With the large-scale integration of renewable energy into multi-level power grids, the traditional deterministic dispatch framework with separated operation among different grid levels faces challenges such as boundary power mismatches and insufficient reserve capacity. To address these issues, this paper proposes a coordinated robust intra-day rolling dispatch model for multi-level power grids based on renewable energy prediction intervals. By coordinating the operation of transmission, distribution, and microgrids, the proposed model exploits the flexibility potential of large-scale flexible resources on the distribution network and microgrid sides, thereby enhancing renewable energy accommodation. An affine adjustable policy is introduced to transform the original robust optimization model into an equivalent quadratic programming formulation, and a multi-parameter space projection decomposition algorithm is developed to efficiently solve the problem. While preserving the convexity of lower-level optimization problems, each grid level only needs to exchange boundary power information and optimal projection functions to achieve the global optimum. Simulation results on the T118-D33-M4 test system demonstrate that the proposed method improves computational efficiency by approximately 8 and 11 times compared with generalized Benders decomposition and the alternating direction method of multipliers, respectively, while effectively promoting renewable energy accommodation.

Intermittency and volatility of photovoltaic (PV) power generation will cause voltage fluctuations in inverter-based microgrids, which require inverter-based voltage control (IBVC) to maintain the grid-wise voltage magnitudes. Considering limited computational and communication abilities of PV inverters, a number of decentralized reinforcement learning (RL) methods have been proposed. However, existing decentralized RL methods cannot provide persistent safety guarantees, which are crucial for IBVC problems, thus hindering their broader application in real world entities. In this paper, we utilize the reachability analysis as a novel safety certificate to provide persistent safety, which adaptively computes the maximum voltage violation throughout the entire control process to characterize the safe set. Under the safe set characterized by reachability analysis, we formulate the IBVC problem as a novel reachability constrained Markov game, and utilize the state-wise Lagrangian method to handle the introduced reachability constraints. Finally, a reachability constrained decentralized RL method is proposed for IBVC in microgrids, which incorporates the multi-agent soft actor-critic algorithm with the reachability analysis and state-wise Lagrangian method, thereby providing persistent safety. The proposed method significantly improves the applicability of current decentralized RL algorithms. Comprehensive comparisons and numerical simulations on various test cases demonstrate the effectiveness of our proposed method.

This paper proposes a flexible interaction method for integrated electricity-hydrogen-gas systems (IEHGSs) that incorporates the entire green hydrogen supply chain (GHSC) with varying operating characteristics. Firstly, a comprehensive structure framework is developed for the GHSC, which encompasses stage-wise precision modeling and multi-stage coordination across production, transportation, storage, and utilization under varying operating conditions. Subsequently, the interaction mechanism between the GHSC and IEHGSs is explored, revealing the flexibility provided by hydrogen-enriched compressed natural gas (HCNG) energy flow regulation, the line-pack effect, and hydrogen volume fraction dynamics. The coupling relationship between variable hydrogen volume fractions and flexible operation model nonlinearity is analyzed. A solution methodology incorporating dynamic hydrogen volume fractions is developed, featuring a model-reformulation method combined with a two-stage solution strategy. Finally, the validity and adaptability of the proposed model and method are verified using two distinct scales of IEHGSs. Simulation results demonstrate that adapting to the varying operating characteristics of the entire GHSC plays a crucial role in enhancing low-carbon performance and economic efficiency during its flexible interaction with IEHGSs.

Power can be harvested in space today. The real challenge is building, operating and certifying a complete energy system in orbit — under extreme conditions, with no one nearby to repair it. Whether space joins the world’s energy supply will depend on that challenge, not just on generation.

