Typhoon-induced extreme weather poses a severe threat to power systems with high offshore wind penetration. Source-side wind turbine tripping and grid-side transmission line failures are likely to occur simultaneously, which may trigger cascading outages and large-scale load shedding. A multi-level source-grid-load-storage preventive resilience dispatch strategy is proposed. A typhoon spatiotemporal evolution model is first established based on the Batts gradient wind model. Failure probability models for offshore wind turbines and overhead transmission lines are developed while considering strong wind and lightning strike effects. The most probable and severe fault scenario is identified using an entropy-based quantification method. A two-stage robust preventive dispatch model is subsequently formulated. In the day-ahead stage, unit commitment, multi-type reserve allocation, and pumped storage scheduling are optimized at a 1 h resolution. In the real-time stage, combined wind-storage systems are coordinated at a 10 min resolution to accommodate rapid wind power ramps caused by high-wind shutdown events. The model is reformulated through Lagrangian duality and solved by the Benders decomposition algorithm. Case studies on a modified IEEE-RTS 24-bus system with three offshore wind farms demonstrate that the proposed strategy reduces wind curtailment by 66.3%, load shedding by 74.6%, and total cost by 14.8% compared with the case without energy storage. The combined operation cost of storage resources accounts for only 3.1% of the total cost, confirming its favorable cost-effectiveness for resilience enhancement. The proposed strategy contributes to the sustainable integration of offshore wind energy by ensuring a reliable power supply during extreme weather events.

False data injection attacks (FDIA) often exploit vulnerabilities in malicious data detection to manipulate state estimation results in power cyber-physical systems, leading to unstable power system operation. Traditional deep learning-based methods often fail to fully consider the spatial correlation between measurement data and power grid topology, as well as the temporal characteristics of power measurement data, and thus the detection performance always is significantly weakened. To address the aforementioned challenges, an efficient FDIA detection model is proposed, which effectively captures both the temporal characteristics of measurement data and the spatial relationships between power grid topology and measurement data. A key innovation is the feature extraction module based on the stacked-LSTM network, which can efficiently and fully extract the multi-scale temporal features of measurement data. In addition, we design a convolutional attention-based residual auto-encoder to efficiently extract the complex spatial features, which can accurately capture the potential correlation between data, and dynamically aggregate key features and spatial topology information. A series of extensive experiments are performed over the IEEE14, IEEE39 and IEEE118 bus systems, and the experimental results demonstrate that the proposed detection model significantly outperforms the state-of-the-art schemes in terms of accuracy, robustness and complexity, and provides reasonable interpretability from the spatial-temporal dimensions.

The increasingly frequent and severe natural disasters have posed significant challenges to the resilience of power systems worldwide, creating an urgent need to investigate the security issues associated with these extreme events and to develop effective risk mitigation strategies. Meanwhile, as one of the leading topics in current research, artificial intelligence (AI) has demonstrated outstanding performance across various domains, such as AI-driven smart grids and smart cities. In particular, its efficiency in processing big data and solving complex computational problems has made AI a powerful tool for supporting decision-making in complex scenarios. This article presents a focused overview of power system resilience against natural disasters, highlighting recent advancements in AI-based approaches aimed at enhancing system security and response capabilities. It begins by introducing various types of natural disasters and their corresponding impacts on power systems. Then, a systematic overview of AI applications in power systems under disaster scenarios is provided, with a classification based on the task categories, i.e., predictive, descriptive and prescriptive tasks. Following this, this article analyzes current research trends and finds a growing shift from knowledge-based models towards data-driven models. Furthermore, this paper discusses the major challenges in this research field, including data processing, data management, and data analytics; the challenges introduced by large language models in power systems; and the limitations related to AI model interpretability and generalization capability. Finally, this article outlines several potential future research directions.

