Power distribution network fault self-recovery method and device and electronic equipment

Abstract

The invention provides a power distribution network fault self-recovery method and device and electronic equipment, and belongs to the technical field of power distribution networks. The method comprises the steps of obtaining fault data of a power distribution network; inputting the fault data into a fault recovery strategy model to obtain a fault recovery strategy output by the fault recovery strategy model; the fault recovery strategy model is trained based on sample fault data of the power distribution network, an action space corresponding to the sample fault data and a reward function corresponding to the sample fault data, and the reward function is constructed based on a sample risk level value of the power distribution network. According to the method, the sample risk level value of the power distribution network is introduced into the reward function of deep reinforcement learning, so that dynamic assessment and intelligent decision-making of the risk are realized when the fault recovery strategy is determined by using the fault recovery strategy model, and rapid fault recovery is realized on the premise of ensuring safety.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network, and in particular to a distribution network fault self-healing method, a distribution network fault self-healing device, an electronic device, a machine-readable storage medium and a computer program product. Background Art

[0002] With the widespread adoption of distributed power sources (such as photovoltaics and wind power) in power systems, their randomness and volatility have placed higher demands on the stable operation of the power grid. Traditional fault handling methods rely on preset rules and manual intervention, making them difficult to cope with complex and changing fault scenarios. They also suffer from slow recovery speeds, which can easily lead to distributed power generation disconnection or power system instability. Existing fault recovery methods often lack dynamic risk assessment and intelligent decision-making capabilities, making it difficult to achieve rapid recovery while ensuring safety. Summary of the Invention

[0003] The purpose of the embodiments of the present invention is to provide a distribution network fault self-healing method, device and electronic equipment to address the defects of the fault recovery methods in the prior art that generally lack dynamic risk assessment and intelligent decision-making capabilities, making it difficult to achieve rapid recovery while ensuring safety.

[0004] To achieve the above objectives, an embodiment of the present invention provides a distribution network fault self-healing method, comprising:

[0005] Obtain fault data of the distribution network;

[0006] Inputting the fault data into a fault recovery strategy model to obtain a fault recovery strategy output by the fault recovery strategy model;

[0007] The fault recovery strategy model is trained based on sample fault data of the distribution network, the action space corresponding to the sample fault data, and the reward function corresponding to the sample fault data. The reward function is constructed based on the sample risk level value of the distribution network.

[0008] Optionally, the sample risk level value of the distribution network is obtained by the following steps:

[0009] Determining target risk scenario data from sample risk scenario data of the distribution network;

[0010] Determine, based on the target risk scenario data, the sum of sample voltage over-limit risk values of all nodes in the distribution network and the sum of sample overload risk values of all lines;

[0011] A sample risk level value of the distribution network is determined based on the sum of the sample voltage over-limit risk values and the sum of the sample overload risk values.

[0012] Optionally, determining target risk scenario data from sample risk scenario data of the distribution network includes:

[0013] A K-medoids algorithm is used to determine target risk scenario data from sample risk scenario data of the distribution network; wherein the K-medoids algorithm determines initial clustering points based on a density function of the sample risk scenario data.

[0014] Optionally, the determining target risk scenario data from sample risk scenario data of the distribution network using a K-medoids algorithm includes:

[0015] Obtain multiple sample risk scenario data for the distribution network;

[0016] Calculating the density function value of all the sample risk scenario data;

[0017] Determining, from a density function value set, sample risk scenario data corresponding to a plurality of density function values as initial clustering points; the density function value set includes a plurality of density function values whose density function values are greater than a set threshold;

[0018] Repeat the following steps until the set stop condition is achieved:

[0019] Assign each non-initial cluster point to the cluster to which the nearest initial cluster point belongs;

[0020] Traversing each of the clusters, replacing the initial cluster point with the first non-initial cluster point of the cluster as the updated cluster point; wherein, if the total distance from all non-updated cluster points to the updated cluster point is smaller than the total distance from all non-initial cluster points to the initial cluster point, then the replacement is accepted, and the first non-initial cluster point is any non-initial cluster point of the cluster;

[0021] Target risk scenario data is determined based on the updated cluster points.

[0022] Optionally, determining a sample risk level value of the distribution network based on the sum of the sample voltage over-limit risk values and the sum of the sample overload risk values includes:

[0023] The sample risk level value of the distribution network is determined based on the sum of the sample voltage over-limit risk values, the sum of the sample overload risk values, and the load-loss operation risk of the distribution network.

[0024] Optionally, the reward function is constructed based on the recovery time of the distribution network, the power supply reliability value, and the risk level value of the distribution network.

[0025] Optionally, the method further includes:

[0026] constructing an objective function for minimizing the sum of the switching operation cost and the line risk level based on costs of multiple switching operations in the distribution network, a switching operation state corresponding to each of the switching operations, risk level values of multiple lines, and a line restoration state corresponding to each of the lines;

[0027] Solving the objective function, and determining a grid reconstruction plan for the distribution network based on the solution of the objective function;

[0028] A droop control method or a virtual synchronous machine control method is used to match the output data of the distributed power source with the power demand of the distribution network, so as to achieve the reconnection of the distributed power source to the grid.

