Power distribution network power supply reliability coupling analysis method for distributed source and load access

Abstract

The application relates to the technical field of power distribution network reliability analysis, and proposes a power distribution network power supply reliability coupling analysis method for distributed source and load access, which comprises the following steps: collecting current data of nodes of a feeder in a power distribution network, and establishing a current sequence of the nodes; constructing a current feature vector of the nodes, determining fluctuation severity coefficients and similar abnormality degrees of the nodes, and calculating source and load fluctuation indexes of the nodes; calculating current correlation coefficients and weighted implicit coupling strengths of different nodes, combining the source and load fluctuation indexes, and obtaining coupling fluctuation indexes of the different nodes; constructing a proposed distribution covariance matrix, using a Markov chain Monte Carlo method to obtain a posterior joint distribution sample set of all nodes at a collection time, and combining the coupling fluctuation indexes to realize power distribution network power supply reliability coupling analysis. The application can improve the precision of power distribution network power supply reliability coupling analysis.

Description

Technical Field

[0001] This application relates to the field of distribution network reliability analysis technology, specifically to a coupled analysis method for distribution network power supply reliability oriented towards distributed source-load access. Background Technology

[0002] Accurately quantifying the impact of distributed power sources and loads on the reliability of power distribution networks, identifying weak links in the grid, and optimizing planning and operation strategies are key to ensuring the safe and economical operation of the power grid and improving the capacity for renewable energy absorption. Currently, Monte Carlo methods, Markov chains, and artificial intelligence technologies are commonly integrated to conduct coupled analysis of power distribution network reliability, enabling grid reliability assessment under uncertainties in power sources and loads.

[0003] However, a high proportion of distributed power generation and load integration brings about strong spatiotemporal uncertainties. Distributed photovoltaic power generation is affected by micro-meteorological factors, resulting in drastic output fluctuations ranging from seconds to hours. The power generation and load at each node of the same feeder exhibit nonlinear, time-varying, and spatiotemporally correlated characteristics. Existing quantitative reliability assessment methods for distribution networks under source-load coupling assume that sources and loads are independent, ignoring the implicit coupling relationship between them caused by factors such as weather and electricity prices. This makes it difficult to accurately analyze complex and uncertain coupling characteristics, resulting in insufficient accuracy in reliability analysis and hindering the safe and efficient operation of distribution networks with high distributed power generation and load penetration. Summary of the Invention

[0004] This application provides a method for coupled analysis of power supply reliability in distribution networks with distributed source-load access, in order to solve the problem that the accuracy of coupled analysis of power supply reliability in distribution networks is insufficient due to the complex implicit coupling relationship between source and load. The specific technical solution adopted is as follows: One embodiment of this application provides a method for coupled analysis of power supply reliability in distribution networks with distributed source-load access. The method includes the following steps: Collect current data for each node of each feeder in the distribution network and establish the current sequence of the node at the time of collection. Based on the difference in current data of nodes at the acquisition time and within a second preset time before the acquisition time, a current feature vector of the node at the acquisition time is constructed. The current feature vectors of all nodes at the acquisition time are clustered. Based on the clustering results, the fluctuation severity coefficient and the degree of similar anomalies of the node at the acquisition time are determined. Combining the randomness of the current sequence of the node at the acquisition time, the source load fluctuation index of the node at the acquisition time is calculated. Based on the correlation of current data of different nodes at the acquisition time and the second preset time before the acquisition time, the current correlation coefficient of different nodes at the acquisition time is calculated. Based on the mutual information value of the current sequence of different nodes at the acquisition time and the degree of similar anomalies of the nodes at the acquisition time, the weighted implicit coupling strength of different nodes at the acquisition time is calculated. Combining the source-load fluctuation index and the current correlation coefficient, the coupling fluctuation index of different nodes at the acquisition time is obtained. Based on the current sequence, source-load fluctuation index, and coupling fluctuation index of different nodes at the acquisition time of all nodes, a proposed distribution covariance matrix is ​​constructed to obtain the posterior joint distribution sample set of all nodes at the acquisition time. Combined with the coupling fluctuation index, the power supply reliability coupling analysis of the distribution network is realized.