Typhoon weather conditions induce severe fluctuations in offshore wind power generation， posing a significant threat to the secure and stable operation of power systems. Accurate prediction of wind power ramp events is therefore crucial for ensuring grid security. To address the limitations of existing methods in capturing the complex and dynamic spatiotemporal correlations of offshore wind farms and their limited ramp prediction accuracy， this paper proposes a novel prediction method for offshore wind power ramp events in typhoon weather conditions. Firstly， an adaptive graph convolutional network （AGCN） is employed to capture the complex and varying spatial features of the offshore wind farm. Secondly， a bidirectional long short-term memory （BiLSTM） network is adopted to extract bidirectional dependencies within the time series， while an attention mechanism （AM） is introduced to enhance the learning of critical features and temporal information. Building on this， the traditional swing door algorithm （SDA） is improved by incorporating a Bump event merging strategy and an extreme point correction strategy to enhance the detection accuracy of ramp events. Finally， experimental results based on SCADA data from a wind farm in Shanghai demonstrate that the proposed combined AGCN-BiLSTM-AM model can effectively capture the complex dynamic spatiotemporal relationships of the offshore wind farm， exhibiting high accuracy and strong applicability for predicting wind power ramp events under typhoon conditions.

The emergence of high-voltage direct current (HVDC) integrated power systems introduces significant out-of-distribution challenges to machine learning-based analysis of cascading failures (CFs). Existing algorithms often suffer from poor interpretability and generalisation in handling diverse CF propagation modes. To address these issues, this paper proposes a novel causal graph attention (CGAT) model that captures critical information from CF events to accurately predict propagation paths and impacts. Specifically, by employing a graph attention mechanism, the model extracts both causal and trivial features as distinctive attention dual subgraphs, which aids in differentiating shortcuts in causal variables and improving prediction accuracy. These subgraphs also provide visual representations that highlight the key elements influencing CF propagation. By incorporating interventions, the CGAT model enhances prediction capabilities across a variety of scenarios that include HVDC control strategies and renewable energy sources. Extensive experiments demonstrate the effectiveness and superiority of the proposed CGAT model, which achieves prediction accuracies of 87.57% for CF chains and 92.89% for power losses on the 30-bus system. Even on the 2737-sop system, the accuracies remain at 84.69% and 88.14%, respectively. This model provides a highly efficient online early-warning tool for preventing large-scale blackouts in modern power systems.

With the increasing scale of offshore wind farms, the spatial-temporal correlation of wind turbines is commonly considered in predicting wind power generation. Meanwhile, the seasonal variation of offshore wind conditions necessitates the consideration of the spatial relationship of wind farms with dynamic changes. This paper proposes a new power prediction model for offshore wind farms, namely the feature attention graph convolutional neural network with temporal transformers (FAGTTN). Specifically, the feature attention module is utilised to extract important features from the offshore wind power supervisory control and data acquisition (SCADA) system data. Then, the adaptive graph convolutional neural network (AGCN) is employed to learn the embedding of multiple wind turbine nodes, uncovering the hidden spatial dependence in the data to express the dynamic spatial relationship of offshore wind farms. Besides, the temporal transformer is used to capture time dependence and temporal patterns in the time series. The proposed method is validated using the real-world data from the offshore wind farm at Donghai Bridge, demonstrating its validity and superiority. The results show that the proposed offshore wind turbine graph topology network can effectively utilise the geographic location information of wind turbines and outperform existing methods in terms of accuracy and interpretability for offshore wind turbine output prediction.

Over the past decade, typhoons have emerged as a primary natural hazard threatening the secure operation of power grids in coastal urban areas. Addressing the day-ahead dispatch challenge for power systems that integrate offshore wind generation, this paper presents a multi-scenario stochastic programming framework that explicitly incorporates typhoon-related uncertainties. Firstly, considering typhoon uncertainty, a typhoon scene set is constructed using the typhoon path, maximum wind speed error model, and probability circle. On this basis, a system loss risk cost model and an offshore wind power output uncertainty set under typhoon uncertainty are proposed. Then, the optimal scheduling model of power system with offshore wind power based on multi-scenario stochastic programming is constructed. With the goal of minimizing the total cost of load loss risk cost, scheduling cost and wind power operation cost under the influence of typhoon, solve the model to get day-ahead scheduling case. Finally, simulations are performed on a modified IEEE 39-bus example. The outcomes demonstrate that the presented method can more accurately reflect the actual uncertainty of typhoons, lessen the uncertainty and conservatism of day - ahead scheduling based on a single typhoon forecast situation, and demonstrate the validity of the proposed method and model.

