The occurrence of wildfires and line tripping caused by tree-line faults has emerged as a critical issue in power grids. To investigate the mechanism of streamer discharge under wildfires and provide theoretical support for realistic simulation of tree-line faults, a two-dimensional axisymmetric streamer discharge fluid model has been established. The microscopic quantities that cannot be measured experimentally such as densities of electrons and positive/negative ions, electric field spatial distribution during discharge process are calculated. The reliability of the model is verified by comparison with the existing research results. The calculation results show that with the increase of temperature, the electric field required for the streamer discharge is reduced, and the streamer radius and velocity are increased. The polarization electric field generated around the solid particles is larger than the maximum electric field of the original streamer. The combined effects of high temperature and solid particles greatly promote the development of streamer discharge, further increasing the streamer velocity.

Isolated gate drivers are widely adopted for half-bridge, full-bridge and dual boost AC-DC converters for power conversion in low power electromagnetic energy harvesting, in which efficiency is of great importance. However, commercial isolated gate drivers were all designed for higher voltage and higher power applications, which brought severe power consumptions at low power scenario. Aim at low power loss and simple driver architecture, this paper proposes two low power isolated driver circuits based on the transformer and the bootstrap capacitor respectively. The detailed design and loss analysis are carried out. The paper elaborates the design method and scope of application. The prototypes of the two isolated gate drivers are also built to verify the effectiveness on the experimental platform.

The growing demand for large-scale new energy access highlights the need for a domestically produced and independently operated distribution network simulation platform. However, traditional simulation software has high entry thresholds and operational difficulties, requiring users to write new algorithms for new equipment. Moreover, primary simulation tools are foreign software with poor openness, emphasizing the need for a domestically produced and independently operated platform. This paper proposes a hybrid modeling architecture for a dynamic simulation platform of distribution power systems. The proposed architecture utilizes mathematical equation strings modeling methods and transfer functions modeling for modeling to minimize programming while maintaining scalability. This technology can contribute to the development of a domestically produced and independently operated distribution network simulation platform, particularly in the context of new energy access.

This study proposes a method for constructing a multi-scale target detection model specifically for power public safety signage. The article designs a multi-scale target detection network based on the YOLOv5 object detection model to meet the needs of multi-scale transformation and small target detection for safety signage. In addition, the article presents a quantitative perception training method of neural network graph optimization and network quantization, which is used to deploy the model at the edge and improve the inference speed. Experimental results show that this model exhibits high recognition accuracy and fast inference speed on edge devices, making it suitable for real-time safety signage recognition.

In response to the constraints of edge computing environments, this paper conducts an in-depth optimization of the YOLOv5-based object detection model to achieve real-time high-precision object detection on edge devices. Firstly, the paper analyzes the principles of two model compression techniques: graph optimization and network quantization, to achieve structural-level optimization of the model. Then, it optimizes memory access for the model using the unified memory access mechanism of the GPU. Experimental results show that compared to the original model, the optimized YOLOv5 achieves over 5.5 times the inference acceleration on the Nvidia Jetson Xavier NX edge device, with only a slight 1.5% decrease in average precision (MAP), achieving real-time high-precision object detection. This paper provides a reference for optimizing neural network model deployment on edge devices.

In order to truly evaluate the reliability of low-voltage distribution network, considering the influence of distributed energy uncertainty on the reliability evaluation of low-voltage distribution network, a reliability evaluation method of low-voltage distribution network considering distributed energy is proposed. Firstly, considering the load recovery capability of distributed energy after failure, and then combining with distributed energy, a collaborative assessment method based on sequential Monte Carlo for medium-low voltage distribution network reliability is proposed. Finally, taking IEEE RBTS BUS-2 system as an example, the system reliability levels in different scenarios are compared and analyzed, and the effectiveness of the proposed method is verified.

This paper presents the design and development of an onboard intelligent device based on Jetson Xavier NX. The device utilizes the computational and recognition capabilities of AI chips to analyze safety signs and provide feedback, assisting in inspections. The technological components and design of the onboard intelligent device are discussed, including the structural framework of embedded systems built with AI chips, as well as the components and applications of embedded systems. The DJI Maverick series models, M30 and M300, are selected and their applications and performance in the power grid industry are introduced. The unmanned aircraft and onboard intelligent device are connected through the DJI Maverick series interface for expansion and control. Edge computing is implemented using a training and computing separation model, and a development roadmap is provided. The designed onboard intelligent device is installed on DJI drones to real-time recognize video streams captured by the drones, which includes modules for data collection, computational analysis, wireless communication, storage, input/output communication, and power interfaces. Through this design, intelligent identification and real-time monitoring of public safety in the power sector can be achieved.

