerical assessments and the physical errors inherent in mathematical approximations when depicting complex boundaries. It achieves accurate, full-domain two-dimensional visualization of multi-state regions.

The large-scale integration of distributed photovoltaics (PV) and electric vehicle charging stations (EVCS) introduces severe congestion and voltage violations in urban distribution networks. Traditional planning methods heavily rely on physical grid expansion and centralized mathematical optimization, which fail to reconcile the inherent "zero-sum" economic conflicts among the Distribution System Operator (DSO), PV developers, and EVCS operators. To address this challenge, this paper proposes a novel cognitive-physical multi-agent collaborative planning framework empowered by Large Language Model (LLM). First, a generative multi-agent system is constructed to represent the diverse stakeholders. By leveraging the semantic understanding and logical reasoning of LLMs, the agents can dynamically perceive planning scenarios and user intents, auto-generating game-theoretic weights. Second, in the physical layer, an improved genetic algorithm with minimum spanning tree constraints (MST-IGA) is proposed, integrating real-world road-right-of-way data and time-series power flow simulations to ensure engineering feasibility. Furthermore, a "Cognitive Awakening" mechanism based on Non-Wires Alternatives (NWA) is introduced. Under extreme constraints, the LLM agents autonomously negotiate to deploy Battery Energy Storage Systems (BESS) rather than curtailing capacity, effectively transforming the investment deadlock into a positive-sum game. Case studies demonstrate that the proposed framework dynamically adapts to various scenarios. Compared to conventional heuristic methods, the LLM-driven multi-agent gaming strategy significantly reduces the DSO's physical capacity expansion costs, maximizes the PV hosting capacity, and provides highly explainable consensus blueprints for modern power system planning.

Constructing a high-resolution electricity carbon emission factor (ECEF) accounting model is the foundation for the low-carbon transition of new-type power systems. Existing research on zoning methods for carbon emission factor accounting remains insufficient, and green Power Purchase Agreements (PPAs) face the risk of distorted green electricity rights accounting when transmission channels are congested. To address this issue, this paper proposes a Time- and Region- Specific ECEF accounting model that decouples the physical base-map from the market floating-layer. First, based on multidimensional features such as grid impedance and generation-load characteristics, spatial spectral clustering is performed to reconstruct the physical responsibility base-map, thereby forming Carbon Isomorphic Zones (CIZs). Second, by incorporating PPA data, the market contract floating-layer is constructed to delineate Virtual Trading Synergy Zones (VTSZs). Furthermore, a dynamic reachability verification model based on modeled AC power-flow results is introduced to adjust contracted green-electricity attributes under channel-congestion conditions, and the deficit load is returned to its corresponding base-map region for accounting. The case study analysis demonstrates that the proposed model successfully aggregates nodes with high carbon similarity into the same zone and quantifies the spatial reachability range of green electricity environmental attributes. This provides a novel approach for refined electricity-carbon synergy and a more balanced allocation of carbon responsibilities.

The large-scale integration of electric vehicles (EVs) into the power grid provides new flexibility resources for renewable energy accommodation. However, the randomness of their charging behavior and the uncertainty of renewable energy output pose challenges for aggregators participating in the electricity market. This paper proposes a two-stage scheduling of EV aggregators in electricity markets aimed at promoting renewable energy accommodation. First, a generalized energy storage aggregation model for EV clusters is constructed based on Monte Carlo sampling and Minkowski addition to accurately characterize the schedulable potential. Second, a bi-level optimization model for the day-ahead market is designed, considering the responsibility weights for renewable energy accommodation. Finally, in the real-time market stage, a rolling optimization model with the objective of minimizing future comprehensive operational costs is designed to dynamically correct the day-ahead plan and cope with the uncertainties of renewable energy and user behavior. Simulation results show that the proposed strategy can effectively maintain the grid's peak-to-valley difference ratio at an optimal 38.8%, avoiding the severe demand peaks caused by single-mechanism designs. Furthermore, it achieves a 39% relative improvement in the renewable energy accommodation rate compared to conventional price-guided orderly charging, achieving a synergistic optimization of aggregator economic benefits and clean energy accommodation goals while ensuring the safe and stable operation of the power grid.