[0029] On the other hand, an embodiment of the present invention further provides a distribution network fault self-healing device, comprising:

[0030] An acquisition module, used to obtain fault data of the distribution network;

[0031] A self-healing module, configured to input the fault data into a fault recovery strategy model to obtain a fault recovery strategy output by the fault recovery strategy model;

[0032] The fault recovery strategy model is trained based on sample fault data of the distribution network, the action space corresponding to the sample fault data, and the reward function corresponding to the sample fault data. The reward function is constructed based on the sample risk level value of the distribution network.

[0033] Optionally, the sample risk level value of the distribution network is obtained by the following steps:

[0034] Determining target risk scenario data from sample risk scenario data of the distribution network;

[0035] Determining the sum of sample voltage over-limit risk values of all nodes in the distribution network and the sum of sample overload risk values of all lines based on the target risk scenario data;

[0036] A sample risk level value of the distribution network is determined based on the sum of the sample voltage over-limit risk values and the sum of the sample overload risk values.

[0037] Optionally, determining target risk scenario data from sample risk scenario data of the distribution network includes:

[0038] A K-medoids algorithm is used to determine target risk scenario data from sample risk scenario data of the distribution network; wherein the K-medoids algorithm determines initial clustering points based on a density function of the sample risk scenario data.

[0039] Optionally, the determining target risk scenario data from sample risk scenario data of the distribution network using a K-medoids algorithm includes:

[0040] Obtain multiple sample risk scenario data for the distribution network;

[0041] Calculating the density function value of all the sample risk scenario data;

[0042] Determining, from a density function value set, sample risk scenario data corresponding to a plurality of density function values as initial clustering points; the density function value set includes a plurality of density function values whose density function values are greater than a set threshold;

[0043] Repeat the following steps until the set stop condition is achieved:

[0044] Assign each non-initial cluster point to the cluster to which the nearest initial cluster point belongs;

[0045] Traversing each of the clusters, replacing the initial cluster point with the first non-initial cluster point of the cluster as the updated cluster point; wherein, if the total distance from all non-updated cluster points to the updated cluster point is smaller than the total distance from all non-initial cluster points to the initial cluster point, then the replacement is accepted, and the first non-initial cluster point is any non-initial cluster point of the cluster;

[0046] Target risk scenario data is determined based on the updated cluster points.

[0047] Optionally, determining a sample risk level value of the distribution network based on the sum of the sample voltage over-limit risk values and the sum of the sample overload risk values includes:

[0048] The sample risk level value of the distribution network is determined based on the sum of the sample voltage over-limit risk values, the sum of the sample overload risk values, and the load-loss operation risk of the distribution network.

[0049] Optionally, the reward function is constructed based on the recovery time of the distribution network, the power supply reliability value, and the risk level value of the distribution network.

[0050] Optionally, the device further includes:

[0051] A construction module is configured to construct an objective function for minimizing the sum of the switching operation costs and the line risk level based on costs of a plurality of switching operations in the distribution network, a switching operation state corresponding to each of the switching operations, risk level values of a plurality of lines, and a line restoration state corresponding to each of the lines;

[0052] A solution module, configured to solve the objective function and determine a grid reconstruction scheme for the distribution network based on a solution result of the objective function;

[0053] The grid-connected module is used to use a droop control method or a virtual synchronous machine control method to match the distributed power output data with the power demand of the distribution network to achieve distributed power re-connection.

[0054] On the other hand, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned distribution network fault self-healing method when executing the program.

[0055] On the other hand, the present invention also provides a machine-readable storage medium having a computer program stored thereon, which implements the above-mentioned distribution network fault self-healing method when executed by a processor.

[0056] On the other hand, the present invention further provides a computer program product, comprising a computer program, which implements the above-mentioned distribution network fault self-healing method when executed by a processor.

[0057] Through the above technical solution, the fault recovery strategy model in this embodiment of the present invention is trained using deep reinforcement learning, and the reward function in deep reinforcement learning is constructed based on the sample risk level values of the distribution network. By incorporating the sample risk level values of the distribution network into the reward function of deep reinforcement learning, the fault recovery strategy model enables dynamic risk assessment and intelligent decision-making when determining the fault recovery strategy, achieving rapid fault recovery while ensuring safety.

[0058] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following detailed description, they are used to explain the embodiments of the present invention, but do not constitute a limitation of the embodiments of the present invention. In the accompanying drawings:

[0060] Figure 1 This is one of the flow charts of the distribution network fault self-healing method provided by the present invention;

[0061] Figure 2 It is a schematic diagram of the comprehensive operation risk of the distribution network provided by the present invention;

[0062] Figure 3 This is the second flow chart of the distribution network fault self-healing method provided by the present invention;

[0063] Figure 4 It is a structural diagram of the distribution network fault self-healing device provided by the present invention;

[0064] Figure 5It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0065] The following describes the specific implementation of the embodiment of the present invention in detail with reference to the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the embodiment of the present invention and is not used to limit the embodiment of the present invention.

[0066] Method Example

[0067] Please refer to Figure 1 , an embodiment of the present invention provides a distribution network fault self-healing method, comprising:

[0068] Step 100: Obtain fault data of the distribution network.