[0005] Furthermore, the method for constructing the current feature vector is as follows: Calculate the root mean square and peak-to-valley difference of all current data at the acquisition time and within the second preset time period before the acquisition time, and denot them as the root mean square and peak-to-valley difference of current characteristics at the acquisition time, respectively. Arrange the mean, variance, skewness, and kurtosis of the root mean square of current characteristics at all acquisition times and within the second preset time period before the acquisition time, as well as the mean, variance, skewness, and kurtosis of the peak-to-valley difference of current characteristics, in order to obtain the current feature vector at the acquisition time.

[0006] Furthermore, the method for determining the fluctuation severity coefficient is as follows: When the mean of all data in the current feature vector of the cluster center of the cluster is less than the first preset threshold and the variance of all data is less than the second preset threshold, the fluctuation severity coefficient of the node in the cluster at the corresponding collection time is assigned the first preset parameter value. When the variance of all data in the current feature vector of the cluster center of the cluster is greater than or equal to the second preset threshold, and the absolute value of the skewness of all data is less than the third preset threshold, the fluctuation severity coefficient of the node in the cluster at the corresponding acquisition time is assigned the second preset parameter value. When the absolute value of the skewness of all data in the current feature vector of the cluster center of the cluster is greater than or equal to the third preset threshold, and the kurtosis of all data is greater than the fourth preset threshold, the fluctuation severity coefficient of the node in the cluster at the corresponding acquisition time is assigned the third preset parameter value. The fluctuation severity coefficient of the nodes in the remaining clusters at the corresponding collection time is assigned the value of the fourth preset parameter. Among them, the first preset parameter value is less than the second preset parameter value, the second preset parameter value is less than the fourth preset parameter value, and the fourth preset parameter value is less than the third preset parameter value.

[0007] Furthermore, the degree of similar anomalies is: the normalized value of the Euclidean distance from the node to the cluster center within the corresponding cluster.

[0008] Furthermore, the method for calculating the source-load fluctuation index is as follows: Obtain the normalized permutation entropy value of the current sequence of the node at the acquisition time. The product of the sum of the same anomaly degree of the node at the acquisition time and the number 1, the fluctuation severity coefficient, and the normalized permutation entropy value of the current sequence of the node at the acquisition time is recorded as the source load fluctuation index of the node at the acquisition time.

[0009] Furthermore, the current correlation coefficient is: the Pearson correlation coefficient of all current data from two different nodes at the same acquisition time and within a second preset time period before the same acquisition time.

[0010] Furthermore, the method for calculating the weighted implicit coupling strength is as follows: Calculate the mutual information value of the current sequences of two different nodes at the same acquisition time; The difference between the number 1 and the degree of similar anomaly of the node at the time of data collection is recorded as the first difference of the node at the time of data collection. The product of the first differences of two different nodes at the same time of data collection is recorded as the confidence level of the two different nodes at the same time of data collection. The product of the confidence level of two different nodes at the same acquisition time and the mutual information value of the current sequence is denoted as the weighted implicit coupling strength of the two different nodes at the same acquisition time.

[0011] Furthermore, the method for obtaining the coupling fluctuation index is as follows: The arithmetic square root of the product of the source-load fluctuation indices of two different nodes at the acquisition time, and the product of the weighted implicit coupling strength and current correlation coefficient of two different nodes at the acquisition time, are denoted as the coupling fluctuation index of two different nodes at the acquisition time.

[0012] Furthermore, the proposed formula for calculating the elements in the distribution covariance matrix is ​​as follows: in, The first element in the covariance matrix of the proposal distribution of the Metropolis-Hastings algorithm represents the... Line 1 Column elements; and They are nodes and nodes The standard deviation of all current data at the time of acquisition and within a first preset time before the time of acquisition; For nodes and nodes The coupling fluctuation index at the time of acquisition; For the Kroneckerdelta function, when The value is 1 if the condition is met, and 0 otherwise. It is a preset minimum positive number.

