Power distribution network reliability cooperative control method and system under distributed power supply access

Abstract

The invention provides a power distribution network reliability cooperative control method and system under distributed power supply access, relates to the technical field of electric power, and designs a power distribution network reliability index system considering a distributed power supply based on a load demand uncertainty model, and adopts a Monte Carlo simulation method. Evaluating the reliability of the power distribution network in combination with the power distribution network reliability index system, and formulating a cooperative control strategy including distributed power supply output control, load demand response management and network topology reconstruction according to an evaluation result; a multi-target cooperative control optimization model with reliability improvement, economical efficiency optimization and environmental benefit maximization as targets is established, the multi-target cooperative control optimization model is solved by adopting an improved particle swarm algorithm, and an optimal cooperative control strategy is generated; and based on the optimal cooperative control strategy, issuing and executing a control instruction to each control object.

Description

Technical Field

[0001] The present invention relates to electric power technology, and in particular to a distribution network reliability collaborative control method and system under distributed power source access. Background Art

[0002] Traditional distribution network reliability control methods mainly optimize a single objective, such as minimizing power outage time or maximizing power supply reliability indicators. However, in the new situation where distributed power sources are widely connected, these methods often fail to meet multiple needs. For example, only considering reliability indicators may ignore economic and environmental benefits; simply pursuing economic benefits may reduce system reliability. Therefore, it is urgent to develop a control method that can comprehensively consider multiple objectives and achieve coordinated optimization of distribution network reliability, economic and environmental benefits.

[0003] In the existing technology, although some studies have begun to focus on the multi-objective optimization problem of distribution networks, most methods have the following shortcomings: Lack of sufficient consideration of the characteristics of distributed power sources makes it difficult to effectively utilize their flexibility and adjustability; The efficiency of fault location and network reconstruction algorithms is low and it is difficult to meet the needs of real-time control; The optimization goal is single, making it difficult to achieve coordinated optimization of reliability, economy and environmental benefits; The control strategy lacks adaptability and is unable to cope with the dynamic changes in the distribution network operating environment. Summary of the invention

[0004] The present invention provides a distribution network reliability collaborative control method and system under distributed power supply access, which can solve the problems in the prior art.

[0005] A first aspect of the present invention, A distribution network reliability collaborative control method under distributed power generation access is provided, including: Based on the established distribution network topology model, the output characteristics of distributed power sources are analyzed and a load demand uncertainty model is constructed. Based on the load demand uncertainty model, a distribution network reliability index system considering distributed power sources is designed, including power supply reliability rate, average power outage time and load loss index. The Monte Carlo simulation method is used in combination with the distribution network reliability index system to evaluate the reliability of the distribution network, including distribution network state evaluation and reliability analysis. The distribution network state evaluation adopts a probabilistic power flow calculation method, taking into account equipment failure rate, repair time and load importance, and the reliability analysis is based on the conditional value risk assessment index. According to the evaluation results, a collaborative control strategy including distributed power output control, load demand response management and network topology reconstruction is formulated. A multi-objective collaborative control optimization model with the goals of reliability improvement, economic optimization and environmental benefit maximization is established, and distributed power output control, load demand response and network topology reconstruction are used as decision variables. Voltage constraints, power constraints and equipment capacity constraints are considered, and an improved particle swarm algorithm is used to solve the multi-objective collaborative control optimization model to generate the optimal collaborative control strategy. Based on the optimal collaborative control strategy, control instructions are issued to each control object and executed; fuzzy logic controllers are used to achieve real-time output regulation of distributed power sources, and graph theory algorithms are used for fault location and network reconstruction; the system operating status is monitored in real time and control parameters are adjusted dynamically; and the control effect is continuously optimized through a closed-loop feedback mechanism.

[0006] Preferably, Based on the established distribution network topology model, the steps of analyzing the output characteristics of distributed power sources and constructing the load demand uncertainty model include: Establishing a distribution network topology model, wherein the distribution network topology model is represented by a weighted undirected graph, wherein nodes represent buses in the distribution network, edges represent lines connecting nodes, and each edge is assigned a complex weight to represent line impedance; Constructing a node admittance matrix based on the distribution network topology model, and establishing an initial power flow equation using the node admittance matrix; Identifying distributed power access nodes in the distribution network topology model and analyzing the probability distribution of output characteristics of the distributed power access nodes, including: For photovoltaic power generation access nodes, the β probability distribution model is used to describe the randomness of the solar radiation intensity of the node, and the probability distribution of the photovoltaic cell output power of the node is calculated; For wind power generation access nodes, Weibull distribution is used to describe the randomness of the wind speed at the node, and the probability distribution of the wind turbine output power at the node is calculated; Integrate the probability distribution of the output characteristics of the distributed power access node into the initial power flow equation to obtain a modified power flow equation that takes into account the uncertainty of the distributed power output; Identifying load nodes in the distribution network topology model, and extracting initial load data of each load node based on the modified power flow equation; Perform cluster analysis on the extracted initial load data to determine the typical load pattern of each load node; Based on the typical load pattern of each load node, a load demand uncertainty model of the node is constructed; the constructed load demand uncertainty models of each load node are integrated to form a load demand uncertainty model of the entire distribution network.

[0007] Preferably, Based on the load data of each load node, the steps of constructing the load demand uncertainty model of the node include: A Gaussian mixture model is constructed based on the load data of each load node, and the Gaussian mixture model is expressed as: ; Where x is the load data of the load node, θ is the model parameter set, π k is the weight of the Gaussian component, the probability density function of the N Gaussian components, μ k is the Gaussian component mean parameter, is the variance, k is the index variable, and K is the total number of Gaussian components; The load data is preprocessed using the K-means clustering algorithm to obtain the initial value of the Gaussian mixture model parameter θ; Execute the expectation maximization algorithm to iteratively optimize the Gaussian mixture model, including the expectation step and the maximization step; in the expectation step, calculate the posterior probability that each load data point belongs to each Gaussian component; in the maximization step, update the π of each Gaussian component based on the calculated posterior probability k , μ k and ; Calculate the log-likelihood function of the updated Gaussian mixture model; if the log-likelihood function value is less than the preset log-likelihood threshold, terminate the iteration and generate the optimal Gaussian mixture model of the node; The Bayesian information criterion is calculated for different K values ​​to select the optimal model complexity; based on the optimal Gaussian mixture model and the optimal model complexity, the load demand uncertainty model of the node is generated.

[0008] Preferably, The Monte Carlo simulation method is used in combination with the distribution network reliability index system to evaluate the reliability of the distribution network, including distribution network status evaluation and reliability analysis. The distribution network status evaluation adopts a probabilistic power flow calculation method, taking into account equipment failure rate, repair time and load importance. The reliability analysis is based on the conditional value risk assessment index and includes the following steps: The distribution network status assessment adopts probabilistic power flow calculation, and the probabilistic power flow calculation adopts an improved Newton-Raphson method, including: establishing a node power balance equation considering the power injection characteristics of distributed power sources; initializing node voltage and phase angle; calculating power imbalance and Jacobian matrix; solving a linear equation system to obtain voltage and phase angle corrections; updating node voltage and phase angle; if the convergence condition is not met, returning to the step of calculating power imbalance, and if the convergence condition is met, outputting node voltage and power injection results, and the convergence condition is that the maximum absolute value of the power imbalance is less than the power imbalance threshold; Based on the probabilistic power flow calculation results, calculate the node voltage deviation, line power flow and system loss status evaluation indicators; The two-state Markov model is used to describe the equipment operation status and calculate the equipment availability; importance weights are assigned to load nodes; and the probabilistic power supply reliability rate, average power outage time and load loss index are calculated by combining equipment availability and load importance weights; A conditional value risk index is introduced to calculate the conditional value risk of the load loss index, wherein the conditional value risk is obtained by sorting the load loss index samples, determining the risk value, and calculating the average value of the samples exceeding the risk value; Through the Monte Carlo simulation method, the probability flow calculation and the reliability analysis based on the conditional value risk of the calculated load loss index are repeatedly performed to generate samples and perform statistical analysis. Based on the statistical analysis of the Monte Carlo simulation method, the evaluation results of the distribution network reliability are output, including the power supply reliability rate, the average power outage time and the load loss index.

