The increasing integration of distributed generation (DG) and energy storage systems (ESS) has significantly enhanced the flexibility and efficiency of distribution networks. However, the growing frequency of extreme weather events has exposed the vulnerability of distribution lines, posing serious challenges to the reliability and resilience of such systems. Existing DG and ESS planning models often neglect this vulnerability dimension, leading to suboptimal siting decisions and reduced system robustness. To address this issue, this paper proposes a comprehensive multi-objective optimization framework that coordinates the allocation of DG and ESS and explicitly incorporates line vulnerability under extreme weather conditions. The vulnerability index of each distribution line is first evaluated through Monte Carlo simulations that capture the probabilistic influence of micro-climatic and terrain factors. This assessment serves as a pre-processing stage that screens out high-risk lines and thereby constrains the optimization decision space to more reliable nodes for DG and ESS deployment. Building upon this filtered network, a multi-objective optimization model is established to determine the optimal siting and capacities of DG and ESS. The optimization simultaneously minimizes the total annual cost, which includes investment, operation, and maintenance expenses, as well as network power losses, while improving overall system resilience. A case study on a modified IEEE 33-bus distribution system verifies the effectiveness of the proposed method. The results demonstrate that vulnerability-aware planning achieves a better balance between cost and reliability compared with conventional approaches. Specifically, the proposed strategy reduces annual network losses and outage durations while maintaining voltage stability with respect to climate-adjusted line failure rates. Furthermore, the integration of ESS enables effective peak shaving and valley filling, improving system efficiency and operational flexibility. These findings confirm that incorporating line vulnerability into DG and ESS planning provides a practical and scalable pathway for enhancing the resilience and economy of distribution networks.

The variable and unpredictable output from distributed generation (DG) like wind and solar creates new reliability concerns for distribution networks. Integrating DG on a large scale can unbalance the power supply and compromise quality, making accurate reliability assessment essential. This paper puts forward a new assessment method using a Transformer network. The proposed framework integrates physical modeling with deep learning. First, an improved minimum path algorithm is employed to theoretically evaluate system reliability, specifically modeling the load restoration capability of islanded microgrids. The resulting reliability indices are then discretized into specific intervals to construct a labeled dataset. Subsequently, the Transformer network is innovatively applied to learn the mapping between the stochastic output characteristics of DG and these reliability intervals. By transforming the difficult prediction challenge into a classification task, this method effectively overcomes the problem of non-smoothness in reliability data caused by discrete load restoration. We demonstrate the method's effectiveness on Feeder 4 of the IEEE RBTS 6-node test system. The proposed framework achieves fast online prediction, enabling dynamic monitoring, and proactive warnings against operational risks in the grid.

Fangchenggang City, Guangxi, China (CNFAN), is a typical region for severe thunderstorms, with the annual thunderstorm days exceeding 85. Frequent thunderstorms have led to high rate of lightning strike trip-out on transmission and distribution lines, thereby threatening the operational stability of power grid. A great engineering value has become increasingly apparent, for developing an early warning system of severe thunderstorm potential (12 24 h in advance). This study established a two-stage artificial neural network (ANN) based method for predicting severe thunderstorm potential and intensity, using six-year continuous daily ground flash frequency data and meteorological parameters across 35 geographic grids covering the CNFAN area. In the first stage, a probabilistic classification model was developed to predict the occurrence potential of severe thunderstorms. In the second stage, a regression model was constructed to further predict ground flash frequencies and geometric mean (GM) of lightning current amplitudes for severe thunderstorms. Validation results indicate that this method achieved a probability of detection (POD) of 94.48

Ensuring reliability in distribution networks is essential under increasing operational and economic constraints. Traditional planning models rely on power flow calculations, leading to high computational costs and poor scalability. This study proposes a quantitative decomposition framework that establishes a direct linkage among reliability improvement measures, reliability parameters, and reliability indices, enabling fast and analytical reliability evaluation without power flow analysis. A bi-objective optimization model is developed to minimize both reliability indices (SAIDI) and investment costs, solved using Pareto-based multi-objective PSO combined with the TOPSIS method. Case studies on a 519-node distribution network demonstrate that the proposed approach achieves significant reliability improvement with superior computational efficiency, offering a practical and scalable tool for reliability-oriented distribution planning.

This study addresses the temporal variability and uncertainty of wind and solar outputs that, when integrated at scale into distribution networks, can cause supply–demand imbalances, power-quality deterioration, and reduced reliability. A Transformer-Based reliability assessment method is developed for distribution networks with large-scale distributed energy: first, an improved minimal-path approach is formulated for networks containing distributed resources; when upstream component failures trigger outages in areas with distributed generation, those areas shift to islanded operation, prioritizing the restoration of essential loads within the island. Subsequently, a Transformer-Based procedure is introduced to determine reliability-index intervals by reframing prediction as a classification task, effectively mitigating the non-smooth behavior of the indices. The method’s effectiveness is demonstrated on the IEEE RBTS 6-node system (feeder 4). The proposed approach enables real-time monitoring and early warning of operational risks in distribution networks.