Current approaches to enhancing system resilience, which prioritize the asset allocation and emergency scheduling optimization, may face challenges related to inadequate asset availability and insufficient, inequitable response capabilities. To address these challenges, this paper designs an insurance mechanism to enhance the resilience of power distribution systems against extreme events. The strategic behavior of the insurer regarding insurance provision and asset aggregation is modeled under a risk-based decision-making framework to maximize the effectiveness of the insurer mechanism. A cooperative framework is formulated to incentivize the emergency service provision from the V2G-ca-pable vehicle aggregator (V2GA). The optimal insurance provisions are determined by addressing the user-insurer-V2GA negotiation process through a tailored two-level nested alternating direction method of multipliers (ADMM) algorithm. Specifically, the outer ADMM coordinates user–insurer contract negotiation, the inner ADMM handles insurer-V2GA coordination, and the Nash bargaining theory is employed for profit allocation between the insurer and the V2GA. Numerical results verify that the proposed insurer optimization approach could offer a promising solution to enhance system resilience against extreme events.

In this letter, we consider the problem of keeping the system frequency of the off-grid networked microgrid system (NMG) within a safe set during the whole control process. The control problem is formulated as a state-dependent quadratic program using control barrier functions (CBFs). By introducing a set of auxiliary variables, the proposed controller can be implemented in a distributed manner while satisfying the same safety constraints at all times. Disturbance observers are also introduced in the control framework to eliminate the impact of external disturbances on safety guarantees established by CBFs.

The control characteristics of active distribution system (ADS) establish a fundamental dependency on information and communication technology (ICT), which will evolve into cyber-physical active distribution systems (CPADS). Consequently, evaluation methodologies incorporating the impacts of ICT enable accurate assessments of the reliability and economic performance in CPADS. This paper proposes a novel cyber-physical cooperative planning method for CPADS to address critical gaps in existing research. ICT strategies in cyber domain and ADS operation strategies in physical domain are organically unified. A multi-scenario contingency method is used to quantify reliability impacts of smart terminal-assisted reconfiguration with outcomes reverse-projected into the planning results. Experimental validation on modified IEEE 33bus distribution systems confirms the effectiveness of our approach, demonstrating that the co-planning strategy achieves the most significant reduction in total planning costs. Key improvements include a substantial reduction in outage duration through smart switch configuration, decreased network losses and mitigated voltage deviations via ESS and PV smart terminals. The proposed planning method provides a comprehensive solution for enhancing the reliability and economy of CPADS planning.

The weak grid condition of distribution systems prevalent in remoted areas significantly constrains the enhancement of renewable energy (RE) hosting capacity (HC) and total supply capability (TSC). Coordinated charging dispatch of abundant electric vehicles (EVs) resources in demand side is a promising solution for this challenge. Thus, an ordered charging method of EV groups is proposed to improve the photovoltaic (PV) HC and TSC of the regional distribution systems in weak grid condition. A cost driven temporal-spatial EV coordinated charging dispatch model is established to proactively shift EV charging behaviors across different time and areas, so as to expand the PV HC and TSC of the target area without requiring upgrades to fixed infrastructure within the distribution system. Compared to traditional approaches relying on flexible interconnection for economical load transfer, the proposed method demonstrates superiority in PV HC expansion and TSC promotion with less cost, while safe operation of the distribution system can be maintained.

To address the security risks caused by the disorderly charging of large-scale electric vehicles to the power grid and the delayed handling of power grid security risks due to the slow response and complex calculation of traditional scheduling methods, an EV adaptive charging management with embedded operation regions is proposed. The load of electric vehicles and the output of distributed power generation are considered state variables for security analysis, and an operation regions model is established, taking into account capacity and voltage constraints. The state variables of electric vehicles and other loads are monitored in real-time, and the method of power flow simulation calculation is used to determine whether there is an over-limit situation. According to the over-limit situation of the operation regions, the event-triggered alternating direction method of multipliers decomposes the optimization problem and distributes it to the EV terminals within their regions, achieving local and rapid handling of power grid security risks. An example simulation using the IEEE 33-node system is conducted, and the results verify that the proposed method for electric vehicles can achieve local and rapid accurate handling of power grid security risks.