[0069] Step 200: Input the fault data into a fault recovery strategy model to obtain a fault recovery strategy output by the fault recovery strategy model.

[0070] The fault data may be operating data of the distribution network, including at least one of load data, distributed power generation output data, and network topology information of the distribution network. The load data may include at least one of node voltage, node current, node power, frequency, and line load factor of the distribution network. The electronic device inputs the fault data into a fault recovery strategy model to obtain a fault recovery strategy output by the fault recovery strategy model. The fault recovery strategy model is trained based on sample fault data of the distribution network, the action space corresponding to the sample fault data, and a reward function corresponding to the sample fault data. The reward function is constructed based on sample risk level values of the distribution network.

[0071] The sample fault data may be historical fault data of the distribution network. The historical fault data may be at least one of historical load data, historical distributed power output data, and historical network topology information of the distribution network collected from a SCADA (Supervisory Control And Data Acquisition) system, a PMU (phasor measurement unit), a distributed power supply monitoring system, etc. The historical load data may include at least one of the historical node voltage, historical node current, historical node power, historical frequency, and historical line load rate of the distribution network. The electronic device performs data cleaning, normalization, and feature extraction on the collected historical load data, removes noise and outliers, and extracts key features (such as load fluctuation rate, distributed power output change rate, etc.).

[0072] The embodiment of the present invention constructs a fault recovery strategy model through a deep reinforcement learning algorithm. The embodiment of the present invention defines the state space (referred to as state) as at least one of the historical node voltage, historical node current, historical node power, historical frequency, and historical line load rate of the distribution network; and defines the possible action space (referred to as action) as the switch operation of the distribution network, the output data adjustment of the distributed power supply, etc. The reward function (referred to as reward) is defined as being constructed based on the sample risk level value of the distribution network. The sample risk level value is obtained by considering the potential risk of the comprehensive state of the distribution network.

[0073] Deep reinforcement learning combines deep neural networks with Q learning, and uses deep neural networks to approximate the Q-value function. First, a deep neural network is used as a function approximator: the Q-value function is approximated by a deep neural network, the input is the state, and the output is the Q-value of each action. Then experience replay is performed: during the interaction with the environment, the agent stores the experience (state, action, reward, new state) of each time step in an experience replay pool, and randomly extracts small batches of samples for training during the training process to break the correlation between samples and stabilize the training process. Finally, the target network (fault recovery strategy model) is introduced as a calculation for the stable target value, and the parameters of the target network are updated to the parameters of the current deep Q network (referred to as Q network) every certain number of steps. In one embodiment, the fault recovery strategy model is obtained by the following steps:

[0074] Initialize the parameters of the experience replay pool, deep Q network and target network;

[0075] Repeat the following steps until the set conditions are achieved:

[0076] Obtain the state (i.e., sample fault data) and select actions (switching operations, output data adjustment of distributed generation, etc.) based on the ε-greedy strategy:

[0077] Execute actions and observe the rewards and new states calculated based on the reward function;

[0078] Store the experience (state, action, reward and new state) in the replay pool;

[0079] Randomly sample a set number of experiences from the replay pool;

[0080] Calculate the target Q value and predicted Q value based on each sample;

[0081] Based on the target Q value and predicted Q value of each sample, the loss function is calculated using mean square error; based on the loss function, the model parameters of the deep Q network are updated through back propagation and optimizer (such as Adam);

[0082] The model parameters of the deep Q network are copied to the target network (fault recovery strategy model) every fixed number of steps (e.g., every 100 training steps).

[0083] It should be noted that during distribution network operation, real-time data collection can be used to update the model parameters of the fault recovery strategy model online, improving the adaptability and accuracy of the fault recovery strategy model. By training the fault recovery strategy model based on deep reinforcement learning, the fault recovery strategy can dynamically adapt to complex and changing fault scenarios, improving the intelligent level of fault recovery.

[0084] In one embodiment, the sample risk level value of the distribution network is obtained by the following steps:

[0085] Step 11: Determine target risk scenario data from sample risk scenario data of the distribution network.

[0086] The electronic device determines the target risk scenario data from the sample risk scenario data of the distribution network. The sample risk scenario data can generate risk scenario data in various scenarios. However, in order to avoid increasing the computational burden, scenario reduction clustering processing should be performed to improve the operation speed while ensuring that the simulation accuracy of uncertain variables is not affected. Therefore, the electronic device can determine the target risk scenario data from the sample risk scenario data of the distribution network through clustering processing (such as K-means algorithm, K-medoids algorithm, etc.). However, the K-means algorithm is susceptible to interference in the selection of the initial cluster center point or the use of the data mean or weighted mean within the cluster as the value particle of the cluster. In contrast, the K-medoids algorithm selects the point closest to the cluster center as the representative point of the cluster. It can effectively reduce the interference of outlier data and noise data. However, both methods have the problem of randomly selecting initial cluster points, which leads to the problem of local optimality of clustering results.

[0087] Therefore, an embodiment of the present invention adds a density function to the K-medoids algorithm to optimally select initial clustering points, thereby determining target risk scenario data from sample risk scenario data of the distribution network. Specifically, in one embodiment, step 11, determining target risk scenario data from sample risk scenario data of the distribution network, includes: utilizing the K-medoids algorithm to determine target risk scenario data from sample risk scenario data of the distribution network; wherein the K-medoids algorithm determines initial clustering points based on the density function of the sample risk scenario data. The scenario distance is defined using Euclidean distance.