The purpose of this article is to propose a medium to long-term load forecasting method based on BP neural network and S-curve fitting. This article first uses a knowledge graph to determine the factors that affect power load forecasting, and selects the key factors that affect load forecasting. Secondly, in order to compensate for the shortcomings of the BP neural network method for forecasting, the paper proposes a load forecasting method based on BP neural network and S-curve fitting. This method compensates for the uncertainty of the neural network prediction results and the low accuracy of S-curve fitting in predicting load development trends. Finally, the accuracy of this method for load forecasting was verified through simulation examples.

In order to better count the reliability indexes of low-voltage distribution network, the application advantages of smart meters are used to construct a scientific and reasonable intelligent analysis mode to improve the operation quality of low-voltage power grid. This paper briefly introduces the main functions of smart meters, and analyzes the regional division of smart meters and the selection of iconic users in low-voltage distribution network faults. It focuses on the specific application of smart meter user outage information statistics in the statistics of power grid reliability indices.

The topology of the distribution network is complex, with many branches.Users have increasingly higher requirements for power supply reliability. Once a fault occurs, how to quickly and accurately locate the fault point and reduce the power outage time is an important aspect of ensuring the safe operation of the power system and providing power supply reliability. With the continuous improvement of the monitoring level of the distribution network, the fault location based on the traveling wave characteristics of the distribution network has gradually become a current research hotspot. Based on the analysis of the propagation characteristics of the traveling wave in the distribution network and the attenuation characteristics of the traveling wave in the distribution line, this paper uses the Stacked Autoencoder (SAE) model to fit the difference between the attenuation characteristics of the traveling wave and the location of the fault point. The simulation results show that the scheme is less affected by the transition resistance, the initial phase angle of the fault, and the fault type, and the positioning accuracy is high. It can accurately locate the fault points on the line with high accuracy.

With the increasing proportion of renewable energy in the distribution network, the uncertainty of renewable energy poses a challenge to the operation analysis and decision-making of distribution network. Scenario is an important tool to control and optimization of the power system under uncertainty. Therefore, how to generate scenarios efficiently and accurately is the focus of power system uncertainty analysis. In order to solve the above problems, this paper uses Generative Adversarial Nets to generate distribution network operation scenarios. By analyzing the operation scenario generation of distribution network and the principle of Generative Adversarial Nets, the structure and training method of Generative Adversarial Nets for time-series power flow data are proposed and verified in an example based on IEEE33 bus system. The results show that the designed network can learn the characteristics and distribution of operation scenarios of distribution network. This paper also proposes a general distribution network operation scenario generation platform, and gives a general method from scenario set construction to network training, testing and application.

With the development of smart grids and emerging measurement technologies, the massive growth of distribution grid data may impact the reliable, economic, and safe operation of the distribution network. For a large-scale distribution network state estimation, the strategy of measuring data for a distribution grid is critical to its economy and reliability. This paper proposed a distribution network state estimation model based on a graph convolutional neural network. The proposed algorithm uses a genetic algorithm to optimize the sensor locations, frequency of sensors in the distribution network and configures of devices to guarantee the proposed state estimation data accuracy. With minimizing costs of investment and operation, the proposed graph convolutional neural network provides super-resolution state estimation data of the distribution network by using low-resolution measurement data. The proposed method is tested and verified by the IEEE33 distribution network system and the testing result demonstrates the feasibility and effectiveness of the proposed model and algorithm.

Using historical data to predict future energy demand of power system plays a key role in solving the challenge of supply and demand balance of power system brought by renewable energy. In this paper, a multi-time scale power load forecasting method based on wavelet threshold denoising and Prophet is proposed. Firstly, the wavelet threshold denoising algorithm is used to de-noise the historical load data to reduce the influence of inherent noise caused by acquisition equipment and transmission equipment on the results. Then the Prophet algorithm is used to build a time series model of historical data, so as to predict the future power load. This method can predict the power load at different time scales according to different demands. Simulation results show that the proposed method has high prediction accuracy and stable prediction results for different time scales.

Because the fault location of active distribution network is a complex optimization model, the traditional genetic algorithm has the problems of premature and slow convergence when solving complex optimization problems. The switch function that can adapt to the switching of multiple distributed power sources in real time is proposed. Here, we proposed a novel method to locate the fault of distributed power distribution network based on a multi-group genetic algorithm. When locating the fault zone, The algorithm stipulates that the direction of the generator flowing to the electrical equipment is the forward direction of the transmission line. In this paper, multiple populations are used to search the solution range at the same time to ensure that the final solution is the global optimal solution, and the optimal individual retention algebra is used as the convergence condition to fully improve the solution speed, which is suitable for complex distribution networks with distributed power generation. The fault location model proposed in this paper is simulated by a numerical example, and the results show that the model can accurately find the fault point, and has validity and accuracy.