[0009] Preferably, A multi-objective collaborative control optimization model with the goals of improving reliability, optimizing economy and maximizing environmental benefits is established. Distributed power output control, load demand response and network topology reconstruction are used as decision variables. Voltage constraints, power constraints and equipment capacity constraints are considered. An improved particle swarm algorithm is used to solve the multi-objective collaborative control optimization model. The steps to generate the optimal collaborative control strategy include: Constructing a reliability improvement objective function, wherein the reliability improvement objective function is represented by a weighted sum of a system average power outage frequency index and a system average power outage duration index, and a weight coefficient is determined by a hierarchical analysis method; Constructing an economic optimization objective function, wherein the economic optimization objective function takes into account the system operation cost and the economic benefits brought about by the reliability improvement, wherein the system operation cost includes the power generation cost, the network loss cost and the load removal cost, and the economic benefits brought about by the reliability improvement are quantified by the reduction of the power outage cost; Constructing an environmental benefit maximization objective function, wherein the environmental benefit maximization objective function is expressed by system carbon emissions and is calculated by carbon emission coefficients and outputs of various power generation units; Set constraints, including voltage constraints, power constraints, and equipment capacity constraints. The power constraints take into account the active and reactive output of the generator, the active and reactive demand of the load, the node voltage amplitude and phase angle, and the network admittance matrix. An improved particle swarm algorithm is used to solve the multi-objective collaborative control optimization model. The improved particle swarm algorithm includes: The non-dominated sorting strategy is introduced, and the population is stratified using the fast non-dominated sorting method; Introduce crowding distance calculation to maintain population diversity; Design an adaptive inertia weight strategy to dynamically adjust the inertia weight according to the number of iterations and particle fitness to balance global search and local search capabilities; Introducing Gaussian mutation operation to enhance the global search capability of the algorithm; A set of Pareto optimal solutions is obtained through the improved particle swarm algorithm to generate an optimal collaborative control strategy, which includes a distributed power output control scheme, a load demand response scheme and a network topology reconstruction scheme.

[0010] Preferably, The method further comprises: The Chebyshev decomposition method is used to transform the multi-objective problem into a series of single-objective sub-problems, and a weight vector is assigned to each sub-problem so that the weight vector is evenly distributed in the objective space. The optimal solution for each sub-problem is maintained to form an external elite file. Design adaptive scaling factors and crossover probabilities, perform differential evolution operations on each particle, generate test vectors, evaluate the fitness of the test vectors according to the Chebyshev method, and update the current particle position; assign a neighborhood to each particle, and dynamically adjust the neighborhood size as the iteration proceeds. When updating the particle position, consider the global optimal solution, individual optimal solution, and neighborhood optimal solution to achieve collaborative learning; design adaptive mutation probability, perform Gaussian mutation on each dimension, and the mutation amplitude decreases as the iteration proceeds; Maintain an external elite archive, add non-dominated solutions to the external elite archive after each iteration, prune the external elite archive based on the crowding distance when the size exceeds the archive size threshold, and dynamically adjust the size of the external elite archive; Perform local search on the solutions in the elite archive, use pattern search method to explore in the neighborhood of the current solution, and update the elite archive if a better solution is found; The algorithm terminates when the maximum number of iterations is reached or the preset convergence condition is met, and the non-dominated solution set in the elite archive is output as the Pareto optimal solution to generate the optimal collaborative control strategy.

[0011] Preferably, The steps of using fuzzy logic controller to realize real-time output regulation of distributed power sources and adopting graph theory algorithm for fault location and network reconstruction include: The distribution network is abstracted as an undirected graph, where nodes represent substations, switches and load points, and edges represent lines; The undirected graph is traversed by using a depth-first search algorithm, a voltage deviation threshold and a current deviation threshold are set, and when an abnormal node is detected, the node is added to a potential fault point set, and the previous normal node is traced back to mark the line between the two as a potential fault line; based on the obtained potential fault point set and potential fault line, the final fault location is determined; The improved Kruskal algorithm is used for network reconstruction, including the introduction of load imbalance index, radial topology constraints, and dynamic weight adjustment mechanism; all edges are sorted according to dynamic weights, and edges that meet the constraints are selected one by one to be added to the minimum spanning tree; in the process of selecting edges, voltage constraints, current constraints and load balance constraints are checked at the same time; according to the final generated minimum spanning tree, the reconstructed network topology is determined.

[0012] The second aspect of the present invention, Provides a distribution network reliability collaborative control system under distributed power access, including: The first unit is used to analyze the output characteristics of distributed power sources based on the established distribution network topology model and build a load demand uncertainty model; based on the load demand uncertainty model, a distribution network reliability index system considering distributed power sources is designed, including power supply reliability rate, average power outage time and load loss index; The second unit is used to evaluate the reliability of the distribution network by using the Monte Carlo simulation method in combination with the distribution network reliability index system, including distribution network status evaluation and reliability analysis, wherein the distribution network status evaluation adopts the probabilistic power flow calculation method, taking into account the equipment failure rate, repair time and load importance, and the reliability analysis is based on the conditional value risk assessment index; according to the evaluation results, formulate a collaborative control strategy including distributed power output control, load demand response management and network topology reconstruction; establish a multi-objective collaborative control optimization model with the goals of reliability improvement, economic optimization and environmental benefit maximization, taking distributed power output control, load demand response and network topology reconstruction as decision variables, considering voltage constraints, power constraints and equipment capacity constraints, and using an improved particle swarm algorithm to solve the multi-objective collaborative control optimization model to generate the optimal collaborative control strategy; The third unit is used to issue and execute control instructions to each control object based on the optimal collaborative control strategy; use fuzzy logic controller to realize real-time output regulation of distributed power sources, and adopt graph theory algorithm to locate faults and reconstruct the network; monitor the system operation status in real time and dynamically adjust the control parameters; and continuously optimize the control effect through a closed-loop feedback mechanism.

[0013] The third aspect of the present invention, An electronic device is provided, comprising: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0014] A fourth aspect of the present invention, A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0015] The method based on multi-objective collaborative control optimization adopted in this paper has many advantages. First, it can simultaneously consider the three objectives of reliability, economy and environmental benefits to achieve comprehensive optimization of the distribution network. Secondly, by taking distributed power output control, load demand response and network topology reconstruction as decision variables, the flexibility resources of the distribution network are fully utilized and the operation efficiency of the system is improved. Furthermore, the improved particle swarm algorithm can effectively handle such complex multi-objective optimization problems, obtain a set of Pareto optimal solutions, and provide decision makers with a variety of options. Finally, the application of this method helps to improve the reliability of the distribution network, reduce operating costs, reduce carbon emissions, and achieve the coordination and unification of economic, social and environmental benefits. By implementing the optimal collaborative control strategy, distribution network operators can better cope with challenges such as load fluctuations and renewable energy access, improve the flexibility and resilience of the system, and provide users with more reliable, economical and clean power services.

[0016] The method of combining fuzzy logic control and graph theory algorithm adopted in this paper has many advantages. First, the fuzzy logic controller can flexibly handle the uncertainty and complexity in the system, realize the intelligent regulation of distributed power sources, and improve the stability and economy of the system. Secondly, the fault location algorithm based on graph theory can quickly and accurately identify the fault location and reduce the scope and duration of power outages. Finally, the improved Kruskal algorithm takes into account network loss, load balance and various constraints when reconstructing the network, and can generate a more optimized network topology. This comprehensive method significantly improves the reliability, flexibility and economy of the distribution network, and provides strong support for the operation and management of the smart distribution network. By adjusting the output of distributed power sources in real time, quickly locating and isolating faults, and intelligently reconstructing the network, the power outage time can be greatly reduced, the power supply quality can be improved, the operating cost can be reduced, and users can be provided with more reliable and high-quality power services. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of a method for coordinated control of distribution network reliability under distributed power access according to an embodiment of the present invention; Figure 2 It is a schematic diagram of the structure of a distribution network reliability collaborative control system under distributed power access according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0020] Figure 1 FIG. 1 is a flow chart of a method for collaboratively controlling reliability of a distribution network under distributed power access according to an embodiment of the present invention. Figure 1 As shown, the method includes: S1. Based on the established distribution network topology model, analyze the output characteristics of distributed power sources and construct a load demand uncertainty model; based on the load demand uncertainty model, design a distribution network reliability index system considering distributed power sources, including power supply reliability rate, average power outage time and load loss index; S2. The Monte Carlo simulation method is used in combination with the distribution network reliability index system to evaluate the reliability of the distribution network, including distribution network status evaluation and reliability analysis, wherein the distribution network status evaluation adopts the probabilistic flow calculation method, taking into account the equipment failure rate, repair time and load importance, and the reliability analysis is based on the conditional value risk assessment index; according to the evaluation results, a collaborative control strategy including distributed power output control, load demand response management and network topology reconstruction is formulated; a multi-objective collaborative control optimization model with the goals of improving reliability, optimizing economy and maximizing environmental benefits is established, with distributed power output control, load demand response and network topology reconstruction as decision variables, considering voltage constraints, power constraints and equipment capacity constraints, and using an improved particle swarm algorithm to solve the multi-objective collaborative control optimization model to generate the optimal collaborative control strategy; S3. Based on the optimal collaborative control strategy, control instructions are issued to each control object and executed; fuzzy logic controllers are used to achieve real-time output regulation of distributed power sources, and graph theory algorithms are used for fault location and network reconstruction; the system operating status is monitored in real time, and control parameters are adjusted dynamically; and the control effect is continuously optimized through a closed-loop feedback mechanism.