With the growing penetration of county-level photovoltaic (PV) generation, distribution networks are facing increasing challenges in maintaining voltage stability, reducing network losses, and ensuring economic operation. This paper proposes a coordinated optimization framework that integrates energy storage system (ESS) allocation and network reconfiguration for distribution systems with high PV integration. The optimization jointly considers reactive power compensation and storage operation over multiple time periods, aiming to enhance system efficiency and reliability. A two-stage method is developed: the first stage determines the optimal allocation of ESS and initial network topology, while the second stage conducts timeseries operation optimization considering active/reactive power dispatch of ESS and PV. The proposed approach is validated on a modified IEEE 33-bus distribution system with real PV generation and load data. Simulation results demonstrate that the proposed method significantly reduces network losses, improves voltage profiles, and enhances the overall economic performance of the system.

Traditional distribution network reliability planning models often face inefficiencies and challenges due to complex power flow constraints and high-dimensional decision variables. This paper presents an optimization approach for reliability improvement measures based on quantitative decomposition, introducing reliability parameters as intermediary variables to quantify their impact on key reliability indices such as SAIDI. A bi-objective optimization model is developed, aiming to minimize both reliability index values and investment costs. The model is solved using a multi-objective particle swarm optimization (PSO) algorithm, incorporating Pareto dominance principles. The method is validated on a real urban distribution network in China, demonstrating the ability to identify cost-efficient strategies that significantly reduce reliability indices across multiple zones. This study offers practical guidance for utilities to optimize distribution network investments, balancing cost with reliability enhancements in an efficient manner.

Driven by the dual-carbon goals, the increasing penetration of distributed generations (DGs) enhances the flexibility and reliability of distribution networks. However, existing DG configuration studies mainly focus on ensuring the power supply to critical loads and improving system reliability, while often neglecting line vulnerability. To address this gap, this paper proposes a comprehensive line vulnerability assessment model integrating both structural and state perspectives. The line betweenness index and voltage stability index are fused using Dempster–Shafer evidence theory to accurately identify vulnerable lines. On this basis, a DG optimization configuration model considering line vulnerability is developed, which jointly optimizes DG siting and sizing. Simulation results verify that the proposed method effectively enhances system resilience and operational efficiency, achieving secure and optimal DG deployment.

The access of distributed generation will increase the complexity of reliability evaluation, which will affect the accuracy of theoretical reliability evaluation value. Therefore, it is necessary to allocate the deviation between the true value and the theoretical value of the reliability index to the parameters of each component, and to correct the parameters purposefully that are allocated to more deviations. In this paper, firstly, the reliability model of distributed generation is established. Then, the improved proportional allocation method is used to calculate the contribution rate of each parameter to the system index, and the index deviation is allocated based on this. Next, after eliminating the parameters that contribute more to the deviation, the values of these parameters are calculated inversely through establishing the optimization model by using the known system reliability indexes, realizing the correction of the parameters. Finally, a distribution system for RBTS Bus 6 is introduced as standard case to demonstrate the accuracy of the proposed method.

The flood disaster has caused extensive damage and economic losses worldwide, and the Power outage caused by it is particularly common and has an important impact. Therefore, it is of great significance to use effective methods to predict Power outage in advance of flood disaster. A method of using random forest classification to predict the probability of Power outage in flood disaster is proposed. Firstly, supervised learning, decision tree and ensemble learning methods are introduced as the basis of random forest. Secondly, the basic principle of random forest is elaborated in detail. Thirdly, based on the random forest classification method, according to the steps of feature selection, model training and model evaluation, a flood disaster Power outage prediction model is constructed. Finally, a set of data is used to verify and analyze the model, and the performance and accuracy of the model are evaluated. The results show that this method has high accuracy in predicting the probability of Power outage in flood disaster, and provides an effective strategy for the prediction of Power outage in flood disaster in advance, which has important practical application value.

The traditional real-time load warning method for distribution transformers has problems such as low recall rate, low warning accuracy, and long warning time, which may lead to potential equipment failures or overload situations not being detected and dealt with in a timely manner, increasing the safety risk of transformer operation and potentially causing safety issues such as equipment damage, fire, or power outage. Therefore, a real-time early warning method of distribution transformer load considering meteorological factor data is designed. The meteorological factor data are collected by the light sensor, humidity sensor, temperature sensor and rainfall sensor, and the load data collection architecture is built by the load monitor, central master station and maintenance station to realize the load data collection of the distribution transformer. The K-nearest neighbor (KNN) method is used to process the missing values of the data, and the LOF algorithm is used to determine the local outliers and eliminate the outliers in the data set to achieve data cleaning. Considering the load loss, hot spot temperature and meteorological factors of the distribution transformer, an early warning model is built, and the cleaned data are input into the model to realize Real-time early warning of the distribution transformer load. The experimental results show that the recall rate of this method varies from 95% to 97%, the accuracy rate of early warning is always above 94%, and the maximum value of early warning time is 0.63s. Having good early warning ability.