The rapid evolution of energy structures has positioned microgrids as pivotal components of next-generation power systems, offering enhanced resilience and renewable energy integration. However, the inherent low inertia, complex dynamics, and poor model conditions of microgrids necessitate advanced data-driven frequency control strategies. Although reinforcement learning (RL) has demonstrated certain potential and advantages, existing RL methods often struggle with generalization across diverse microgrid configurations and lack adaptability to unseen environments, particularly when explicit system parameters are unavailable. To address these challenges, in this paper, we introduce a novel prompt decision transformer (Prompt-DT) architecture for microgrid frequency control. Unlike traditional approaches that rely on hard-to-obtain environmental characteristic parameters, the proposed method leverages few-shot expert historical trajectories as prompts to guide autonomous perception and adaptive decision-making. In addition, we propose a context-aware training and execution mechanism utilizing self-supervised contrastive learning to enhance environment recognition and prompt utilization efficiency. In addition, a physics-informed prompt design technique that filters prompts based on cumulative reward and frequency volatility is proposed, ensuring high-quality physical guidance during online execution. Finally, to ensure generalization in unseen environments with limited data, we develop a lightweight finetuning approach that achieves performance comparable to full-parameter finetuning with minimal adjustments.

Random wind speed fluctuations can significantly influence a wind turbine's ability to release rotor kinetic energy during frequency regulation. To fully exploit the turbine's adjustable capacity, this paper establishes a dynamic model of a doubly fed induction generator and derives its adjustable capacity considering ultra-short-term wind speed forecasting. The nonlinear states are then projected into a high-dimensional linear space with the Koopman operator, enabling accurate global linearization. Based on this model, a distributed, capacity-aware Model Predictive Control (cVMPC) framework is formulated, where the ‘c’ denotes its core feature: capacity-awareness. To further improve performance, the adjustable capacity parameter is proactively predicted and incorporated into the MPC's control weighting coefficients. This enables an adaptive regulation strategy that explicitly accounts for the predicted kinetic energy reserves of each turbine. The proposed cVMPC approach is validated on a 3-machine, 9-bus system. Simulation results confirm that the strategy alleviates the initial frequency nadir, accelerates rotor speed recovery, and prevents secondary frequency dips.

Grid-forming (GFM) doubly-fed induction generators (DFIGs) are promising solutions to enhance the stability of wind farms under weak grid conditions. They typically adopt power-based synchronization strategies and use voltage/flux inner loops to regulate stator-side characteristics. However, existing inner-loop designs are structurally complex or lack sufficient dynamic performance. Moreover, the coupling between power regulation and synchronization further reduces the speed of power regulation. To address this issue, this paper proposes a rotor-flux-based GFM control strategy for DFIGs. First, a virtual flux oscillator (VFO) is designed to achieve synchronization under weak grid conditions, where both active and reactive power are used to regulate the internal flux vector. Then, a reference-feedforward rotor flux controller is proposed based on the induction machine flux characteristics. This controller has a simple structure, imposes low inner-loop bandwidth requirements and enables fast power regulation. In addition, an automatic power regulator is designed to track either the optimal power command or the dispatched power command. Small-signal analysis and T-S multi-model-based large-signal stability analysis show that the proposed method avoids undamped oscillations and has a wider domain of attraction (DOA). The effectiveness of the proposed control strategy is further validated through hardware-in-the-loop (HIL) experiments.

The growing integration of distributed photovoltaics (PVs) into active distribution networks (ADNs) has exacerbated operational challenges, making it imperative to coordinate diverse equipment to mitigate voltage violations and enhance power quality. Although existing data-driven approaches have demonstrated effectiveness in the voltage control problem, they often require extensive trial-and-error exploration and struggle to incorporate heterogeneous information, such as day-ahead forecasts and semantic-based grid codes. Considering the operational scenarios and requirements in real-world ADNs, in this paper, we propose a hybrid knowledge-data-driven approach that leverages dynamic collaboration between a large language model (LLM) agent and a reinforcement learning (RL) agent to achieve two-stage voltage control. In the day-ahead stage, the LLM agent receives coarse region-level forecasts and generates scheduling strategies for on-load tap changer (OLTC) and shunt capacitors (SCs) to regulate the overall voltage profile. Then in the intra-day stage, based on accurate node-level measurements, the RL agent refines terminal voltages by deriving reactive power generation strategies for PV inverters. On top of the LLM-RL collaboration framework, we further propose a self-evolution mechanism for the LLM agent and a pretrain-finetune pipeline for the RL agent, effectively enhancing and coordinating the policies for both agents. The proposed approach not only aligns more closely with practical operational characteristics but also effectively utilizes the inherent knowledge and reasoning capabilities of the LLM agent, significantly improving training efficiency and voltage control performance. Comprehensive comparisons and ablation studies demonstrate the effectiveness of the proposed method.