As the scale of offshore wind power grid-connection continues to expand, the impacts on the adequacy of power system are also increasing. This study proposes an adequacy evaluation method for power systems of high offshore wind power penetrations, which considers the capacity constraints of grid-connected systems and also the uncertainty of offshore wind power. Specifically, to characterize the uncertainty of offshore wind power, the model of prediction plus error is firstly utilized to simulate offshore wind power output. Then, Monte Carlo probability estimation based on cluster analysis is proposed to obtain the capacity probability samples of the grid-connected system. On this basis, an improved probabilistic power flow for AC/DC power system is presented to analyze the impacts of the grid-connected system and the offshore wind power uncertainty under each scenario, with several typical indices on grid adequacy calculated accordingly. Finally, the proposed model and method are verified on the modified IEEE 30 bus test system.

The increasing penetration of distributed generation (DG) brings about great economic and environmental benefits, while also negatively affecting the operation of distribution networks due to its high intermittency. Although distributed energy storage (DES) can effectively deal with the problems caused by massive DG penetrations by decoupling the generation and consumption of electricity, the placement of DES significantly determines the effectiveness of its capabilities. Unfortunately, existing DES placement studies are commonly based on a balanced network model, whereas practical distribution networks are unbalanced. In addition, existing DES placement studies are mostly based on an extreme scenario and rarely consider the operational complexity resulting from the uncertainties of DGs and loads. To address the aforementioned challenges, this paper proposes a hierarchical and sequential DES placement strategy in distribution networks by considering multi-scenario operations. Specifically, the proposed hierarchical framework for DES placement includes three sequential layers: outer, inter, and inner. In the outer layer, a multi-scenario comprehensive loss sensitivity index (MSCLSI) is first introduced to search for the most effective DES placement location. Subsequently, the sizing and scheduling of DES for the selected location are conducted through coordinated optimization across the inter and inner layers, which can be solved using a hybrid method combining particle swarm optimization and second-order cone programming (PSO-SOCP). Finally, a series of detailed simulations are carried out over the IEEE-33 test system and the experimental results demonstrate that the proposed scheme can provide significant effectiveness and superiority compared to the state-of-the-art schemes.

Massive integration of distributed generators with inherently high intermittency and volatility leads to frequent voltage violations. Thus, an appropriate voltage/Var control (VVC) is essential for the secure and economic operation of distribution systems. The increasing penetrations of rooftop photovoltaic units strengthens the coupling between the medium and low voltage (MV-LV) distribution networks spatially, causes temporal interference of VVCs with each other on different timescales, and worsens the network unbalance profile especially on the LV sides. To address this challenge, this study proposes a spatial-temporal coordinated VVC for MV-LV unbalanced distribution networks. Based on 'decomposition-coordination' principle, the developed strategy consists of three modules of real-time evaluation of reactive capability of LV feeders, long-short-time coordinated VVC of MV network, and parallel short-time VVC of LV feeders. Based on the proposed strategy, the VVC optimization problems for these modules are formulated and jointly solved by mixed-integer second-order cone programming and linear programming methods, to minimize the voltage deviations within the MV-LV unbalanced distribution systems in real time. Finally, based on the joint simulation platform of MATLAB and Python, the effectiveness and superiority of the proposal are numerically verified on a real Australian distribution system.