[0013] Furthermore, the specific methods for achieving coupled analysis of power supply reliability in the distribution network by combining the coupling fluctuation index are as follows: For each state combination in the posterior joint distribution sample set, the sum of the product of the normalized sample value corresponding to the node and the preset current range and the preset current minimum value is recorded as the predicted current of the node at the corresponding acquisition time. When the predicted current of the node at the corresponding acquisition time is greater than 1.2 times the rated current of the node, it is determined that the current limit of the corresponding state combination has been exceeded. The probability of the current limit exceeding of all state combinations in the posterior joint distribution sample set is recorded as the probability of the current limit exceeding of the node at the corresponding acquisition time. The two nodes corresponding to the maximum values ​​of the coupling fluctuation index of all different nodes at the acquisition time are denoted as the key node pair at the acquisition time. Based on the key node pairs at the time of data acquisition and the current over-limit probability of all nodes at the time of data acquisition, a power distribution network reliability coupling analysis report is output.

[0014] The beneficial effects of this application are: This application first considers the characteristics of distributed photovoltaic power generation efficiency, which is susceptible to changes in cloud movement, aerosol concentration, and random electricity consumption behavior. Based on the quantification results of the randomness and volatility of node currents, and the degree of fluctuation anomalies of a node relative to other nodes in the same cluster, it evaluates the likelihood that a node is a high-risk source of fluctuation in the distribution network and obtains the source-load fluctuation index of the node at the time of data collection. Then, considering that after distributed source-loads are connected to the distribution network, the time-series changes of currents at different nodes on the same feeder generally exhibit time misalignment, this application quantifies the similarity of the current fluctuation waveforms of different nodes after time axis constraints, and obtains the source-load fluctuation index of different nodes at the time of data collection. The current correlation coefficient is calculated, and the weighted implicit coupling strength of different nodes at the acquisition time is calculated to reduce the pseudo-coupling interference caused by isolated anomalies. The possibility that different nodes are strong coupling node pairs that need to be focused on in the distribution network is analyzed, and the coupling fluctuation index of different nodes at the acquisition time is obtained. Based on the implicit coupling relationship between source and load and the current fluctuation, a proposed distribution covariance matrix is ​​constructed, and the posterior joint distribution sample set of all nodes at the acquisition time is obtained. Combined with the coupling fluctuation index, the coupling analysis of power supply reliability of the distribution network is realized, which solves the problem that the complexity of the implicit coupling relationship between source and load leads to insufficient accuracy of the coupling analysis of power supply reliability of the distribution network, and improves the accuracy of the coupling analysis of power supply reliability of the distribution network. Attached Figure Description

[0015] Figure 1 This is a schematic flowchart of a power supply reliability coupling analysis method for distributed source-load access in a distribution network, provided in one embodiment of this application. Detailed Implementation

[0016] Please see Figure 1The diagram illustrates a flowchart of a power supply reliability coupling analysis method for distributed source-load access in a distribution network according to an embodiment of this application. The method includes the following steps: Step S001: Collect current data of each node of each feeder in the distribution network and establish the current sequence of the node at the time of collection.

[0017] Each feeder in the distribution network is treated as a basic analysis unit. The grid connection points of all distributed power sources, energy storage devices, and conventional loads on the feeder are defined as nodes, and current data of each node is collected.

[0018] Specifically, taking the power distribution network of an urban commercial park as an example, the node division rules for the power distribution network of an urban commercial park are as follows: 1. Each distributed power source's grid connection point is treated as an independent node. The distributed power source is such as a photovoltaic inverter or a wind turbine converter. In this embodiment, there are 8 distributed power sources, specifically 6 photovoltaic inverters and 2 wind turbine converters. The data acquisition module built into the inverter of the distributed power source is used to collect current data.