[0021] In an optional embodiment, Based on the established distribution network topology model, the steps of analyzing the output characteristics of distributed power sources and constructing the load demand uncertainty model include: Establishing a distribution network topology model, wherein the distribution network topology model is represented by a weighted undirected graph, wherein nodes represent buses in the distribution network, edges represent lines connecting nodes, and each edge is assigned a complex weight to represent line impedance; Constructing a node admittance matrix based on the distribution network topology model, and establishing an initial power flow equation using the node admittance matrix; Identifying distributed power access nodes in the distribution network topology model and analyzing the probability distribution of output characteristics of the distributed power access nodes, including: For photovoltaic power generation access nodes, the β probability distribution model is used to describe the randomness of the solar radiation intensity of the node, and the probability distribution of the photovoltaic cell output power of the node is calculated; For wind power generation access nodes, Weibull distribution is used to describe the randomness of the wind speed at the node, and the probability distribution of the wind turbine output power at the node is calculated; Integrate the probability distribution of the output characteristics of the distributed power access node into the initial power flow equation to obtain a modified power flow equation that takes into account the uncertainty of the distributed power output; Identifying load nodes in the distribution network topology model, and extracting initial load data of each load node based on the modified power flow equation; Perform cluster analysis on the extracted initial load data to determine the typical load pattern of each load node; Based on the typical load pattern of each load node, a load demand uncertainty model of the node is constructed; the constructed load demand uncertainty models of each load node are integrated to form a load demand uncertainty model of the entire distribution network.

[0022] Exemplarily, the establishment of a distribution network topology model is the basis for analyzing the output characteristics of distributed power sources and constructing a load demand uncertainty model. The model is represented by a weighted undirected graph, where nodes represent buses in the distribution network and edges represent lines connecting nodes. Each edge is assigned a complex weight to represent the line impedance. For example, in a distribution network model containing 10 nodes, the line between node 1 and node 2 may be assigned a complex weight of (0.1+0.2j) ohms, indicating that the resistance of the line is 0.1 ohms and the reactance is 0.2 ohms.

[0023] Based on the established distribution network topology model, the node admittance matrix is ​​constructed. The node admittance matrix is ​​a complex matrix whose diagonal elements represent the sum of the admittances of all lines connected to the node, and the non-diagonal elements represent the negative admittance of the line connecting two nodes. Through the node admittance matrix, the initial power flow equation can be established, which describes the relationship between power flow and node voltage in the distribution network.

[0024] In the distribution network topology model, it is necessary to identify the distributed power access nodes and analyze the probability distribution of the output characteristics of these nodes. For photovoltaic power generation access nodes, the β probability distribution model is used to describe the randomness of the solar radiation intensity of the node. The parameters of the β distribution can be estimated from historical data. For example, the β distribution parameters of a node may be α=2.5 and β=5.0. Based on the β distribution model, the probability distribution of the photovoltaic cell output power of the node can be calculated.

[0025] For wind power access nodes, Weibull distribution is used to describe the randomness of the wind speed at the node. The scale parameter and shape parameter of Weibull distribution can also be estimated through historical data. For example, the Weibull distribution parameters of a node may be scale parameter c=8.0 and shape parameter k=2.0. Based on the Weibull distribution model, the probability distribution of the wind turbine output power at the node can be calculated.

[0026] The probability distribution of the output characteristics of the distributed generation access node obtained by analysis is integrated into the initial power flow equation to obtain a modified power flow equation that takes into account the uncertainty of the distributed generation output. This modification process involves replacing the original deterministic power generation term with a probability distribution expression.

[0027] The load nodes are identified in the distribution network topology model, and the initial load data of each load node is extracted based on the modified power flow equation. These initial load data are the basis for building the load demand uncertainty model.

[0028] Perform cluster analysis on the extracted initial load data to determine the typical load pattern of each load node. Cluster analysis can use K-means algorithm or hierarchical clustering algorithm to classify similar load curves into one category. For example, a load node may have three typical load patterns: weekdays, weekends, and holidays.

[0029] Based on the typical load pattern of each load node, the load demand uncertainty model of the node is constructed. Here, the Gaussian mixture model can be used to describe the uncertainty of load demand. The Gaussian mixture model consists of multiple Gaussian components, each of which corresponds to a typical load pattern. The model parameters include the weight, mean and variance of each Gaussian component, which can be estimated by the expectation maximization (EM) algorithm.

[0030] For example, for a node with three typical load patterns, its Gaussian mixture model may contain three Gaussian components, with weights of each component being 0.6, 0.3, and 0.1, respectively. The corresponding means may be 500kW, 300kW, and 200kW, and the variances may be 10,000, 5,000, and 2,000.

[0031] Finally, the constructed uncertainty models of the load nodes are integrated to form the uncertainty model of the load demand of the entire distribution network. This integration process needs to consider the correlation between nodes, which can be achieved by introducing the correlation coefficient matrix.

[0032] This method based on graph theory and probability model can effectively capture the uncertainty of distributed power output and load demand in the distribution network. By constructing a detailed topology model and probability distribution model, the operating status of the distribution network can be described more accurately, providing a reliable basis for subsequent optimization scheduling and risk assessment.

[0033] In an optional embodiment, Based on the load data of each load node, the steps of constructing the load demand uncertainty model of the node include: A Gaussian mixture model is constructed based on the load data of each load node, and the Gaussian mixture model is expressed as: ; Where x is the load data of the load node, θ is the model parameter set, π k is the weight of the Gaussian component, the probability density function of the N Gaussian components, μ k is the Gaussian component mean parameter, is the variance, k is the index variable, and K is the total number of Gaussian components; The load data is preprocessed using the K-means clustering algorithm to obtain the initial value of the Gaussian mixture model parameter θ; Execute the expectation maximization algorithm to iteratively optimize the Gaussian mixture model, including the expectation step and the maximization step; in the expectation step, calculate the posterior probability that each load data point belongs to each Gaussian component; in the maximization step, update the π of each Gaussian component based on the calculated posterior probability k , μ k and ; Calculate the log-likelihood function of the updated Gaussian mixture model; if the log-likelihood function value is less than the preset log-likelihood threshold, terminate the iteration and generate the optimal Gaussian mixture model of the node; The Bayesian information criterion is calculated for different K values ​​to select the optimal model complexity; based on the optimal Gaussian mixture model and the optimal model complexity, the load demand uncertainty model of the node is generated.

[0034] Exemplarily, constructing a Gaussian mixture model based on the load data of each load node is the core step to achieve load demand uncertainty modeling. The Gaussian mixture model is a powerful probabilistic model that can effectively capture the multimodal characteristics and uncertainties of load data. In this model, the load data is regarded as a mixed distribution consisting of multiple Gaussian components. Each Gaussian component represents a typical load pattern and is described by three parameters: weight, mean, and variance. For example, for a node with three typical load patterns, its Gaussian mixture model may contain three Gaussian components with weights of 0.5, 0.3, and 0.2, respectively. The corresponding means may be 500kW, 300kW, and 200kW, and the variances may be 10,000, 5,000, and 2,000.