False data injection attacks (FDIAs) refer to attackers exploiting vulnerabilities in the detection of bad data in smart grid energy management systems to maliciously manipulate state estimation results in the cyber-physical system, resulting in unstable operation of the power system. Existing deep learning-based detection schemes often fail to capture the spatial topology features and long-term dependencies in power grid data well. Meanwhile, the complexity of deep detection models makes them a "difficult-to-interpret system," reducing the credibility of detection results. To address the aforementioned challenges, this article presents a dependency-aware deep interpretable FDIA detection model. The proposed model first introduces graph sample and aggregate (GraphSAGE) network to extract spatial topological features, which are used to represent the deep spatial topological dependencies of adjacent data nodes. Subsequently, we build a Bidirectional Long Short-Term Memory (BiLSTM) network with a Squeeze-and-Excitation (SE) attention module, which can efficiently aggregate attack characteristics and long-term dependency information by dynamically capturing the potential correlations between FDIAs detection and measurement data. Furthermore, the SHapley Additive exPlanations (SHAP) method is used to demonstrate the interpretability of the model in the spatial-temporal dimensions and then provide the basis for high-precision detection results. A series of extensive experiments are carried out over the IEEE 14-bus and 118-bus test systems. The experimental results demonstrate that the proposed model presents a superior overall performance comparing with several state-of-the-art FDIA detection models, and provides reasonable interpretability from the spatial-temporal dimensions.

The dynamic changes of wind speed and direction have a significant impact on the power generation efficiency of offshore wind turbines (WTs). This paper proposes a coordinated yaw control of multiple offshore wind turbines especially considering the wind direction. First, on the foundation of Gaussian-Curl Hybrid model, we establish a wake model considering real-time dynamic wind direction. Then we propose Model Predictive Control based on this wake model to adjust the yaw angle of single WT, and establish a coordinated yaw control across multiple WTs on the basis of Serial Refinement which obtains WTs’ optimal yaw angle by ergodic search. Simulation results demonstrate that this approach effectively improves power generation while reducing fatigue loads.

Improving the accuracy of offshore wind power forecasting is one of the most effective means to enhance the security and stability of offshore wind power integration. To this end， this paper proposes a review of offshore wind power forecasting studies. First， based on recent studies in offshore wind power forecasting， the review investigates three key aspects， such as data preprocessing techniques for offshore wind power， offshore wind power forecasting models， and forecasting for large-scale offshore wind farm clusters. Subsequently， the challenges currently faced in offshore wind power forecasting are analyzed. Finally， potential future research directions are explored by incorporating emerging technologies. The findings of this study can serve as a reference for the operation and maintenance of offshore wind farms as well as system dispatch after grid integration.

The emerging of modern power system poses significant out-of-distribution (OOD) challenges to machine learning (ML)-based cascading failure (CF) analysis. This paper pioneering proposes a Causal Graph Attention Network (CGAT) model, integrating structural causal model (SCM) identify pivotal information within CF events, thereby guiding the predictive accuracy of CF propagation path. Specifically, graph attention mechanism is employed to extract both causal and environmental features as distinctive subgraphs from the topology of power system. The subgraphs help distinguish confounding factors between causal variables and predictions, and their visual representation enables to clarify critical components driving CF propagation. Furthermore, by distinguishing between causal and environmental features, CGAT model predicts CF sequences across diverse topology configurations and varying scenarios of renewable energy sources (RES). Extensive experiments on an augmented IEEE 30-bus system illustrate the effectiveness and superiority of CGAT model.

An adaptive droop control strategy is proposed for a parallel distributed multi-energy storage system of an isolated DC microgrid with unmatched line impedance and abnormal communication. System line impedance mismatch can cause unbalanced load power distribution and reduce service life of distributed energy storage unit (DESU). Therefore, an improved droop control based on mixed coefficient compensation of state of charge (SOC) and voltage is designed, which can adaptively adjust a power characteristic curve according to the sampling period, so as to ensure precise distribution of load power while minimizing voltage deviation. Considering the stability of the system under communication anomaly, a non-communication backup control based on Metropolis acceptance criterion is proposed, which only uses internal data to adaptively adjust the droop coefficient. In addition, a gradual smooth handover strategy is designed through gradient coefficient to optimize control stability under communication anomaly. Finally, effectiveness and correctness of the proposed control strategy are verified by mathematical analysis and RTDS/DSP hardware-in-the-loop experiments.