[0019] 2. Treat the common connection point of each distribution network power supply energy storage system as an independent node. In this embodiment, there are two common connection points of the distribution network power supply energy storage system. Use the existing distribution automation terminal or the new intelligent fusion terminal of the distribution network power supply energy storage system to collect current data.

[0020] 3. A node is set at the end of each feeder and at the sectionalizing switch in the distribution network. In this embodiment, there are a total of 8 feeder end nodes and 8 sectionalizing switch nodes. Open-type current transformers are installed at the nodes at the end of the feeders and at the sectionalizing switches to collect current data.

[0021] Each node is assigned a unique number, and the total number of nodes is counted.

[0022] In this embodiment, the current data acquisition frequency of the node is set to 1Hz, and the current data is normalized. This embodiment adopts the maximum-minimum value normalization method. Based on all current data of the node at the acquisition time and in the 24 hours before the acquisition time, the current data of each node is normalized separately. In actual application, implementers can use other methods such as the Z-Score standard normalization method for normalization, which is not limited here.

[0023] Arrange all current data of the node at the acquisition time and within a first preset time before the acquisition time in sequence to obtain the current sequence of the node at the acquisition time.

[0024] In this embodiment, the first preset time is set to 15 minutes.

[0025] At this point, the current sequence of each node at each acquisition time has been obtained.

[0026] Step S002: Based on the difference in current data of the node at the acquisition time and within a second preset time before the acquisition time, construct the current feature vector of the node at the acquisition time, cluster the current feature vectors of all nodes at the acquisition time, determine the fluctuation severity coefficient and the degree of similar anomalies of the node at the acquisition time based on the clustering results, and calculate the source load fluctuation index of the node at the acquisition time by combining the randomness of the current sequence of the node at the acquisition time.

[0027] Distributed photovoltaic power generation efficiency is easily affected by cloud movement and sudden changes in aerosol concentration, resulting in non-stationary multimodal fluctuations in output ranging from seconds to minutes. Wind power generation is prone to intermittent turbine shutdowns and power surges caused by gusts in low wind speed ranges. Meanwhile, the load power of the distribution network is significantly affected by users' random electricity consumption behavior, further causing the current distribution at each power supply node of the distribution network to exhibit strong nonlinear and non-Gaussian distribution characteristics.

[0028] The normalized permutation entropy value of the current sequence of the node at the acquisition time is obtained using the permutation entropy analysis algorithm.

[0029] In this embodiment, the embedding dimension is set to 6 and the time delay is 1 when using the permutation entropy analysis algorithm.

[0030] Normalized permutation entropy is used to quantify the randomness of the current at a node at the time of data acquisition. The larger the normalized permutation entropy, the stronger the randomness of the current at the time of data acquisition, and the more significant the uncertainty and complexity of the distributed source-load. The smaller the normalized permutation entropy, the more significant the periodic or quasi-periodic characteristics of the current at the time of data acquisition.

[0031] Permutation entropy can effectively characterize the degree of current randomness at distribution network nodes, but it cannot depict the similarity of fluctuations and spatial clustering characteristics between different nodes. Multiple photovoltaic nodes on the same feeder in a distribution network are often affected by similar cloud movements, exhibiting similar shading variation patterns; user electricity consumption behavior is affected by the synchronicity of commuting schedules, easily forming regional load spikes. If node clusters with the same fluctuation patterns cannot be accurately identified, the curse of dimensionality can easily occur in subsequent source-load coupling analysis.

[0032] For any given acquisition time, calculate the root mean square (RMS) and peak-to-valley difference (PKD) of all current data within the acquisition time and the second preset time period prior to the acquisition time. These are denoted as the current characteristic RMS and current characteristic PKD at the acquisition time, respectively. Arrange the mean, variance, skewness, and kurtosis of the current characteristic RMS and the current characteristic PKD at all acquisition times within the second preset time period prior to the acquisition time in sequence to obtain the current characteristic vector at the acquisition time. Cluster the current characteristic vectors of all nodes at the same acquisition time to obtain four clusters. Calculate the current characteristic vector of the cluster center for each cluster.