Rising distributed energy resources (DERs) penetration in distribution systems boosts grid flexibility for low-carbon transition, and rational DER management can address volatility and ensure better dispatch compliance. To this end, this paper proposes a novel flexibility support method for distribution networks. First, a flexibility evaluation neural network is constructed to approximate the mapping relationship between heterogeneous DERs and the feasible power range at the root node of DS, thereby delineating the feasible boundary of flexibility support. Second, a deep reinforcement learning (DRL)-based algorithm integrated with an enhanced model-based expert module is developed to achieve balanced allocation of the upper-grid injected power among flexible units. To ensure real-time adaptability under abrupt load fluctuations or DER outages, a gradient-based indicator combined with real-time action/value deviation measurement is adopted to dynamically optimize parameter selection in the distribution system operator's (DSO) policy network, facilitating online adaptive adjustments. Numerical simulations carried out on modified IEEE 33-bus and 141-bus test systems demonstrate the effectiveness of the proposed method, with voltage, branch power, and critical generator output deviations constrained within 0.01 p.u., 0 p.u., and 0.048 p.u., respectively. Furthermore, the proposed framework achieves a more favorable trade-off between economic efficiency and operational security, exhibiting the most balanced performance among state-of-the-art DRL algorithms.

The calculation of g-functions is essential for the design and simulation of geothermal boreholes. However, existing methods, such as the stacked finite line source (SFLS) model, face challenges regarding computational efficiency and accuracy, particularly with fine-grained discretization. This paper introduces a novel discretization method to address these limitations. We reformulate the g-function calculation as the solution to spatio-temporal integral equations and employ the Gauss-Legendre quadrature to approximate the spatial integrals, avoiding the assumption of a stepwise structure of heat extraction rate, thereby improving accuracy. This formulation transforms the time-consuming segment-to-segment integral calculations in SFLS model into simpler and analytical point-to-point response factors, which tackles the long-standing issue that higher discretization orders generally reduce computational error at the cost of speed. Furthermore, we identify that the governing integral equations are of the Fredholm first kind, leading to ill-conditioned linear systems that can cause g-function to diverge at high discretization orders. To address this, a regularization technique is implemented to ensure stable and convergent solutions. Numerical tests demonstrate that the proposed method is significantly more efficient, achieving comparable or improved accuracy at speeds 20 to 200 times faster than the SFLS model with optimized nonuniform discretization schemes.

This paper presents a novel and efficient approach to calculate the thermal response function (g-function) of inclined geothermal boreholes. Traditional analytical methods based on the finite line source model require computationally expensive segment-to-segment numerical integration, which becomes particularly challenging for inclined boreholes due to the complex three-dimensional geometry. We propose a semi-analytical discretized formulation that eliminates the need for complicated double integrals by directly discretizing the governing spatio-temporal integral equations using numerical quadrature rules. This approach replaces segment-to-segment response factors with efficient point-to-point coupling factors that have closed-form analytical expressions. Additionally, we develop an intuitive distance aggregation technique that exploits the symmetrical properties of borehole layouts to reduce computational complexity. The proposed method unifies the treatment of vertical and inclined boreholes within a single computational framework without introducing additional complexity for inclined configurations. Numerical validation against the reference implementation of finite line source models demonstrates excellent agreement while achieving significant computational speedups up to three orders of magnitudes times faster than state-of-the-art finite line source methods.

Electricity prices exhibit strong volatility due to stochastic demand fluctuations, renewable integration, and unexpected market events, making probabilistic forecasting essential for risk-aware operational decisions. This paper proposes a reasoning-enhanced probabilistic forecasting framework (RE-Price) based on a parameter-efficient large language model (LLM). The semantic information distillation mechanism extracts compact features from price-relevant news and operational reports. The heterogeneous market information is then fused by cross-modal attention which enables the LLM to jointly interpret temporal dynamics and textual signals. To efficiently adapt the pre-trained model, a parameter-efficient fine-tuning strategy is employed based on quantized low-rank adaptation. Probabilistic forecasts are constructed by generating multiple trajectories through Monte Carlo dropout, where dropout masks remain active during inference. The standard and tail-aware trajectories are combined as a risk-enhanced scenario distribution and then processed by adaptive kernel density estimation, allowing flexible modeling of skewed and fat-tailed price distributions. Experiments on three electricity market datasets demonstrate that RE-Price outperforms state-of-the-art benchmarks, achieving improvements of 15.63%, 17.72%, and 15.70% in MAE, RMSE, and CRPS, respectively, providing accurate forecasts and expressive uncertainty quantification for electricity market decision-making.