[0088] Specifically, the K-medoids algorithm is used to determine the target risk scenario data from the sample risk scenario data of the distribution network, including:

[0089] Step 111: Acquire multiple sample risk scenario data of the distribution network.

[0090] The electronic device obtains multiple sample risk scenario data for the distribution network. The multiple sample risk scenario data can be sample voltage over-limit risk values of nodes and sample overload risk values of lines in various scenarios (different weather scenarios, different power load scenarios). Different weather scenarios can include sunny, rainy, foggy, windy, or snowy days. Different power load scenarios can include peak power consumption scenarios, low power consumption scenarios, and low power consumption scenarios between peak and low power consumption scenarios.

[0091] The embodiment of the present invention considers the potential risk of the comprehensive state of the distribution network. It should be combined with the actual power risk assessment application. Generally, the indicators are considered from the spatial layer, which are divided into equipment layer and distribution network layer risk indicators, such as Figure 2 As shown. The comprehensive operation risk of the distribution network mainly includes the sample low voltage risk value (low voltage risk) of the node, the sample overvoltage risk value (overvoltage risk) and the sample overload risk value (line overload risk) of the line, etc., which is intended to qualitatively analyze the risk. In other embodiments, the comprehensive operation risk of the distribution network also includes the risk of the distribution network losing load operation (distribution network losing load risk). The sample low voltage risk value of the node, the sample overvoltage risk value, and the sample overload risk value of the line are calculated respectively by the following formulas:

[0092]

[0093] in, are the sample low voltage risk value and sample overvoltage risk value of the i-th node at any time, and Ω is the scenario set. t,s Indicates the probability of occurrence of various scenarios; It represents the sample overload risk value of the jth line in any period; Sev νh (v) represents the severity function of each risk; B(v) is a scalar from 0 to 1, representing whether the potential risk failure occurs. t represents the time. s represents the scenario, v i,t,s Indicates that the computational effort is in the scenario of node i at time t, v j,t,s Indicates the amount of computation in the scenario of line j at time t and s.

[0094] Step 112: Calculate the density function value of all the sample risk scenario data.

[0095] The electronic device calculates the density function value of all the sample risk scenario data. Among all the data samples, there are data points x and x', where x is any sample risk scenario data, and x' is the influence function of x, namely the density function Density(x,x'), which is essentially the sum of the influence functions of all the nearest neighboring data points on it within a given neighborhood parameter δ. Suppose the data sample set X composed of all the sample risk scenario data is {x1,x2,…,x n} for each data point x i , where x i ∈X, the calculated density value within the neighborhood radius is as follows:

[0096]

[0097] Where, d(x i ,x j ) 2 is the Euclidean distance, δ is the neighborhood radius, usually set to 0.5. i There are k nearest neighbor data points within the neighborhood radius δ, x j represents the nearest neighbor data point within the neighborhood radius δ.

[0098] Step 113: Determine sample risk scenario data corresponding to multiple density function values from the density function value set as initial clustering points; the density function value set includes multiple density function values whose density function values are greater than a set threshold.

[0099] The set of density function values may be a set of density function values ranked highest (e.g., the top 10) in a descending order of density function values. The electronic device obtains the density function of each data point in the data sample set. Then, based on the maximum distance principle, K initial cluster centers (initial clustering points) are selected from the set of density function values. Therefore, by adding the density function to the K-medoids algorithm, optimal selection of initial clustering points is achieved, avoiding the problem of randomly selecting initial clustering points, which may lead to local optimality in the clustering results.

[0100] Repeat the following steps until the set stop condition is achieved:

[0101] Step 114: Assign each non-initial cluster point to the cluster to which the nearest initial cluster point belongs.

[0102] Step 115: traverse each of the clusters and replace the initial cluster point with the first non-initial cluster point of the cluster as the updated cluster point; wherein, if the total distance from all non-updated cluster points to the updated cluster point is smaller than the total distance from all non-initial cluster points to the initial cluster point, the replacement is accepted, and the first non-initial cluster point is any non-initial cluster point of the cluster.

[0103] The electronic device assigns each non-initial cluster point to the cluster to which the closest initial cluster point among the K initial cluster points belongs. The electronic device then iterates through each cluster, replacing the initial cluster point with each non-initial cluster point in the cluster as an updated cluster point. If the total distance from all non-updated cluster points to the updated cluster point is less than the total distance from all non-initial cluster points to the initial cluster point, the replacement is accepted. This process is repeated until no better medoids are available or the maximum number of iterations is reached.

[0104] Step 116: Determine target risk scenario data based on the updated cluster points.

[0105] The electronic device identifies all updated cluster points as target risk scenario data. By using the updated cluster points as target risk scenario data to replace the sample risk scenario data within the corresponding cluster, the data size can be reduced while retaining key information. This embodiment of the present invention improves computational speed by reducing the risk scenario clustering process.

[0106] Step 12: Determine the sum of the sample voltage over-limit risk values of all nodes in the distribution network and the sum of the sample overload risk values of all lines based on the target risk scenario data.