Extreme weather such as typhoon leads to dramatic fluctuations in offshore wind power, which brings challenges to the safe and stable operation of power systems. Thus, it is particularly important to improve the accuracy of offshore wind power prediction under typhoon weather. To address the challenge of insufficient mining of complicated and dynamic spatio-temporal correlations of offshore wind farms under extreme weather in existing studies, this study proposes an AGCN-BiLSTM-AM based offshore wind power prediction model. Firstly, the spatial correlation of offshore wind farms topology is constructed as graphs, and an adaptive graph convolution network (AGCN) is then utilized to quantify and mine the topological changes due to wind fluctuation and wake effect. Secondly, the BiLSTM module is used to extract bidirectional temporal features to capture the drastic fluctuations of wind speed under typhoon weather. Additionally, the feature attention mechanism (AM) is introduced to enhance the learning of key features. Finally, the proposed model is verified on real wind farm data, with a higher prediction accuracy obtained under typhoon weather.

In power systems, topology attacks can trigger cascading reactions and hidden failures and seriously affect the stable operation of the system, causing huge losses. Aiming at the dependence of spatial modeling of current detection algorithms on predefined graphs and their poor accuracy, this paper proposes a model based on adaptive graph learning and multi-head temporal self-attention (AGL-MTSA). Specifically, the adaptive graph learning module (AGL) automatically learns the topological correlations among nodes, and the multi-head temporal self-attention module (MTSA) accurately captures the importance of different time series on topological attack events. The effectiveness and superiority of the proposed AGL-MTSA model are verified by detailed simulations on a typical ${160}$-node cyber-physical power system.

Situation awareness is a critical foundation for the safe operation of offshore wind power. To address the uncertainty risk caused by offshore wind power penetration in the extremely complex marine environment, a novel security situation awareness method for the offshore wind power networking system is proposed on the basis of the vague-CNN-LSTM model. Firstly, a partition model of offshore wind system topology is built according to the location of offshore wind farm access nodes, which can quickly capture the elements of the system situation. Secondly, a vague set is introduced to propose an interlaced offshore wind power interval division method integrating the truth-membership degree and pseudo-membership degree functions. Thirdly, the vague set interval prediction model based on vague-CNN-LSTM for offshore wind power is established to forecast the future fluctuation range of offshore wind power from different criteria, including support, opposition and hesitation uncertainty. Fourthly, early warning indexes for the security situation of the offshore wind power networking system are proposed so that the real-time and future security risks of the entire system can be perceived from the multiple layers of node-branch-area-network. Finally, the effectiveness of the proposed method is validated using a case study of an actual offshore wind farm in China.

Virtual power plant (VPP) is a critical platform for modern distribution systems with distributed generators (DGs). However, its cybersecurity is susceptible to cyber-attacks such as false data injection attacks (FDIAs). The impacts of FDIAs on VPP-distribution cyber–physical power systems have not been thoroughly investigated in the literature. This study concentrates on the distribution–VPP joint system and designs a new FDIA framework, topology-distributed-generator attack (TDA), that manipulates power network topology and DG outputs. An attack vector is designed carrying incorrect topology, falsified DG outputs, and tampered power flow information that can bypass the existing bad data detection and topology error identification, misleading the decision-making in the control center. Additionally, TDA models are formulated to optimize attack vectors based on objectives of attack investment, VPP economic loss, and operational security. A hybrid solution framework is then proposed for the optimization problem above, where the corresponding submodules realize the bad data detection, topology error identification, and optimal dispatching in the optimal attack vector. The effectiveness and superiority of the proposal are numerically verified on a 62-node cyber–physical system. Key findings highlight that VPP-integrated distribution systems are more vulnerable under low-level renewable energy penetration and the urgent need for enhancing backup power supplies to mitigate such threats.