The growing renewable energy penetration increases the risks of sub- and super-synchronous oscillations (sub/super SOs) in power systems, making accurate identification of oscillation frequencies critical to grid security. This paper proposes a continuous wavelet transform (CWT)-inspired deep learning (DL) method for dominant frequency identification in sub/super SOs. The designed sub/super-SO-oriented architecture leverages multi-scale representation and scale-selective feature aggregation, adaptively emphasizing informative temporal characteristics under diverse oscillation conditions. The proposed method offers several advantages over existing approaches. First, unlike conventional DL methods that employ generic network architectures, it adopts sub/super-SO-oriented structure inspired by CWT based analysis. Second, in comparison with the CWT and other related spectral analysis techniques, it demonstrates low sensitivity to parameter settings and robust performance even in the presence of data loss. Moreover, the dominant convolutional kernel scales exhibit adaptive switching behavior under complete, isolated-missing, and consecutive-missing data scenarios, thereby enhancing the model's effectiveness and credibility. Experimental validation confirms the method's superiority over existing algorithms. It also exhibits strong generalization to real-world data with grid structures, noise conditions, and oscillation causes that differ from their counterparts in the training set, highlighting its practical applicability.

The planning of fast charging stations (CSs) not only satisfies the escalating charging demand of electric vehicles (EVs), but reshapes its spatio-temporal distribution. However, most existing CS planning studies neglect this endogenous feedback and treat charging demand as an exogenous input, resulting in inefficient infrastructure utilization, heightened grid strain and skewed energy evaluations. To address this gap, this article proposes a two-layer robust renewable CS (RCS) planning and operation model considering spatio-temporal flexibility and continuous decision-dependent uncertainty (DDU) of charging demand. In the first layer, an RCS-capacity-dependent charging demand transfer model is developed based on human mobility empirics, and the corresponding continuous DDU set is constructed. In the second layer, a carbon-aware robust RCS planning and operation model is formulated by integrating charging demand response, capacity-dependent demand uncertainty, and capacity-scaled renewable uncertainty. Since the continuous DDU sets introduce nonlinear and nonconvex couplings into the planning model, an approximation optimization model and a combined alternating direction method and convex-concave procedure algorithm are developed. Numerical studies based on real-world transportation and distribution networks demonstrate that the proposed model improves demand–service matching, reduces transportation-network cost, and achieves a better balance among infrastructure investment, grid operation, and low-carbon charging.

With numerous renewable generators and energy storage systems integrated into the power grids, the security-constrained DC optimal power flow (DCOPF) is essential for power system operation. For large-scale power grids, traditional CPU-based optimization algorithms (such as the simplex and barrier methods) have saturated in computational efficiency and are inherently difficult to parallelize. To tackle these issues, by incorporating the symmetric Gauss-Seidel (sGS) decomposition, this work develops a GPU-based Halpern Peaceman-Rachford algorithm, termed the sGS-HPR, which enjoys an O(1/k) iteration complexity in terms of the KKT residual. Moreover, the closed-form solutions for all subproblems are derived, which only consist of matrix-vector multiplications and vector operations, and thus can be easily parallelized on GPUs. As a consequence, the developed sGS-HPR algorithm enjoys a O(N-L & times; n/is an element of) non-ergodic computational complexity in terms of floating-point operations for obtaining an is an element of-optimal solution measured by the KKT residual for large-scale DCOPF problems, where n represents the variable dimension, and N-L denotes the number of branches in the power grid. Extensive numerical tests on large-scale power grids, reaching up to the 9241-bus PEGASE system, demonstrate the scalability and superior efficiency of the developed GPU-based sGS-HPR algorithm compared to state-of-the-art methods. Notably, the proposed method achieves a 6 & times; speedup compared with Gurobi for largescale instances. Additionally, for ultra-large-scale cases, Gurobi throws an "out-of-memory" error, while the proposed sGS-HPR algorithm maintains its computational scalability and efficiency.