[0107] The electronic device determines the sum of the sample voltage over-limit risk values of all nodes of the distribution network and the sum of the sample overload risk values of all lines based on the target risk scenario data. In another embodiment, the embodiment of the present invention defines the target risk scenario data, including the sum of the sample overload risk values of all lines, the sum of the sample voltage over-limit risk values of all nodes, and the frequency instability risk value. The electronic device uses a fuzzy comprehensive evaluation method or a Bayesian network to quantitatively assess the risk. The sample low voltage risk value of all nodes, the sample overvoltage risk value, the sample line overload risk value of all lines, and the load-loss operation risk of the distribution network are calculated as shown in the following formula:

[0108]

[0109] Among them, T represents the evaluation period, N b Represents the total number of nodes in the distribution network, N i Indicates the total number of lines in the distribution network. is the sum of the sample low voltage risk values and the sample overvoltage risk values of all nodes; is the sum of the sample overload risk values of all lines; is the risk of load loss operation of distribution network, R i,t load represents the load loss operation risk of the distribution network at the i-th node, and t represents the time.

[0110] Step 13: Determine a sample risk level value of the distribution network based on the sum of the sample voltage over-limit risk values and the sum of the sample overload risk values.

[0111] In one embodiment, the electronic device forms a comprehensive feeder-level low-voltage operation risk and overvoltage operation risk by simply calculating the sum of the sample low-voltage risk values and the sample overvoltage risk values of each node at the device level. The cumulative sum of the sample overload risk values of each line forms the distribution network line overload operation risk. The sum of these operation risks weighted by the weight coefficient is calculated to obtain the sample risk level value of the distribution network:

[0112]

[0113] Among them, R all is the comprehensive risk within the time range, where the total weight W = w1 + w2 + w3 = 1.

[0114] In another embodiment, step 13, determining the sample risk level value of the distribution network based on the sum of the sample voltage over-limit risk values and the sum of the sample overload risk values, includes: determining the sample risk level value of the distribution network based on the sum of the sample voltage over-limit risk values, the sum of the sample overload risk values and the load-loss operation risk of the distribution network.

[0115] By simply calculating the sum of the sample low voltage risk values and the sample overvoltage risk values of each node at the quantitative device level, a comprehensive feeder-level low voltage and overvoltage operation risk is formed. The cumulative sum of the overload risk values of each line forms the distribution network line overload operation risk. The sum of these operation risks and the distribution network load loss operation risk weighted by the weight coefficient is calculated to obtain the sample risk level value of the distribution network:

[0116]

[0117] Among them, R all is the comprehensive risk within the time range, where the total weight W = w1 + w2 + w3 + w4 = 1.

[0118] In the embodiment of the present invention, the action with the best Q value is selected as the fault recovery strategy from among the multiple actions output by the fault recovery strategy model. For example, the line with a lower load rate is preferentially selected for recovery operation. The embodiment of the present invention can also dynamically adjust the fault recovery strategy in the action space based on the calculation result of the reward function. For example, when the model is trained, based on the calculation result of the reward function, it is found that the load rate of a certain line is close to the upper limit, that is, there is a risk of line overload, then the action of "automatically switching to the backup recovery path" is added to the action space. Therefore, the embodiment of the present invention effectively reduces the potential risks in the recovery process and improves the safety and stability of the system. The embodiment of the present invention fully considers the various risks in fault recovery by adding the sample risk level value of the distribution network to the reward function, and realizes dynamic risk assessment and intelligent decision-making when determining the fault recovery strategy using the fault recovery strategy model, thereby achieving rapid fault recovery while ensuring safety.

[0119] The fault recovery strategy model in this embodiment of the present invention is trained using deep reinforcement learning, and the reward function in deep reinforcement learning is constructed based on sample risk level values of the distribution network. By incorporating these sample risk levels into the deep reinforcement learning reward function, the fault recovery strategy model enables dynamic risk assessment and intelligent decision-making when determining the fault recovery strategy, achieving rapid fault recovery while ensuring safety.

[0120] In other aspects of the embodiments of the present invention, in order to further consider the recovery time and power supply reliability of the distribution network during the fault recovery of the distribution network, in one embodiment, a reward function can be constructed based on the recovery time of the distribution network, the power supply reliability value, and the risk level value of the distribution network. Specifically, the reward function can be calculated using the following formula:

[0121] R=w1·T recovery +w2·P reliability +w3·R risk ;

[0122] Among them, T recovery is the recovery time, P reliability is the power supply reliability value, R risk is the risk level, and w1, w2, and w3 are weighting factors. It should be noted that the present invention allows preset recovery times for different actions. Thus, given a specific action, the corresponding recovery time can be determined. The power supply reliability value can be calculated based on the ratio of the user's effective power supply time to the total statistical time. For example, if the user's average annual power outage duration is 5 minutes, the power supply reliability value is ≥ 99.999%.

[0123] For other aspects of the present invention, please refer to Figure 3 , the method further comprises:

[0124] Step 300: Based on the costs of multiple switch operations in the distribution network, the switch operation status corresponding to each of the switch operations, the risk level values of multiple lines, and the line recovery status corresponding to each of the lines, construct an objective function for minimizing the sum of the switch operation costs and the line risk level.