[0033] When the mean of all data in the current feature vector of the cluster center of a cluster is less than a first preset threshold and the variance of all data is less than a second preset threshold, the cluster is designated as a stationary cluster, and the fluctuation severity coefficient of the nodes in the stationary cluster at the corresponding acquisition time is assigned the first preset parameter value. When the variance of all data in the current feature vector of the cluster center of a cluster is greater than or equal to the second preset threshold and the absolute value of the skewness of all data is less than a third preset threshold, the cluster is designated as a random fluctuation cluster, and the fluctuation severity coefficient of the nodes in the random fluctuation cluster at the corresponding acquisition time is assigned the second preset parameter value. When the absolute value of the skewness of all data in the current feature vector of the cluster center of a cluster is greater than or equal to the third preset threshold and the kurtosis of all data is greater than a fourth preset threshold, the cluster is designated as a spike impact cluster, and the fluctuation severity coefficient of the nodes in the spike impact cluster at the corresponding acquisition time is assigned the third preset parameter value. The remaining clusters are designated as pulsating clusters, and the fluctuation severity coefficient of the nodes in the pulsating cluster at the corresponding acquisition time is assigned the fourth preset parameter value.

[0034] In this embodiment, the values ​​of the first preset threshold, the second preset threshold, the third preset threshold, and the fourth preset threshold are 0.2, 0.1, 0.5, and 5, respectively.

[0035] The values ​​of the first preset parameter value, the second preset parameter value, the third preset parameter value, and the fourth preset parameter value should simultaneously satisfy the following: the first preset parameter value is less than the second preset parameter value, the second preset parameter value is less than the fourth preset parameter value, and the fourth preset parameter value is less than the third preset parameter value. In this embodiment, the values ​​of the first preset parameter value, the second preset parameter value, the third preset parameter value, and the fourth preset parameter value are 0.1, 0.4, 1.0, and 0.7, respectively.

[0036] Wherein, the second preset time is less than the first preset time, and in this embodiment, the value of the second preset time is 1 minute; this embodiment uses the K-means clustering algorithm to implement clustering, and clustering is a well-known technology, so it will not be described in detail here.

[0037] The four clusters are classified into a stationary cluster, a random fluctuation cluster, a spike impact cluster, and a pulsating cluster, based on the intensity of current fluctuations in the nodes within each cluster at the corresponding acquisition time. The pulsating cluster corresponds to current fluctuations caused by factors such as gusts of wind.

[0038] The normalized value of the Euclidean distance from a node to the cluster center within its corresponding cluster is denoted as the degree of anomaly of the node at the corresponding data collection time.

[0039] It should be noted that this embodiment uses the maximum-minimum normalization method to calculate the normalized value. The normalized value is the ratio of the Euclidean distance from a node to the cluster center within the corresponding cluster to the maximum value of the Euclidean distances from all nodes within the cluster to the cluster center. In practical applications, implementers may use other methods of existing technology, such as the Z-Score standard normalization method or the sigmoid function, to calculate the normalized value, which is not limited here.

[0040] The degree of similar anomalies of a node at the time of data collection is used to evaluate the degree of fluctuation anomalies of a node relative to other nodes in the same cluster.

[0041] The source-load fluctuation index of the node at the time of acquisition is denoted as the product of the sum of the same anomaly degree of the node at the time of acquisition and the number 1, the fluctuation severity coefficient, and the normalized permutation entropy value of the current sequence of the node at the time of acquisition.

[0042] The source-load fluctuation index of a node at the time of data collection is used to evaluate the likelihood that the node is a high-risk fluctuation source in the distribution network.

[0043] At this point, the source load fluctuation index of the node at the time of data acquisition is obtained.