Flexible loads, such as thermal loads and electric vehicles (EVs), act as Virtual Energy Storage Sources (VESS) in microgrid scheduling to enhance economic performance via load shifting. Due to centralized optimization methods face scalability and privacy limitations in flexible loads scheduling, distributed methods like Alternating Direction Method of Multipliers (ADMM) provide scalable alternatives. However, performance of ADMM is highly sensitive to the penalty parameter, which must be tailored to the problem's intrinsic characteristics. This dependency necessitates complex and labor-intensive manual tuning. To address this issue, this paper proposed a novel parameter tuning mechanism that leverages large language models (LLM). Under a domain-specific crafted prompt, LLM dynamically refines the penalty parameter to improve convergence efficiency and adaptability, reducing the reliance on expert knowledge and manual intervention. Moreover, a high-resolution temperature constraint is established through Predicted Mean Vote (PMV) method to ensure user thermal comfort during scheduling interval. Simulation results demonstrate: 1) Compared to fixed-parameter and heuristic-based tuning methods, the LLM-assisted parameter tuning method accelerates ADMM convergence by up to 50% (with average improvements exceeding 30%), 2) the method's effectiveness is validated across diverse LLMs, and its high performance is critically dependent on trial-and-error search process, and 3) the integrated indoor temperature constraint ensures thermal comfort compliance while reducing microgrid operational costs through VESS-enabled load shifting and price optimization.

Highlights V2G technology is employed in optimal charging coordination. MDP is employed for optimal scheduling of charging coordination of electric vehicles. DQN is employed for multi-cycle global decision optimization.Abstract To further enhance the active participation of electric vehicles in grid interaction and reduce the decision-making costs for electric vehicle aggregators, this paper addresses the challenges in current EV charging and V2G (Vehicle-to-Grid) management. Considering the owners' willingness to participate, an optimal charging and V2G model for EV charging stations based on a Deep Q-Network is established. The paper analyzes in detail the mutual influence between the level of EV owner participation and the strategies of EV aggregators. Based on the owners' willingness and the physical constraints of the EVs, an evaluation metric for EV participation in charging scheduling is developed. The Deep Q-Network is employed to make decisions regarding EV participation, thereby enhancing the decision-making capability of the EV aggregator, reducing the instability of its scheduling plans, and improving the reliability of these plans. Simulation results demonstrate that this method can dynamically consider EV owners' willingness to participate, adaptively optimize the scheduling margin ratio, make global decisions across multiple time periods, and formulate charging and V2G scheduling plans for the EV aggregator.

The proliferation of distributed energy resources introduces multi-source uncertainties, including implicit uncertainties arising from third-party operators’ partial observability of security constraints, challenging traditional distribution network planning methods dependent on model simplification and predefined scenarios. We address this gap via an adaptive hierarchical learning architecture that co-optimizes distributed energy resources location, capacity, and operational strategies data-drivenly, enabling autonomous learning of implicit constraints without full model knowledge. Our framework embeds a bi-level Stackelberg structure where Monte Carlo Tree Search autonomously generates planning schemes at the upper level, while multi-agent reinforcement learning directly learns operational policies from real-time data at the lower level under partial observability. Validation on both benchmark and large-scale practical distribution systems shows lower investment costs and faster solutions while maintaining voltage stability, demonstrating superior scalability and adaptiveness to implicit uncertainties versus scenario-based methods. Yue Xiang and colleagues propose an adaptive hierarchical learning framework to address implicit uncertainties in distributed energy resource planning when third-party operators lack full grid visibility. Their method co-optimizes investment and operational decisions data-drivenly, reducing costs while maintaining voltage stability without relying on predefined scenarios or model simplifications.