[0125] Step 400: Solve the objective function, and determine a grid reconstruction plan for the distribution network based on the solution of the objective function.

[0126] The electronic device may also construct an objective function for minimizing the sum of the switching operation cost and the line risk level based on the cost of multiple switching operations in the distribution network, the switching operation status corresponding to each of the switching operations, the risk level values of multiple lines, and the line restoration status corresponding to each of the lines. In one embodiment, the objective function is:

[0127] Among them, C i is the cost of the i-th switch operation, x i For switch operation state, R j is the risk level value of the jth line, y j = is the line recovery state, m is the total number of switch operations, and n is the total number of lines. Embodiments of the present invention may also set constraints including topology connectivity constraints, voltage constraints, and load rate constraints. The electronic device may employ a heuristic algorithm (such as a genetic algorithm or a particle swarm optimization algorithm) to solve the objective function and determine a distribution network reconfiguration scheme based on the solution to the objective function, thereby achieving distribution network reconfiguration, i.e., reconfiguring the distribution network topology.

[0128] Step 500: Utilize a droop control method or a virtual synchronous machine control method to match the output data of the distributed power source with the power demand of the distribution network, so as to achieve the reconnection of the distributed power source to the grid.

[0129] Electronic devices can use droop control to simulate the active power-frequency (Pf) and reactive power-voltage (QV) droop characteristics of traditional synchronous generators. By adjusting the output frequency and voltage of distributed power sources, they achieve autonomous power distribution and frequency / voltage stability. When system frequency decreases, active power output increases; when frequency increases, active power output decreases. When voltage decreases, reactive power output increases; when voltage increases, reactive power output decreases. Alternatively, electronic devices can use virtual synchronous machine control to simulate the mechanical and electromagnetic characteristics of synchronous generators, giving distributed power sources inertia, damping, and frequency / voltage regulation capabilities. Mechanical equations: The rotor motion equations of synchronous generators are simulated by introducing the moment of inertia (J) and damping coefficient (D). Electromagnetic equations: The electromagnetic torque of synchronous generators is simulated by using the power-voltage / frequency droop characteristics. Furthermore, electronic devices can coordinate the output of multiple distributed power sources through distributed optimization algorithms to avoid secondary impacts on the power grid. For example, electronic devices can coordinate the output of multiple distributed power sources using a consensus algorithm. Through local communication, each distributed power source reaches consensus on target values (such as power allocation), eliminating the need for a central controller. For example, distributed power sources (DGs) dynamically adjust their output to match the power demand of the distribution network through communication with adjacent nodes. This embodiment of the present invention coordinates grid reconstruction and DG reconnection to the grid, enabling rapid power transfer and self-healing recovery, significantly improving the safe absorption capacity of DGs.

[0130] In summary, the embodiments of the present invention utilize real-time and historical data to train a fault recovery strategy model, dynamically optimize the fault recovery strategy, and assess and mitigate potential risks during the recovery process to ensure system safety. Furthermore, by adjusting the distribution network topology, rapidly restoring power supply paths, and coordinating the output of distributed power sources to ensure it matches system requirements, rapid power transfer and self-healing recovery are achieved, thereby improving the safe absorption capacity of distributed power sources.

[0131] Device embodiment

[0132] Please refer to Figure 4 On the other hand, an embodiment of the present invention further provides a distribution network fault self-healing device, comprising:

[0133] An acquisition module 401 is used to acquire fault data of the distribution network;

[0134] A self-healing module 402 is configured to input the fault data into a fault recovery strategy model to obtain a fault recovery strategy output by the fault recovery strategy model;

[0135] The fault recovery strategy model is trained based on sample fault data of the distribution network, the action space corresponding to the sample fault data, and the reward function corresponding to the sample fault data. The reward function is constructed based on the sample risk level value of the distribution network.

[0136] Optionally, the sample risk level value of the distribution network is obtained by the following steps:

[0137] Determining target risk scenario data from sample risk scenario data of the distribution network;

[0138] Determining the sum of sample voltage over-limit risk values of all nodes in the distribution network and the sum of sample overload risk values of all lines based on the target risk scenario data;

[0139] A sample risk level value of the distribution network is determined based on the sum of the sample voltage over-limit risk values and the sum of the sample overload risk values.

[0140] Optionally, determining target risk scenario data from sample risk scenario data of the distribution network includes:

[0141] A K-medoids algorithm is used to determine target risk scenario data from sample risk scenario data of the distribution network; wherein the K-medoids algorithm determines initial clustering points based on a density function of the sample risk scenario data.

[0142] Optionally, the determining target risk scenario data from sample risk scenario data of the distribution network using a K-medoids algorithm includes:

[0143] Obtain multiple sample risk scenario data for the distribution network;

[0144] Calculating the density function value of all the sample risk scenario data;

[0145] Determining, from a density function value set, sample risk scenario data corresponding to a plurality of density function values as initial clustering points; the density function value set includes a plurality of density function values whose density function values are greater than a set threshold;

[0146] Repeat the following steps until the set stop condition is achieved:

[0147] Assign each non-initial cluster point to the cluster to which the nearest initial cluster point belongs;

[0148] Traversing each of the clusters, replacing the initial cluster point with the first non-initial cluster point of the cluster as the updated cluster point; wherein, if the total distance from all non-updated cluster points to the updated cluster point is smaller than the total distance from all non-initial cluster points to the initial cluster point, then the replacement is accepted, and the first non-initial cluster point is any non-initial cluster point of the cluster;

[0149] Target risk scenario data is determined based on the updated cluster points.