[0035] In order to improve the training efficiency and convergence of the Gaussian mixture model, the K-means clustering algorithm (K-means clustering algorithm) is first used to preprocess the load data. The K-means algorithm divides the load data into K clusters, with the center of each cluster as the initial mean of the Gaussian component, the average square of the distance between the data points in the cluster and the center as the initial variance, and the proportion of data points in the cluster as the initial weight. This preprocessing method can provide a good starting point for subsequent parameter optimization. For example, for a node containing 1,000 load data points, if K=3 is selected, the K-means algorithm may divide the data into three clusters, containing 500, 300, and 200 data points, respectively, and the corresponding initial weights are 0.5, 0.3, and 0.2.

[0036] Next, the expectation maximization (EM) algorithm is executed to iteratively optimize the parameters of the Gaussian mixture model. The EM algorithm includes an expectation step (E-step) and a maximization step (M-step). In the expectation step, the posterior probability of each load data point belonging to each Gaussian component is calculated. This probability reflects the "affinity" of the data point to each Gaussian component. In the maximization step, the weight, mean, and variance of each Gaussian component are updated based on the calculated posterior probability. For example, if a data point has a high posterior probability for the first Gaussian component, then this data point will have a greater impact on the parameters of the first component when updating.

[0037] After each iteration, the log-likelihood function value of the updated Gaussian mixture model is calculated. The log-likelihood function reflects the degree of fit of the model to the observed data. If the growth rate of the log-likelihood function value is less than the preset threshold (such as 0.001), the model is considered to have converged and the iteration is terminated. At this time, the optimal Gaussian mixture model of the node is obtained.

[0038] In order to determine the optimal model complexity, that is, the number of Gaussian components K, it is necessary to calculate the Bayesian Information Criterion (BIC) for different K values. BIC seeks a balance between model fit and model complexity to avoid overfitting. Usually, a series of K values ​​are tried, such as 1 to 10, and the BIC corresponding to each K value is calculated. The K value that minimizes the BIC is selected as the optimal model complexity. For example, if the BIC is minimized when K=4, then the model containing 4 Gaussian components is selected as the final load demand uncertainty model.

[0039] Based on the optimal Gaussian mixture model and the optimal model complexity, a load demand uncertainty model for the node is generated. This model contains the optimal number of Gaussian components, each with an optimized weight, mean, and variance. For example, the final model may contain 4 Gaussian components with weights of 0.4, 0.3, 0.2, and 0.1, means of 600kW, 400kW, 300kW, and 200kW, and variances of 15,000, 10,000, 5,000, and 2,000.

[0040] This load demand uncertainty modeling method based on the Gaussian mixture model has many advantages. First, it can effectively capture the multimodal characteristics of load data and adapt to different time periods and different types of load patterns. Second, through the iterative optimization of the EM algorithm, the model can accurately fit the distribution characteristics of the actual load data. Furthermore, the BIC criterion is introduced to select the optimal model complexity, achieving a good balance between model accuracy and computational efficiency. Finally, this probabilistic model can not only describe the expected value of the load, but also quantify the uncertainty of the load, providing a solid foundation for subsequent risk assessment and robust optimization.

[0041] In an optional embodiment, The Monte Carlo simulation method is used in combination with the distribution network reliability index system to evaluate the reliability of the distribution network, including distribution network status evaluation and reliability analysis. The distribution network status evaluation adopts a probabilistic power flow calculation method, taking into account equipment failure rate, repair time and load importance. The reliability analysis is based on the conditional value risk assessment index and includes the following steps: The distribution network status assessment adopts probabilistic power flow calculation, and the probabilistic power flow calculation adopts an improved Newton-Raphson method, including: establishing a node power balance equation considering the power injection characteristics of distributed power sources; initializing node voltage and phase angle; calculating power imbalance and Jacobian matrix; solving a linear equation system to obtain voltage and phase angle corrections; updating node voltage and phase angle; if the convergence condition is not met, returning to the step of calculating power imbalance, and if the convergence condition is met, outputting node voltage and power injection results, and the convergence condition is that the maximum absolute value of the power imbalance is less than the power imbalance threshold; Based on the probabilistic power flow calculation results, calculate the node voltage deviation, line power flow and system loss status evaluation indicators; The two-state Markov model is used to describe the equipment operation status and calculate the equipment availability; importance weights are assigned to load nodes; and the probabilistic power supply reliability rate, average power outage time and load loss index are calculated by combining equipment availability and load importance weights; A conditional value risk index is introduced to calculate the conditional value risk of the load loss index, wherein the conditional value risk is obtained by sorting the load loss index samples, determining the risk value, and calculating the average value of the samples exceeding the risk value; Through the Monte Carlo simulation method, the probability flow calculation and the reliability analysis based on the conditional value risk of the calculated load loss index are repeatedly performed to generate samples and perform statistical analysis. Based on the statistical analysis of the Monte Carlo simulation method, the evaluation results of the distribution network reliability are output, including the power supply reliability rate, the average power outage time and the load loss index.

[0042] For example, the reliability assessment of the distribution network is a key link to ensure the safe and stable operation of the power system. The Monte Carlo simulation method, combined with a complete distribution network reliability index system, can comprehensively assess the reliability of the distribution network. This assessment process mainly includes two parts: distribution network status assessment and reliability analysis. The distribution network status assessment uses a probabilistic flow calculation method, and the reliability analysis is based on the conditional value risk assessment index.

[0043] The distribution network state assessment adopts probabilistic power flow calculation, specifically using the improved Newton-Raphson method (Newton iteration method). First, establish the node power balance equation that takes into account the power injection characteristics of distributed power sources. For example, for a node that contains photovoltaic power generation, its power balance equation needs to consider the randomness and intermittency of photovoltaic power generation. Initialize the node voltage and phase angle, and assume that the voltage amplitude of all nodes is 1.0 standard value and the phase angle is 0 degrees. Then calculate the power imbalance and Jacobian matrix. The power imbalance reflects the difference between the actual injected power and the calculated power of the node in the current state, while the Jacobian matrix describes the sensitivity of voltage and phase angle changes to power.

[0044] Next, solve the linear equations to obtain the corrections for the voltage and phase angle. For example, for a distribution network with 100 nodes, this step may involve solving a 200×200 linear equation system. After updating the node voltage and phase angle, check whether the convergence condition is met. Usually, the convergence condition is set as the maximum absolute value of the power imbalance is less than the preset power imbalance threshold, such as 0.0001 standard value. If the convergence condition is not met, return to the step of calculating the power imbalance; if the convergence condition is met, output the final node voltage and power injection results.

[0045] Based on the results of the probabilistic power flow calculation, the state evaluation indicators such as node voltage deviation, line power flow and system loss are calculated. For example, for a distribution network with a rated voltage of 10kV, if the voltage calculation result of a node is 9.8kV, its voltage deviation is 2%. Line power flow can be used to check whether there is an overload condition, while system loss reflects the operating efficiency of the network.

[0046] In reliability analysis, a two-state Markov model is used to describe the operating status of equipment. Each device has two states: normal operation and failure. The state transition is described by the failure rate and repair rate. For example, the annual failure rate of a transformer may be 0.05 times / year, and the average repair time is 10 hours. Based on this, its availability can be calculated to be about 99.94%. At the same time, importance weights are assigned to load nodes to reflect the priority of different loads. For example, the load node of a hospital may be given a weight of 1.5, while the weight of an ordinary residential area is 1.0.

[0047] Combining equipment availability and load importance weights, the probabilistic power supply reliability, average power outage time, and load loss index are calculated. For example, if the average number of power outages per year at a load point is 2, and each power outage lasts 1 hour, then its average power outage time is 2 hours / year. The load loss index can be estimated by multiplying the power outage time by the load size.

[0048] In order to better assess the risks in extreme situations, the conditional value risk indicator is introduced to calculate the conditional value risk of the load loss indicator. The specific steps include sorting the load loss indicator samples, determining the risk value, and calculating the average value of the samples that exceed the risk value. For example, if the risk level is set to 95%, the conditional value risk is the average load loss in the worst 5% of cases.

[0049] Through the Monte Carlo simulation method, the above-mentioned probabilistic power flow calculation and reliability analysis process are repeatedly executed to generate a large number of samples and perform statistical analysis. For example, 10,000 simulations may be required to obtain stable statistical results. Based on these statistical analyses, the evaluation results of the reliability of the distribution network are finally output, including the power supply reliability rate, the average power outage time, and the load loss index. For example, the evaluation results may show that the annual average power supply reliability rate of a distribution network is 99.99%, the average power outage time is 52.56 minutes / year, and the annual average load loss is 10,000 kWh.