The growing influence of Demand Response (DR) creates heterogeneous electric vehicle (EV) charging loads by intertwining controlled responses with natural user behavior, challenging conventional forecasting. This paper proposes a two-stage decomposition framework to physically disentangle these components. The first stage separates the natural load patterns from a high-frequency signal containing DR-induced dynamics, which is then refined in the second stage. An enhanced CNN-BiLSTM model is tailored to forecast these distinct components. Experiments show the framework achieves state-of-the-art accuracy . More significantly, this work pioneers the use of component-wise forecasting as a diagnostic tool. We demonstrate how the separated forecasts of ”natural load” versus ”DR-induced load” can support the dynamic hosting capacity assessment of distribution grids, transforming prediction into an instrument for structural analysis.

In recent years, the evolution of Li-ion battery material components, cell architectures, and application scenarios has posed significant challenges for the rapid adaptation of battery management systems (BMS). Accurate health diagnostics and prognostics are fundamental to reliable battery operation. However, traditional approaches based on empirical equations, physical models, or handcrafted features often suffer from limited generalization, heavy data demands, and time-consuming development. Representation learning, a major advancement in deep learning, is emerging as a powerful tool to accelerate battery health modeling. Under novel chemistries and unseen operating conditions, it mitigates data scarcity through generative learning and enables rapid model adaptation via transfer learning, which was overlooked in earlier reviews. We systematically summarize representation learning architectures tailored for battery data, highlight their applications in data augmentation and cross-domain transfer, and further identify key challenges and future opportunities in data privacy, multimodal information integration, and model interpretability. Overall, representation learning establishes a solid foundation for the efficient development of next-generation intelligent BMS.

With the global pursuit of carbon neutrality, Virtual Power Plant (VPP) is becoming essential for power system flexibility and decarbonization. Carbon trading mechanisms reshape generation marginal costs through carbon prices and emission allowance constraints, significantly increasing the complexity of VPP bidding decisions. To address this issue, a novel carbon-aware bidding strategy for VPP participation in the electricity spot market is proposed. In the proposed strategy, an output-dependent carbon emission factor model is incorporated to characterize the carbon-emission features of conventional generating units and analyze its impact on the spot market. Then, an electricity spot market framework incorporating carbon constraints is constructed, based on which a bi-level bidding model for the VPP is developed, where the upper level maximizes VPP economic benefits and the lower level simulates the market-clearing process. To address issues such as incomplete information in the spot market, a multi-agent soft actor-critic algorithm is introduced to solve the game process between the VPP and other market participants. This approach enables explicit modeling of multi-agent interactions, thus improving strategy optimization efficiency. Various case studies have been conducted to demonstrate that the proposed model can effectively enhance VPP profits under the carbon trading mechanism while reducing system carbon emissions.

With the increasing penetration of renewable energy sources (RES) in distribution network, the coordination of demand-side electric vehicle (EV) becomes crucial for RES accommodation. By introducing the concept of customer directrix load, this paper proposes a novel closed-loop corrective strategy to enhance the effectiveness of EV demand response. Firstly, a robust composite customer directrix load (CCDL) model is constructed to enhance resilience against uncertainties in EV responses, which integrates control characteristics and response fluctuations to provide tailored full-time guidance for heterogeneous EVs. On the basis, considering that the implementation of demand response is a sequential coupled process, a closed-loop correction mechanism is further proposed to facilitate the multi-stage coordination across RES prediction, CCDL optimization, and deviation feedback. By developing a direction-aware combined prediction model to cope with asymmetric prediction effects, and designing a policy-based gradient descent method for closed-loop refinement, the proposed mechanism enables self-correction against uncertain deviations for demand response implementation. Numerical comparison using real-world case demonstrates that the constructed CCDL model could effectively promote the EV-coordinated RES accommodation, while the proposed deviation correction mechanism can improve the demand response accuracy by 17.88 % over traditional sequential mode.