[0150] Optionally, determining a sample risk level value of the distribution network based on the sum of the sample voltage over-limit risk values and the sum of the sample overload risk values includes:

[0151] The sample risk level value of the distribution network is determined based on the sum of the sample voltage over-limit risk values, the sum of the sample overload risk values, and the load-loss operation risk of the distribution network.

[0152] Optionally, the reward function is constructed based on the recovery time of the distribution network, the power supply reliability value, and the risk level value of the distribution network.

[0153] Optionally, the device further includes:

[0154] A construction module is configured to construct an objective function for minimizing the sum of the switching operation costs and the line risk level based on costs of a plurality of switching operations in the distribution network, a switching operation state corresponding to each of the switching operations, risk level values of a plurality of lines, and a line restoration state corresponding to each of the lines;

[0155] A solution module, configured to solve the objective function and determine a grid reconstruction scheme for the distribution network based on a solution result of the objective function;

[0156] The grid-connected module is used to use a droop control method or a virtual synchronous machine control method to match the distributed power output data with the power demand of the distribution network to achieve distributed power re-connection.

[0157] The distribution network fault self-healing device includes a processor and a memory. The acquisition module 401, self-healing module 402, etc. are stored in the memory as program units, and the processor executes the program units stored in the memory to implement corresponding functions.

[0158] The processor includes a kernel, which calls the corresponding program unit from the memory. There can be one or more kernels.

[0159] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0160] Figure 5 An example of a physical structure diagram of an electronic device is shown below. Figure 5As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other via the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute a distribution network fault self-healing method, which includes: obtaining fault data of the distribution network; inputting the fault data into a fault recovery strategy model to obtain a fault recovery strategy output by the fault recovery strategy model; the fault recovery strategy model is trained based on sample fault data of the distribution network, the action space corresponding to the sample fault data, and the reward function corresponding to the sample fault data, and the reward function is constructed based on the sample risk level value of the distribution network.

[0161] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0162] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a machine-readable storage medium. When the computer program is executed by a processor, the computer can execute a distribution network fault self-healing method, which includes: obtaining fault data of the distribution network; inputting the fault data into a fault recovery strategy model to obtain a fault recovery strategy output by the fault recovery strategy model; the fault recovery strategy model is trained based on sample fault data of the distribution network, the action space corresponding to the sample fault data, and the reward function corresponding to the sample fault data, and the reward function is constructed based on the sample risk level value of the distribution network.

[0163] On the other hand, the present invention also provides a machine-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform a distribution network fault self-healing method, the method comprising: obtaining fault data of the distribution network; inputting the fault data into a fault recovery strategy model to obtain a fault recovery strategy output by the fault recovery strategy model; the fault recovery strategy model is trained based on sample fault data of the distribution network, the action space corresponding to the sample fault data, and the reward function corresponding to the sample fault data, and the reward function is constructed based on the sample risk level value of the distribution network.

[0164] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0165] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0166] 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 make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A distribution network fault self-healing method, characterized in that: include: Obtain fault data of the distribution network; Inputting the fault data into a fault recovery strategy model to obtain a fault recovery strategy output by the fault recovery strategy model; The fault recovery strategy model is trained based on sample fault data of the distribution network, the action space corresponding to the sample fault data, and the reward function corresponding to the sample fault data. The reward function is constructed based on the sample risk level value of the distribution network.

2. The distribution network fault self-healing method according to claim 1, characterized in that: The sample risk level value of the distribution network is obtained by the following steps: Determining target risk scenario data from sample risk scenario data of the distribution network; Determining the sum of sample voltage over-limit risk values of all nodes in the distribution network and the sum of sample overload risk values of all lines based on the target risk scenario data; A sample risk level value of the distribution network is determined based on the sum of the sample voltage over-limit risk values and the sum of the sample overload risk values.

3. The distribution network fault self-healing method according to claim 2, characterized in that: Determining target risk scenario data from sample risk scenario data of the distribution network includes: A K-medoids algorithm is used to determine target risk scenario data from sample risk scenario data of the distribution network; wherein the K-medoids algorithm determines initial clustering points based on a density function of the sample risk scenario data.

4. The distribution network fault self-healing method according to claim 3, characterized in that: The method of using the K-medoids algorithm to determine target risk scenario data from sample risk scenario data of the distribution network includes: Obtain multiple sample risk scenario data for the distribution network; Calculating the density function value of all the sample risk scenario data; Determining, from a density function value set, sample risk scenario data corresponding to a plurality of density function values as initial clustering points; the density function value set includes a plurality of density function values whose density function values are greater than a set threshold; Repeat the following steps until the set stop condition is achieved: Assign each non-initial cluster point to the cluster to which the nearest initial cluster point belongs; Traversing each of the clusters, replacing the initial cluster point with the first non-initial cluster point of the cluster as the updated cluster point; wherein, if the total distance from all non-updated cluster points to the updated cluster point is smaller than the total distance from all non-initial cluster points to the initial cluster point, then the replacement is accepted, and the first non-initial cluster point is any non-initial cluster point of the cluster; Target risk scenario data is determined based on the updated cluster points.