[0050] This distribution network reliability assessment method based on Monte Carlo simulation and conditional value risk has many advantages. First, it can comprehensively consider various uncertain factors in the distribution network, including load fluctuations, distributed power output and equipment failure. Secondly, by introducing the conditional value risk index, this method can better assess the system risk under extreme conditions and provide decision makers with more comprehensive risk information. Furthermore, the use of the Monte Carlo simulation method makes the assessment results have high statistical reliability. Finally, this method can not only assess the reliability status of the existing distribution network, but also be used to compare the reliability performance of different planning schemes, providing an important basis for the planning and transformation of the distribution network. In general, the application of this assessment method will help improve the reliability level of the distribution network, reduce system risks, improve the power supply quality, and ultimately provide users with more reliable and high-quality power services.

[0051] In an optional embodiment, A multi-objective collaborative control optimization model with the goals of improving reliability, optimizing economy and maximizing environmental benefits is established. Distributed power output control, load demand response and network topology reconstruction are used as decision variables. Voltage constraints, power constraints and equipment capacity constraints are considered. An improved particle swarm algorithm is used to solve the multi-objective collaborative control optimization model. The steps to generate the optimal collaborative control strategy include: Constructing a reliability improvement objective function, wherein the reliability improvement objective function is represented by a weighted sum of a system average power outage frequency index and a system average power outage duration index, and a weight coefficient is determined by a hierarchical analysis method; Constructing an economic optimization objective function, wherein the economic optimization objective function takes into account the system operation cost and the economic benefits brought about by the reliability improvement, wherein the system operation cost includes the power generation cost, the network loss cost and the load removal cost, and the economic benefits brought about by the reliability improvement are quantified by the reduction of the power outage cost; Constructing an environmental benefit maximization objective function, wherein the environmental benefit maximization objective function is expressed by system carbon emissions and is calculated by carbon emission coefficients and outputs of various power generation units; Set constraints, including voltage constraints, power constraints, and equipment capacity constraints. The power constraints take into account the active and reactive output of the generator, the active and reactive demand of the load, the node voltage amplitude and phase angle, and the network admittance matrix. An improved particle swarm algorithm is used to solve the multi-objective collaborative control optimization model. The improved particle swarm algorithm includes: The non-dominated sorting strategy is introduced, and the population is stratified using the fast non-dominated sorting method; Introduce crowding distance calculation to maintain population diversity; Design an adaptive inertia weight strategy to dynamically adjust the inertia weight according to the number of iterations and particle fitness to balance global search and local search capabilities; Introducing Gaussian mutation operation to enhance the global search capability of the algorithm; A set of Pareto optimal solutions is obtained through the improved particle swarm algorithm to generate an optimal collaborative control strategy, which includes a distributed power output control scheme, a load demand response scheme and a network topology reconstruction scheme.

[0052] For example, in order to achieve efficient, reliable and environmentally friendly operation of the distribution network, it is an effective method to establish a multi-objective collaborative control optimization model with the goals of improving reliability, optimizing economy and maximizing environmental benefits. The model takes distributed power output control, load demand response and network topology reconstruction as decision variables, while considering voltage constraints, power constraints and equipment capacity constraints, and finally generates the optimal collaborative control strategy through the improved particle swarm algorithm.

[0053] First, the reliability improvement objective function is constructed. This function is expressed as the weighted sum of the system average outage frequency index (SAIFI) and the system average outage duration index (SAIDI). For example, assuming that the SAIFI of a distribution network is 2 times / year and the SAIDI is 4 hours / year, the weight of SAIFI is determined to be 0.6 and the weight of SAIDI is 0.4 through the analytic hierarchy process. The reliability improvement objective function can be expressed as the weighted sum of these two indicators. The analytic hierarchy process determines the weight coefficient by constructing a judgment matrix, calculating the eigenvector, and performing a consistency test.

[0054] Secondly, construct an economic optimization objective function. This function takes into account the system operating costs and the economic benefits brought by improved reliability. The system operating costs include power generation costs, network loss costs, and load shedding costs. For example, assuming that the daily power generation cost of a distribution network is 10,000 yuan, the network loss cost is 1,000 yuan, and the load shedding cost is 500 yuan, the total operating cost is 11,500 yuan / day. The economic benefits brought by improved reliability are quantified by the reduction in power outage costs. If the improved reliability reduces the annual power outage loss from 1 million yuan to 800,000 yuan, it can bring 200,000 yuan of economic benefits each year.

[0055] Next, construct the objective function of maximizing environmental benefits. This function is expressed in terms of system carbon emissions, which is calculated through the carbon emission coefficients and outputs of various power generation units. For example, assuming that the distribution network includes gas generators and photovoltaic power generation systems, the carbon emission coefficient of the gas generators is 0.5kg / kWh, and the daily power generation is 10,000kWh, and the carbon emission coefficient of the photovoltaic power generation system is 0kg / kWh, and the daily power generation is 5,000kWh, then the system's daily carbon emissions are 5,000kg.

[0056] Constraints are set in the model, including voltage constraints, power constraints, and equipment capacity constraints. Voltage constraints require that node voltages be maintained within an allowable range, such as 0.95~1.05 standard values. Power constraints consider the active and reactive output of generators, the active and reactive demands of loads, node voltage amplitudes and phase angles, and the network admittance matrix to ensure system power balance. Equipment capacity constraints require that the operating status of each device does not exceed its rated capacity, such as a transformer load rate of no more than 95%.

[0057] In order to solve this complex multi-objective optimization problem, an improved particle swarm algorithm is used. The algorithm first introduces a non-dominated sorting strategy and uses a fast non-dominated sorting method to stratify the population. For example, a population of 100 particles may be divided into 5 non-dominated layers. Then the crowding distance calculation is introduced to maintain the diversity of the population. The crowding distance reflects the distribution density of particles in the solution space, and particles with larger crowding distances are more likely to be retained.

[0058] During the algorithm iteration process, an adaptive inertia weight strategy is designed to dynamically adjust the inertia weight according to the number of iterations and particle fitness to balance the global search and local search capabilities. For example, a larger inertia weight (such as 0.9) can be used in the early iteration to enhance the global search capability, while a smaller inertia weight (such as 0.4) can be used in the later iteration to enhance the local search capability. At the same time, the Gaussian mutation operation is introduced to enhance the global search capability of the algorithm. Gaussian mutation can be performed on some dimensions of some particles, and the mutation probability can be set to 0.01.

[0059] Through the improved particle swarm algorithm, a set of Pareto optimal solutions are finally obtained to generate the optimal collaborative control strategy. The strategy includes distributed power output control scheme, load demand response scheme and network topology reconstruction scheme. For example, the distributed power output control scheme may include increasing the output of the photovoltaic power generation system during peak load hours; the load demand response scheme may include reducing 20% ​​of the controllable load during peak electricity price hours; the network topology reconstruction scheme may include opening a tie line in certain circumstances to reduce network losses.

[0060] This method based on multi-objective collaborative control optimization has many advantages. First, it can simultaneously consider the three objectives of reliability, economy and environmental benefits to achieve comprehensive optimization of the distribution network. Secondly, by taking distributed power output control, load demand response and network topology reconstruction as decision variables, the flexibility resources of the distribution network are fully utilized and the operation efficiency of the system is improved. Furthermore, the improved particle swarm algorithm can effectively handle such complex multi-objective optimization problems, obtain a set of Pareto optimal solutions, and provide decision makers with a variety of options. Finally, the application of this method helps to improve the reliability of the distribution network, reduce operating costs, reduce carbon emissions, and achieve the coordination and unification of economic, social and environmental benefits. By implementing the optimal collaborative control strategy, distribution network operators can better cope with challenges such as load fluctuations and renewable energy access, improve the flexibility and resilience of the system, and provide users with more reliable, economical and clean power services.