Abstract The global warming and El Nino effects are becoming more and more obvious now. These result in frequent rainstorms, especially in coastal areas. It is necessary to deeply study evolution mechanism of rainstorm-flood disasters, assess disaster risk of power grid, and formulate efficient power grid flood control strategies. Firstly, the research progress of disaster modeling is summarized according to the disaster evolution mechanism. The differences between hydrological models and machine learning models are compared, and the possibility of combining them is discussed. Secondly, current disaster risk assessment methods of power grid and their application in rainstorm-flood disasters to quantify flood prevention capacity are reviewed. Thirdly, the power grid flood control measures are studied from the perspective of pre-disaster grid reinforcement and post-disaster repair to improve resilience. Finally, in view of the above content, the shortcomings of current researches and feasible research directions in the future are proposed.

In the power system, power outages are usually caused by the distribution system. Therefore, it is necessary to allocate the system reliability indexes to each component, so as to identify the components that have great influence on the system reliability, and to optimize them in order to improve the system reliability. In this paper, firstly, multi-state output power model of distributed generation is established, using historical distributed generation data. Then, based on Shapley value, the system reliability indexes are allocated to each component, and by standardizing them, the contribution rate of each component to the system index can be obtained. Finally, a distribution system for RBTS Bus 6 is introduced as standard case to demonstrate the validity of the proposed method. By analyzing the calculation results, it can be obtained that the component contribution rate is related to the size of the component reliability parameters and the location of the component.

In order to address the fossil fuel crisis and climate change issues, vigorously developing renewable energy has become the dominant direction for the global energy low-carbon transition. However, with the large-scale access of distributed generations such as photovoltaic panels and wind turbines, the safe and reliable operation of the distribution system faces severe challenges. Therefore, it is necessary to propose a rapid reliability evaluation method to provide a decision basis for operation dispatchers. This paper proposes a dynamic reliability evaluation method for distribution systems based on sparse polynomial chaos (PC). Firstly, a PC model is established using historical power output of each distributed generation and its corresponding system reliability index. Then, the least angle regression (LAR) method is used to screen out the items in the polynomial that have the greatest impact on the output response, achieving the sparseness of the polynomial and efficient reliability dynamic evaluation. A distribution system for RBTS Bus 6 is introduced as standard case to demonstrate the accuracy of calculating the system reliability index and the effectiveness of reducing the terms of polynomials and alleviating the dimension disaster problem of PC method.

Because the reliability data statistics work is complex, dynamic, and vulnerable to human factors. There may be some errors or missing in the reliability parameters of each component in the distribution system with distributed generations. It will lead to wrong evaluation results, which may mislead the planning, transformation and other decisions of the distribution system, and even affect the safe and reliable operation of the future planning system. Therefore, this paper studies the anomaly detection and correction method of component reliability parameters. Firstly, the models of component failure rate and repair rate are established respectively. Then, FCM (Fuzzy C-means) method is used to detect outliers in the failure rate or repair rate collected by a component at each moment, based on its model. Next, after eliminating the abnormal parameters detected above, the values of these parameters are calculated inversely through establishing the optimization model by using the known system reliability indexes, realizing the correction of abnormal parameters. Finally, a distribution system for RBTS Bus 6 is introduced as standard case to demonstrate the accuracy of the proposed method.

The fault outage time caused by the distribution network accounts for 80% of the total outage time. Due to the access of a large number of distributed power sources, it is easy to cause cascading failures and large-scale power outages after the failure occurs. This paper proposes a fault blocking method for distribution network based on offline reinforcement learning. The proposed method does not need to interact with the real distribution network nor does it need to accurately model the distribution network. Only historical data is needed to obtain a feasible distribution network fault blocking strategy through offline reinforcement learning methods. The IEEE 16 node distribution network system is used to verify the proposed method. The results show that the proposed method reduces the average fault recovery time by 8.5% and the network loss by 6.7%, outperforming previous reinforcement learning methods.

The disaster chain triggered by typhoon has long duration, wide impact range, and complex disaster mechanism. These disaster chain has caused significant losses to distribution network. This paper reviews current research progress and issues that need to be addressed from three perspectives: outage risk analysis, outage prediction, and repair path optimization for distribution network. Firstly, the complex disaster mechanism faced by distribution network under typhoon disaster chain is analyzed. It emphasizes the necessity of studying the outage risk analysis by taking disaster chains as a whole. Secondly, it elaborates on the problems of data imbalance and poor interpret ability in outage forecast research. The physical-data fusion modeling's feasibility will be explored to enhance distribution network's early warning capability. Finally, the interaction mechanism among typhoon disaster chain, transportation and distribution network are discussed. By discussing emergency repair path optimization method, the distribution network's disaster response ability is enhanced.

With the large-scale integration of distributed generation, the characteristics of the distribution network have changed significantly, resulting in a notable decline in the accuracy of traditional fault localization methods. To address this issue, this paper proposes a fault location method based on a conservative Q-learning (CQL) algorithm. Specifically, to tackle the problem of potential misjudgment, an objective function/fitness value correction method for bidirectional power flow induced by distributed generation is introduced. To overcome the challenges of slow convergence speed and poor robustness in existing models, a CQL algorithm is employed to enhance the efficiency and stability of the algorithm. The IEEE-33 distribution network is used as a case study to demonstrate the accuracy and efficiency of the proposed fault localization method.