With the increasing awareness of environmental protection, integrated energy systems have gained an increasingly attention. However, the significance of resilience of integrated energy systems has been increasingly emphasized at a time when extreme events occur frequently. Therefore, this paper focuses on the concept of resilience. Firstly, the resilience evaluation index of integrated energy system is established. The index considers the damage frequency of each device in the system, the scale of load loss caused by failure and the time of the disaster. Subsequently, this paper gives the resilience evaluation process based on the simulation of the load loss of the system in a disaster. By improving the enumeration method, the accuracy of the overall resilience assessment when enumerating low-order cases is improved. The effectiveness of this method is tested in an actual case. Finally, this paper proposes a specific process for resilience strengthening of the system using this resilience assessment method and verifies that these methods can effectively improve the system resilience.

在详细分析"双碳"目标背景下配电网关键特征演化的基础上,对配电网智能感知与故障诊断存在的挑战进行了梳理.随后对配电网智能感知的技术与需求进行了分门别类,对存在的问题以及借助人工智能技术实现智能感知的方法进行了归纳.接着对配电网故障诊断与故障定位的难点进行了分析,介绍了人工智能在配电网故障诊断与故障定位中的应用研究进展.为进一步分析智能感知与故障诊断的相辅相成以及人工智能在其中的作用,以10 kV架空线路局部放电故障检测为例,详细分析了人工智能在配电网智能感知中的作用以及不同人工智能算法进行局部放电故障检测时的性能对比.最后,对人工智能在配电网智能感知与故障诊断中的应用进行了评述.

It is very important for power grid development research and related technical improvement to obtain the disaster situation of fine-scale distribution network, such as the transportation condition evaluation of distribution network and the wind waterlogging disaster prediction of distribution network. Among them, the wind waterlogging disaster prediction of distribution network is the main one, and the prediction of its disaster degree often determines whether the distribution network can be prevented before and rescued after the disaster. Therefore, in view of the above problems, combined with the actual transmission situation of the distribution network, after collecting the measured disaster data of the distribution network in relevant areas, combined with the multi-source data fusion technology and neural network modeling technology, this paper analyzes the disaster degree indicators of different distribution networks and constructs the relevant fuzzy matrix through the fuzzy theory to evaluate the disaster degree, which is verified by the measured data. This distribution network disaster loss prediction model can effectively implement the disaster loss prediction of distribution network and compare its prediction results with the other two different common models. The comparison results show that the prediction accuracy of the multi-source data fusion prediction model constructed in this paper is more than 0.95 compared with the other two models, while the prediction accuracy of the other two models is not more than 0.9, which proves that the model constructed in this paper has smaller errors. It has the advantages of higher accuracy and faster convergence speed.

为寻求配网改造方案可靠性与经济性之间的平衡,提出了一种基于混合整数非线性规划的辐射型配电网可靠性规划方法.首先,选取了可靠性和经济性评估指标用于对配网改造方案可靠性和经济性进行评估分析.然后,以多目标的方式兼顾了可靠性与经济性的需求,构建基于混合整数非线性规划的多目标加权优化模型,在考虑配网技术约束和可靠性指标约束条件下得到Pareto曲线,采用模糊满意决策方法根据可靠性期望水平和配网改造预算选择合适的配网可靠性改造方案.最后通过33节点测试系统对所提方法进行测试,结果表明所提方法能在满足可靠性期望水平前提下帮助供电公司选择经济最优的可靠性改造方案.

More and more distributed generators (DGs) are connected to the distribution networks in recent, but complicates fault location process of DNs. To improve the power supply quality of the distribution network with DGs, higher precision and faster fault location methods are needed to adapt to complex network topologies. This paper proposes a fault locating method for the distribution network based on multiple objective particle swarm optimization (MOPSO). Firstly, according to the fault current characteristics of distribution networks with DGs, the mathematical model of fault location in the distribution network with DGs is established. And then, the switching function and evaluation function suitable of the distribution network with DGs are constructed. Secondly, the principle of MOPSO is introduced and the flow chart of fault location is analyzed. Finally, a simulation is carried out in the IEEE-33 standard network. The results show that the proposed method can locate single faults and multiple faults correctly under conditions of incomplete FTU fault information, and the accuracy and rapidity of the proposed method are improved to a certain extent.

As the core component of the distribution network, overhead lines are the key to ensuring the safe and reliable operation of the power system. However, under the influence of internal aging effects and external environmental effects, it will cause different failure effects on overhead lines. Compared with the traditional failure probability prediction of overhead lines, the real-time failure probability calculation requires the time-varying failure probability value of the overhead line. This paper proposes a time-varying failure probability calculation method for overhead lines in distribution networks based on the combination of historical statistics and meteorological monitoring. First, a calculation method for the internal failure rate of overhead lines based on exponential function fitting is proposed. Then, a technique based on SAE neural network training weather monitoring data to fit the external failure rate is offered. Finally, the Fokker Planck equation is used to calculate the overhead line’s overall time-varying failure probability value. Validation of calculation examples shows that the results calculated in this paper can effectively reflect the failure probability of overhead lines under different operating times and conditions.