[0061] In an optional embodiment, The method further comprises: The Chebyshev decomposition method is used to transform the multi-objective problem into a series of single-objective sub-problems, and a weight vector is assigned to each sub-problem so that the weight vector is evenly distributed in the objective space. The optimal solution for each sub-problem is maintained to form an external elite file. Design adaptive scaling factors and crossover probabilities, perform differential evolution operations on each particle, generate test vectors, evaluate the fitness of the test vectors according to the Chebyshev method, and update the current particle position; assign a neighborhood to each particle, and dynamically adjust the neighborhood size as the iteration proceeds. When updating the particle position, consider the global optimal solution, individual optimal solution, and neighborhood optimal solution to achieve collaborative learning; design adaptive mutation probability, perform Gaussian mutation on each dimension, and the mutation amplitude decreases as the iteration proceeds; Maintain an external elite archive, add non-dominated solutions to the external elite archive after each iteration, prune the external elite archive based on the crowding distance when the size exceeds the archive size threshold, and dynamically adjust the size of the external elite archive; Perform local search on the solutions in the elite archive, use pattern search method to explore in the neighborhood of the current solution, and update the elite archive if a better solution is found; The algorithm terminates when the maximum number of iterations is reached or the preset convergence condition is met, and the non-dominated solution set in the elite archive is output as the Pareto optimal solution to generate the optimal collaborative control strategy.

[0062] For example, in order to further improve the solution efficiency and solution quality of the multi-objective collaborative control optimization model, this method introduces a series of advanced algorithm strategies. These strategies include Chebyshev decomposition, differential evolution, adaptive parameter adjustment, collaborative learning, external elite archive maintenance, and local search techniques, forming a comprehensive optimization framework.

[0063] First, the Chebyshev decomposition method is used to transform the multi-objective problem into a series of single-objective sub-problems. This decomposition method is based on the Chebyshev norm and can effectively handle the nonlinear relationship between objective functions. For example, for a three-objective optimization problem, it can be decomposed into 100 single-objective sub-problems. Assign a weight vector to each sub-problem so that these weight vectors are evenly distributed in the objective space. Assume that the weight vector of the first sub-problem is (0.8, 0.1, 0.1), the weight vector of the second sub-problem is (0.7, 0.2, 0.1), and so on. Maintain the optimal solution for each sub-problem, and these optimal solutions constitute the external elite archive.

[0064] Next, we design adaptive scaling factors and crossover probabilities and perform differential evolution operations on each particle. Differential evolution is an effective evolutionary algorithm that generates new test vectors by vector difference. For example, we can set the initial scaling factor to 0.5 and the initial crossover probability to 0.9, and then dynamically adjust these parameters based on the evolutionary effect of the algorithm. After the test vector is generated, the Chebyshev method is used to evaluate the fitness of the test vector, and the current particle position is updated based on the evaluation results.

[0065] In order to enhance the algorithm's search capability, a neighborhood is assigned to each particle, and the neighborhood size is dynamically adjusted as the iteration proceeds. For example, the neighborhood size can be initially set to 20% of the population size, and then gradually reduced to 5% as the iteration proceeds. When updating the particle position, the global optimal solution, the individual optimal solution, and the neighborhood optimal solution are considered at the same time to achieve collaborative learning. This strategy can balance global search and local search, and improve the algorithm's convergence speed and solution quality.

[0066] In order to maintain population diversity, an adaptive mutation probability is designed to perform Gaussian mutation on each dimension. The mutation probability can be initially set to 0.1 and gradually decreases as the iteration proceeds. The mutation amplitude also decreases as the iteration proceeds, for example, it can be gradually reduced from the initial standard deviation of 0.1 to 0.01. This strategy can increase exploration in the early stages of the search and increase development in the later stages.

[0067] During the operation of the algorithm, it is crucial to maintain the external elite archive. After each iteration, the newly generated non-dominated solutions are added to the external elite archive. When the size of the external elite archive exceeds the preset archive size threshold, pruning is performed based on the crowding distance. For example, the initial archive size threshold can be set to 100. When the number of solutions in the archive exceeds this number, the 100 solutions with the largest crowding distance are retained. At the same time, the size of the external elite archive can be dynamically adjusted. The archive size can be gradually increased according to the convergence of the algorithm, such as from 100 to 200.

[0068] In order to further improve the quality of the solution, a local search is performed on the solutions in the elite archive. The pattern search method is used to explore in the neighborhood of the current solution. For example, a perturbation of plus or minus 0.1% can be performed on each dimension of the current solution to evaluate the fitness of the perturbed solution. If a better solution is found, the elite archive is updated. This local search strategy can help the algorithm escape the local optimum and find a better solution.

[0069] The algorithm terminates when it reaches the maximum number of iterations or meets the preset convergence condition. For example, the maximum number of iterations can be set to 1000, or the algorithm is considered to have converged when the solution in the elite file has not been significantly improved for 50 consecutive iterations. Finally, the non-dominated solution set in the elite file is output as the Pareto optimal solution to generate the optimal collaborative control strategy.

[0070] This improved multi-objective optimization method has many advantages. First, the Chebyshev decomposition method can effectively handle the complex relationship between objective functions and transform the multi-objective problem into a single-objective sub-problem that is easy to solve. Second, the differential evolution operation and adaptive parameter adjustment strategy improve the search efficiency and adaptability of the algorithm. The collaborative learning mechanism balances the global search and local search capabilities of the algorithm by considering global, individual and neighborhood information. The maintenance and dynamic adjustment of external elite archives ensure the preservation of high-quality solutions and the maintenance of population diversity. The local search strategy further improves the quality of the solution. These improvements enable the algorithm to better handle complex multi-objective optimization problems and obtain a more diverse and evenly distributed Pareto optimal solution set. In the multi-objective collaborative control optimization of distribution networks, this method can provide decision makers with more high-quality options and better balance reliability, economy and environmental benefits, thereby realizing the intelligent, efficient and green operation of distribution networks.

[0071] In an optional embodiment, The steps of using fuzzy logic controller to realize real-time output regulation of distributed power sources and adopting graph theory algorithm for fault location and network reconstruction include: The distribution network is abstracted as an undirected graph, where nodes represent substations, switches and load points, and edges represent lines; The undirected graph is traversed by using a depth-first search algorithm, a voltage deviation threshold and a current deviation threshold are set, and when an abnormal node is detected, the node is added to a potential fault point set, and the previous normal node is traced back to mark the line between the two as a potential fault line; based on the obtained potential fault point set and potential fault line, the final fault location is determined; The improved Kruskal algorithm is used for network reconstruction, including the introduction of load imbalance index, radial topology constraints, and dynamic weight adjustment mechanism; all edges are sorted according to dynamic weights, and edges that meet the constraints are selected one by one to be added to the minimum spanning tree; in the process of selecting edges, voltage constraints, current constraints and load balance constraints are checked at the same time; according to the final generated minimum spanning tree, the reconstructed network topology is determined.

[0072] For example, in the multi-objective coordinated control optimization of distribution networks, it is crucial to achieve real-time output regulation of distributed power sources and network fault handling. Fuzzy logic controllers and graph theory algorithms can effectively implement these functions and improve the reliability and flexibility of distribution networks.

[0073] First, a fuzzy logic controller is used to realize real-time output regulation of distributed power sources. Fuzzy logic controllers are based on human expert knowledge and experience and can handle uncertainty and nonlinearity in the system. In this application, the input of the fuzzy logic controller can include load demand changes, system frequency deviations, and voltage deviations, and the output is the output adjustment of distributed power sources. For example, a fuzzy controller with five input linguistic variables (load change rate, system frequency deviation, voltage deviation, current output level, and available capacity) and one output linguistic variable (output adjustment) can be designed. Each linguistic variable can be divided into several fuzzy sets, such as "negative large", "negative small", "zero", "positive small", "positive large", etc. By designing a suitable fuzzy rule base, such as "if the load change rate is positive large and the system frequency deviation is negative small, the output adjustment is positive medium", intelligent output regulation can be realized. The fuzzy reasoning process adopts the Mamdani method, and the defuzzification adopts the centroid method. This method can flexibly adjust the output of distributed power sources according to the real-time status of the system and improve the stability and economy of the system.

[0074] Next, the distribution network is abstracted into an undirected graph to facilitate fault location and network reconstruction. In this undirected graph, nodes represent substations, switches, and load points, and edges represent lines. For example, an undirected graph with 100 nodes and 120 edges can represent a medium-sized distribution network. Each node contains attribute information such as voltage and current, and each edge contains attribute information such as impedance and capacity. This abstraction makes it possible to use graph theory algorithms for network analysis and optimization.