This paper proposes a novel strengthened second-order conic programming (SOCP) relaxation framework to tackle the inexactness of conventional convex relaxations in radial distribution network. We propose compact convex envelope of the omitted concave constraint based on computationally efficient bound tightening method. Numerical tests show that our strengthened formulation achieves less than 5% optimality gap compared to the global optimum and 15~1000 times reduction of relaxation gap than the original conic relaxation.

Fault detection is a critical aspect of system reliability and safety in various fields. Early identification of faults can prevent catastrophic failures, optimize maintenance schedules, and reduce operational costs. Traditional fault detection methods have been fundamental to ensuring system reliability. In recent years, deep learning and machine learning have been employed for fault detection in complex systems. These techniques include hierarchical clustering, the Gaussian mixture model (GMM), and Hotelling’s T-square test. This paper focuses on current state-of-the-art approaches to fault detection. The paper also discusses recent developments in the field of fault detection and its applications.

With the integration of massive distributed energy resources and the widespread participation of novel market entities, the operation of active distribution networks (ADNs) is progressively evolving into a complex multi-scenario, multi-objective problem. Although expert engineers have developed numerous domain specific models (DSMs) to address distinct technical problems, mastering, integrating, and orchestrating these heterogeneous DSMs still entail considerable overhead for ADN operators. Therefore, an intelligent approach is urgently required to unify these DSMs and enable efficient coordination. To address this challenge, this paper proposes the ADN-Agent architecture, which leverages a general large language model (LLM) to coordinate multiple DSMs, enabling adaptive intent recognition, task decomposition, and DSM invocation. Within the ADN-Agent, we design a novel communication mechanism that provides a unified and flexible interface for diverse heterogeneous DSMs. Finally, for some language-intensive subtasks, we propose an automated training pipeline for fine-tuning small language models, thereby effectively enhancing the overall problem-solving capability of the system. Comprehensive comparisons and ablation experiments validate the efficacy of the proposed method and demonstrate that the ADN-Agent architecture outperforms existing LLM application paradigms.

The uncertainties and volatilities introduced by large-scale renewable energy integration present significant challenges to power system operation, making it critical to schedule reasonable reserve capacities across diverse regulation resources. This paper proposes a stochastic generation-reserve co optimization model, where the intra-interval automatic generation control (AGC) process is formulated as a sub-optimization model and embedded in the dispatching stage. To prevent state of charge (SOC) violations of energy storage systems during AGC participation, SOC deviations induced by AGC response are incorporated into the state vector and constrained within a proper range. Moreover, the proposed model accounts for intra-interval load uncertainties and employs a feedback control strategy to address them. The original chance constraints are reformulated into their equivalent deterministic forms, enabling efficient solution of the co-optimization model. Finally, case studies are conducted to verify the effectiveness of the proposed method.

This paper investigates the full-state virtual oscillator control (fVOC) method to realize the flexible operation of 100% PV-powered microgrid. In the grid-connected mode, the enhanced DC voltage stability of fVOC enables the grid-forming PVs to provide the dynamic support while being compatible with the maximum power point tracking function. This dual capability is critical to replace the grid-following operation under the very high penetration of renewable sources. In the islanded situation, the reconnection of the microgrid to the bulk grid leads to large phase jumps, which can trigger the overcurrent saturation and induce the DC voltage instability in grid-forming PVs. To address this issue, a phase adaption current limiting method in fVOC is proposed. The full-order system modeling is performed to guide the parameter design of the proposed method. The Lyapunov function is derived to reveal the enhanced transient stability of the grid-forming PV system during the reconnection of the islanded microgrid. The superiority of the proposed method is verified in both simulations and field tests in a 100% PV powered microgrid. The mode transition between grid-connected and islanded is seamless without islanding detection and pre-synchronization.