5. The distribution network fault self-healing method according to claim 2, characterized in that: Determining a sample risk level value of the distribution network based on the sum of the sample voltage over-limit risk values and the sum of the sample overload risk values includes: The sample risk level value of the distribution network is determined based on the sum of the sample voltage over-limit risk values, the sum of the sample overload risk values, and the load-loss operation risk of the distribution network.

6. The distribution network fault self-healing method according to claim 1, characterized in that: The reward function is constructed based on the recovery time of the distribution network, the power supply reliability value and the risk level value of the distribution network.

7. The distribution network fault self-healing method according to claim 1, characterized in that: The method further comprises: constructing an objective function for minimizing the sum of the switching operation cost and the line risk level based on costs of multiple switching operations in the distribution network, a switching operation state corresponding to each of the switching operations, risk level values of multiple lines, and a line restoration state corresponding to each of the lines; Solving the objective function, and determining a grid reconstruction plan for the distribution network based on the solution of the objective function; A droop control method or a virtual synchronous machine control method is used to match the output data of the distributed power source with the power demand of the distribution network, so as to achieve the reconnection of the distributed power source to the grid.

8. A distribution network fault self-healing device, characterized in that: include: An acquisition module is used to obtain fault data of the distribution network; A self-healing module, configured to input the fault data into a fault recovery strategy model to obtain a fault recovery strategy output by the fault recovery strategy model; The fault recovery strategy model is trained based on sample fault data of the distribution network, the action space corresponding to the sample fault data, and the reward function corresponding to the sample fault data. The reward function is constructed based on the sample risk level value of the distribution network.

9. The distribution network fault self-healing device according to claim 8, characterized in that: The sample risk level value of the distribution network is obtained by the following steps: Determining target risk scenario data from sample risk scenario data of the distribution network; Determining the sum of sample voltage over-limit risk values of all nodes in the distribution network and the sum of sample overload risk values of all lines based on the target risk scenario data; A sample risk level value of the distribution network is determined based on the sum of the sample voltage over-limit risk values and the sum of the sample overload risk values.

10. The distribution network fault self-healing device according to claim 9, characterized in that: Determining target risk scenario data from sample risk scenario data of the distribution network includes: A K-medoids algorithm is used to determine target risk scenario data from sample risk scenario data of the distribution network; wherein the K-medoids algorithm determines initial clustering points based on a density function of the sample risk scenario data.

11. The distribution network fault self-healing device according to claim 10, characterized in that: The method of using the K-medoids algorithm to determine target risk scenario data from sample risk scenario data of the distribution network includes: Obtain multiple sample risk scenario data for the distribution network; Calculating the density function value of all the sample risk scenario data; Determining, from a density function value set, sample risk scenario data corresponding to a plurality of density function values as initial clustering points; the density function value set includes a plurality of density function values whose density function values are greater than a set threshold; Repeat the following steps until the set stop condition is achieved: Assign each non-initial cluster point to the cluster to which the nearest initial cluster point belongs; Traversing each of the clusters, replacing the initial cluster point with the first non-initial cluster point of the cluster as the updated cluster point; wherein, if the total distance from all non-updated cluster points to the updated cluster point is smaller than the total distance from all non-initial cluster points to the initial cluster point, then the replacement is accepted, and the first non-initial cluster point is any non-initial cluster point of the cluster; Target risk scenario data is determined based on the updated cluster points.

12. The distribution network fault self-healing device according to claim 9, characterized in that: Determining a sample risk level value of the distribution network based on the sum of the sample voltage over-limit risk values and the sum of the sample overload risk values includes: The sample risk level value of the distribution network is determined based on the sum of the sample voltage over-limit risk values, the sum of the sample overload risk values, and the load-loss operation risk of the distribution network.

13. The distribution network fault self-healing device according to claim 8, characterized in that: The reward function is constructed based on the recovery time of the distribution network, the power supply reliability value and the risk level value of the distribution network.

14. The distribution network fault self-healing device according to claim 8, characterized in that: The device further comprises: A construction module is configured to construct an objective function for minimizing the sum of the switching operation costs and the line risk level based on costs of a plurality of switching operations in the distribution network, a switching operation state corresponding to each of the switching operations, risk level values of a plurality of lines, and a line restoration state corresponding to each of the lines; A solution module, configured to solve the objective function and determine a grid reconstruction scheme for the distribution network based on a solution result of the objective function; The grid-connected module is used to use a droop control method or a virtual synchronous machine control method to match the distributed power output data with the power demand of the distribution network to achieve distributed power re-connection.

15. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the distribution network fault self-healing method according to any one of claims 1 to 7 is implemented.

16. A machine-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the distribution network fault self-healing method according to any one of claims 1 to 7 is implemented.

17. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the distribution network fault self-healing method according to any one of claims 1 to 7 is implemented.

Images