[0075] The depth-first search (DFS) algorithm is used to traverse the undirected graph to locate the fault. The voltage deviation threshold (such as ±5% of the rated voltage) and the current deviation threshold (such as ±10% of the rated current) are set. The traversal starts from the source node. When an abnormal node is detected, the node is added to the set of potential fault points, and the previous normal node is traced back to mark the line between the two as a potential fault line. For example, if the voltage of node 25 is found to be 0.93 per unit during the traversal process, which is lower than the threshold of 0.95, node 25 is added to the set of potential fault points, and the line connecting node 25 and its upstream normal node is marked as a potential fault line. Based on the obtained set of potential fault points and potential fault lines, combined with other information (such as fault indicator status, user repair information, etc.), the final fault location is determined. This method can quickly and accurately locate faults and provide a basis for subsequent isolation and recovery operations.

[0076] After the fault is located, network reconstruction is required to restore power supply. The improved Kruskal algorithm is used for network reconstruction, which introduces a number of improvements on the basis of the traditional Kruskal algorithm. First, the load imbalance index is introduced to calculate the load distribution between each feeder as an optimization goal for reconstruction. Secondly, the radial topology constraint is introduced to ensure that the reconstructed network still maintains a radial structure to avoid the formation of loops. Thirdly, a dynamic weight adjustment mechanism is adopted to dynamically adjust the weight of each edge according to the current network status. For example, the weight of the edge can be set to the weighted sum of the loss, load rate and importance of its corresponding line, and the weight coefficient is dynamically adjusted according to the current operating status.

[0077] When executing the improved Kruskal algorithm, all edges are first sorted according to the dynamic weights. Then the edges that meet the constraints are selected one by one to be added to the minimum spanning tree. In the process of selecting edges, voltage constraints (such as the node voltage must be within the range of 0.95~1.05 per unit), current constraints (such as the line load rate does not exceed 90%) and load balance constraints (such as the load difference between each feeder does not exceed 20%) are checked at the same time. If the addition of an edge will violate these constraints, the edge is skipped and the next edge is considered. The final minimum spanning tree is the reconstructed network topology. For example, in a network containing 100 nodes, the final minimum spanning tree may contain 99 edges, forming a new radial network structure.

[0078] This method of combining fuzzy logic control and graph theory algorithm has many advantages. First, the fuzzy logic controller can flexibly handle the uncertainty and complexity in the system, realize the intelligent regulation of distributed power sources, and improve the stability and economy of the system. Secondly, the fault location algorithm based on graph theory can quickly and accurately identify the fault location and reduce the scope and duration of power outages. Finally, the improved Kruskal algorithm takes into account network losses, load balance and various constraints when reconstructing the network, and can generate a more optimized network topology. This comprehensive method significantly improves the reliability, flexibility and economy of the distribution network, and provides strong support for the operation and management of the smart distribution network. By adjusting the output of distributed power sources in real time, quickly locating and isolating faults, and intelligently reconstructing the network, the power outage time can be greatly reduced, the power supply quality can be improved, the operating cost can be reduced, and users can be provided with more reliable and high-quality power services.

[0079] Figure 2 Schematic diagram of the structure of the distribution network reliability collaborative control system under the access of distributed power sources according to an embodiment of the present invention. Figure 2 As shown, the system comprises: The first unit is used to analyze the output characteristics of distributed power sources based on the established distribution network topology model and build a load demand uncertainty model; based on the load demand uncertainty model, a distribution network reliability index system considering distributed power sources is designed, including power supply reliability rate, average power outage time and load loss index; The second unit is used to evaluate the reliability of the distribution network by using the Monte Carlo simulation method in combination with the distribution network reliability index system, including distribution network status evaluation and reliability analysis, wherein the distribution network status evaluation adopts the probabilistic power flow calculation method, taking into account the equipment failure rate, repair time and load importance, and the reliability analysis is based on the conditional value risk assessment index; according to the evaluation results, formulate a collaborative control strategy including distributed power output control, load demand response management and network topology reconstruction; establish a multi-objective collaborative control optimization model with the goals of reliability improvement, economic optimization and environmental benefit maximization, taking distributed power output control, load demand response and network topology reconstruction as decision variables, considering voltage constraints, power constraints and equipment capacity constraints, and using an improved particle swarm algorithm to solve the multi-objective collaborative control optimization model to generate the optimal collaborative control strategy; The third unit is used to issue and execute control instructions to each control object based on the optimal collaborative control strategy; use fuzzy logic controller to realize real-time output regulation of distributed power sources, and adopt graph theory algorithm to locate faults and reconstruct the network; monitor the system operation status in real time and dynamically adjust the control parameters; and continuously optimize the control effect through a closed-loop feedback mechanism.

[0080] According to a third aspect of the embodiments of the present invention, An electronic device is provided, comprising: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0081] According to a fourth aspect of the embodiments of the present invention, A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0082] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A distribution network reliability collaborative control method under distributed power generation access, characterized in that: include: Based on the established distribution network topology model, the output characteristics of distributed power sources are analyzed and a load demand uncertainty model is constructed; Design a distribution network reliability index system considering distributed power sources based on the load demand uncertainty model, including power supply reliability rate, average power outage time and load loss index; The Monte Carlo simulation method is used in combination with the distribution network reliability index system to evaluate the reliability of the distribution network, including distribution network status evaluation and reliability analysis. The distribution network status evaluation adopts a probabilistic power flow calculation method, taking into account equipment failure rate, repair time and load importance, and the reliability analysis is based on the conditional value risk assessment index. According to the evaluation results, a collaborative control strategy including distributed power output control, load demand response management and network topology reconstruction is formulated. A multi-objective collaborative control optimization model with the goals of improving reliability, optimizing economy and maximizing environmental benefits is established. Distributed power output control, load demand response and network topology reconstruction are used as decision variables. Voltage constraints, power constraints and equipment capacity constraints are considered. An improved particle swarm algorithm is used to solve the multi-objective collaborative control optimization model and generate the optimal collaborative control strategy. Based on the optimal collaborative control strategy, control instructions are issued to each control object and executed; fuzzy logic controllers are used to achieve real-time output regulation of distributed power sources, and graph theory algorithms are used for fault location and network reconstruction; the system operating status is monitored in real time and control parameters are adjusted dynamically; and the control effect is continuously optimized through a closed-loop feedback mechanism.

2. The method according to claim 1, characterized in that Based on the established distribution network topology model, the steps of analyzing the output characteristics of distributed power sources and constructing the load demand uncertainty model include: Establishing a distribution network topology model, wherein the distribution network topology model is represented by a weighted undirected graph, wherein nodes represent buses in the distribution network, edges represent lines connecting nodes, and each edge is assigned a complex weight to represent line impedance; Constructing a node admittance matrix based on the distribution network topology model, and establishing an initial power flow equation using the node admittance matrix; Identifying distributed power access nodes in the distribution network topology model and analyzing the probability distribution of output characteristics of the distributed power access nodes, including: For photovoltaic power generation access nodes, the β probability distribution model is used to describe the randomness of the solar radiation intensity of the node, and the probability distribution of the photovoltaic cell output power of the node is calculated; For wind power generation access nodes, Weibull distribution is used to describe the randomness of the wind speed at the node, and the probability distribution of the wind turbine output power at the node is calculated; Integrate the probability distribution of the output characteristics of the distributed power access node into the initial power flow equation to obtain a modified power flow equation that takes into account the uncertainty of the distributed power output; Identifying load nodes in the distribution network topology model, and extracting initial load data of each load node based on the modified power flow equation; Perform cluster analysis on the extracted initial load data to determine the typical load pattern of each load node; Based on the typical load pattern of each load node, a load demand uncertainty model of the node is constructed; the constructed load demand uncertainty models of each load node are integrated to form a load demand uncertainty model of the entire distribution network.

3. The method according to claim 2, characterized in that Based on the load data of each load node, the steps of constructing the load demand uncertainty model of the node include: A Gaussian mixture model is constructed based on the load data of each load node, and the Gaussian mixture model is expressed as: ; Where x is the load data of the load node, θ is the model parameter set, π k is the weight of the Gaussian component, the probability density function of the N Gaussian components, μ k is the Gaussian component mean parameter, is the variance, k is the index variable, and K is the total number of Gaussian components; The load data is preprocessed using the K-means clustering algorithm to obtain the initial value of the Gaussian mixture model parameter θ; Execute the expectation maximization algorithm to iteratively optimize the Gaussian mixture model, including the expectation step and the maximization step; in the expectation step, calculate the posterior probability that each load data point belongs to each Gaussian component; in the maximization step, update the π of each Gaussian component based on the calculated posterior probability k , μ k and ; Calculate the log-likelihood function of the updated Gaussian mixture model; if the log-likelihood function value is less than the preset log-likelihood threshold, terminate the iteration and generate the optimal Gaussian mixture model of the node; The Bayesian information criterion is calculated for different K values ​​to select the optimal model complexity; based on the optimal Gaussian mixture model and the optimal model complexity, the load demand uncertainty model of the node is generated.

4. The method according to claim 1, characterized in that: The Monte Carlo simulation method is used in combination with the distribution network reliability index system to evaluate the reliability of the distribution network, including distribution network status evaluation and reliability analysis. The distribution network status evaluation adopts a probabilistic power flow calculation method, taking into account equipment failure rate, repair time and load importance. The reliability analysis is based on the conditional value risk assessment index and includes the following steps: The distribution network status assessment adopts probabilistic power flow calculation, and the probabilistic power flow calculation adopts an improved Newton-Raphson method, including: establishing a node power balance equation considering the power injection characteristics of distributed power sources; initializing node voltage and phase angle; calculating power imbalance and Jacobian matrix; solving a linear equation system to obtain voltage and phase angle corrections; updating node voltage and phase angle; if the convergence condition is not met, returning to the step of calculating power imbalance, and if the convergence condition is met, outputting node voltage and power injection results, and the convergence condition is that the maximum absolute value of the power imbalance is less than the power imbalance threshold; Based on the probabilistic power flow calculation results, calculate the node voltage deviation, line power flow and system loss status evaluation indicators; The two-state Markov model is used to describe the equipment operation status and calculate the equipment availability; importance weights are assigned to load nodes; and the probabilistic power supply reliability rate, average power outage time and load loss index are calculated by combining equipment availability and load importance weights; A conditional value risk index is introduced to calculate the conditional value risk of the load loss index, wherein the conditional value risk is obtained by sorting the load loss index samples, determining the risk value, and calculating the average value of the samples exceeding the risk value; Through the Monte Carlo simulation method, the probability flow calculation and the reliability analysis based on the conditional value risk of the calculated load loss index are repeatedly performed to generate samples and perform statistical analysis. Based on the statistical analysis of the Monte Carlo simulation method, the evaluation results of the distribution network reliability are output, including the power supply reliability rate, the average power outage time and the load loss index.

5. The method according to claim 1, characterized in that A multi-objective collaborative control optimization model with the goals of improving reliability, optimizing economy and maximizing environmental benefits is established. Distributed power output control, load demand response and network topology reconstruction are used as decision variables. Voltage constraints, power constraints and equipment capacity constraints are considered. An improved particle swarm algorithm is used to solve the multi-objective collaborative control optimization model. The steps to generate the optimal collaborative control strategy include: Constructing a reliability improvement objective function, wherein the reliability improvement objective function is represented by a weighted sum of a system average power outage frequency index and a system average power outage duration index, and a weight coefficient is determined by a hierarchical analysis method; Constructing an economic optimization objective function, wherein the economic optimization objective function takes into account the system operation cost and the economic benefits brought about by the reliability improvement, wherein the system operation cost includes the power generation cost, the network loss cost and the load removal cost, and the economic benefits brought about by the reliability improvement are quantified by the reduction of the power outage cost; Constructing an environmental benefit maximization objective function, wherein the environmental benefit maximization objective function is expressed by system carbon emissions and is calculated by carbon emission coefficients and outputs of various power generation units; Set constraints, including voltage constraints, power constraints, and equipment capacity constraints. The power constraints take into account the active and reactive output of the generator, the active and reactive demand of the load, the node voltage amplitude and phase angle, and the network admittance matrix. An improved particle swarm algorithm is used to solve the multi-objective collaborative control optimization model. The improved particle swarm algorithm includes: The non-dominated sorting strategy is introduced, and the population is stratified using the fast non-dominated sorting method; Introduce crowding distance calculation to maintain population diversity; Design an adaptive inertia weight strategy to dynamically adjust the inertia weight according to the number of iterations and particle fitness, balancing global search and local search capabilities; Introducing Gaussian mutation operation to enhance the global search capability of the algorithm; A set of Pareto optimal solutions is obtained through the improved particle swarm algorithm to generate an optimal collaborative control strategy, which includes a distributed power output control scheme, a load demand response scheme and a network topology reconstruction scheme.

6. The method according to claim 5, characterized in that The method further comprises: The Chebyshev decomposition method is used to transform the multi-objective problem into a series of single-objective sub-problems, and a weight vector is assigned to each sub-problem so that the weight vector is evenly distributed in the objective space. The optimal solution for each sub-problem is maintained to form an external elite file. Design adaptive scaling factors and crossover probabilities, perform differential evolution operations on each particle, generate test vectors, evaluate the fitness of the test vectors according to the Chebyshev method, and update the current particle position; assign a neighborhood to each particle, and dynamically adjust the neighborhood size as the iteration proceeds. When updating the particle position, consider the global optimal solution, individual optimal solution, and neighborhood optimal solution to achieve collaborative learning; design adaptive mutation probability, perform Gaussian mutation on each dimension, and the mutation amplitude decreases as the iteration proceeds; Maintain an external elite archive, add non-dominated solutions to the external elite archive after each iteration, prune the external elite archive based on the crowding distance when the size exceeds the archive size threshold, and dynamically adjust the size of the external elite archive; Perform local search on the solutions in the elite archive, use pattern search method to explore in the neighborhood of the current solution, and update the elite archive if a better solution is found; The algorithm terminates when the maximum number of iterations is reached or the preset convergence condition is met, and the non-dominated solution set in the elite archive is output as the Pareto optimal solution to generate the optimal collaborative control strategy.

7. The method according to claim 1, characterized in that The steps of using fuzzy logic controller to realize real-time output regulation of distributed power sources and adopting graph theory algorithm for fault location and network reconstruction include: The distribution network is abstracted as an undirected graph, where nodes represent substations, switches and load points, and edges represent lines; The undirected graph is traversed by using a depth-first search algorithm, a voltage deviation threshold and a current deviation threshold are set, and when an abnormal node is detected, the node is added to a potential fault point set, and the previous normal node is traced back to mark the line between the two as a potential fault line; based on the obtained potential fault point set and potential fault line, the final fault location is determined; The improved Kruskal algorithm is used for network reconstruction, including the introduction of load imbalance index, radial topology constraints, and dynamic weight adjustment mechanism; all edges are sorted according to dynamic weights, and edges that meet the constraints are selected one by one to be added to the minimum spanning tree; in the process of selecting edges, voltage constraints, current constraints and load balance constraints are checked at the same time; according to the final generated minimum spanning tree, the reconstructed network topology is determined.

8. A distribution network reliability collaborative control system with distributed power generation access, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to analyze the output characteristics of distributed power sources and build a load demand uncertainty model based on the established distribution network topology model; Design a distribution network reliability index system considering distributed power sources based on the load demand uncertainty model, including power supply reliability rate, average power outage time and load loss index; The second unit is used to evaluate the reliability of the distribution network by using the Monte Carlo simulation method in combination with the distribution network reliability index system, including distribution network status evaluation and reliability analysis, wherein the distribution network status evaluation adopts the probabilistic power flow calculation method, taking into account the equipment failure rate, repair time and load importance, and the reliability analysis is based on the conditional value risk assessment index; according to the evaluation results, a coordinated control strategy including distributed power output control, load demand response management and network topology reconstruction is formulated; A multi-objective collaborative control optimization model with the goals of improving reliability, optimizing economy and maximizing environmental benefits is established. Distributed power output control, load demand response and network topology reconstruction are used as decision variables. Voltage constraints, power constraints and equipment capacity constraints are considered. An improved particle swarm algorithm is used to solve the multi-objective collaborative control optimization model and generate the optimal collaborative control strategy. The third unit is used to issue and execute control instructions to each control object based on the optimal collaborative control strategy; use fuzzy logic controller to realize real-time output regulation of distributed power sources, and adopt graph theory algorithm to locate faults and reconstruct the network; monitor the system operation status in real time and dynamically adjust the control parameters; and continuously optimize the control effect through a closed-loop feedback mechanism.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images