[0044] Step S003: Based on the correlation of current data of different nodes at the acquisition time and the second preset time before the acquisition time, calculate the current correlation coefficient of different nodes at the acquisition time. Based on the mutual information value of the current sequence of different nodes at the acquisition time and the degree of similar anomalies of the nodes at the acquisition time, calculate the weighted implicit coupling strength of different nodes at the acquisition time. Combine the source-load fluctuation index and the current correlation coefficient to obtain the coupling fluctuation index of different nodes at the acquisition time.

[0045] After distributed power sources and loads are connected to the distribution network, the time-series changes in current at different nodes on the same feeder generally exhibit a time misalignment phenomenon. For example, when clouds move from upstream to downstream along the feeder, the downstream photovoltaic output will experience a power drop due to a delay of tens of seconds relative to the upstream photovoltaic output. At the same time, the power source and load form an implicit coupling relationship through unobserved variables such as electricity price and ambient temperature. Relying solely on the power source-load fluctuation index cannot effectively quantify the intensity and direction of dynamic interactions between nodes.

[0046] Calculate the Pearson correlation coefficient of all current data of two different nodes at the same acquisition time and within a second preset time before the same acquisition time, and denot it as the current correlation coefficient of the two different nodes at the same acquisition time.

[0047] The calculation of the Pearson correlation coefficient is a well-known technique and will not be elaborated further.

[0048] The larger the absolute value of the current correlation coefficient between two different nodes at the acquisition time, the more similar the waveforms of the current fluctuations at the two different nodes will be after time axis constraints.

[0049] Furthermore, this analysis examines whether the aforementioned similarity evaluation results originate from shared environmental driving factors or from genuine implicit coupling between the source and load. Environmental driving factors include photovoltaics, which are simultaneously affected by irradiance, while implicit coupling includes the inverse correlation between electric vehicle and photovoltaic power output under electricity price incentives.

[0050] Using the mutual information algorithm, the edge distribution is estimated using Gaussian kernel density, and the mutual information value of the current sequence of two different nodes at the same acquisition time is calculated. The difference between the number 1 and the degree of similar anomaly of the node at the acquisition time is recorded as the first difference of the node at the acquisition time. The product of the first differences of the two different nodes at the same acquisition time is recorded as the confidence of the two different nodes at the same acquisition time. The product of the confidence of the two different nodes at the same acquisition time and the mutual information value of the current sequence is recorded as the weighted implicit coupling strength of the two different nodes at the same acquisition time.

[0051] Mutual information value is used to evaluate the amount of mutual information between current data of two different nodes. The larger the mutual information value, the stronger the direct dependency between the current data of the two different nodes after excluding common influencing factors.

[0052] The greater the degree of similar anomalies at a node during the data acquisition time, the higher the probability that the current at that node belongs to local abnormal noise. Therefore, the confidence level is calculated based on the first difference. The weighted implicit coupling strength is used to reduce pseudo-coupling interference caused by isolated anomalies.

[0053] The arithmetic square root of the product of the source-load fluctuation indices of two different nodes at the acquisition time, and the product of the weighted implicit coupling strength and current correlation coefficient of two different nodes at the acquisition time, are denoted as the coupling fluctuation index of two different nodes at the acquisition time.

[0054] Based on prior knowledge, when distributed source-load access is connected to the distribution network, two different nodes are more likely to be strongly coupled node pairs that require more attention in the distribution network only when the current data of the two different nodes have stronger fluctuation characteristics, there is still more significant direct information coupling between the two different nodes, and the current waveforms of the two different nodes have higher similarity.

[0055] At this point, the coupling fluctuation index of all different nodes at the time of acquisition is obtained.

[0056] Step S004: Based on the current sequence, source-load fluctuation index, and coupling fluctuation index of different nodes at the acquisition time of all nodes, construct the proposed distribution covariance matrix, obtain the posterior joint distribution sample set of all nodes at the acquisition time, and combine it with the coupling fluctuation index to realize the coupling analysis of power supply reliability of the distribution network.

[0057] Furthermore, a coupled fluctuation index is introduced, and the Markov Chain Monte Carlo (MCMC) method is used to perform coupled analysis on the power supply reliability of the distribution network.

[0058] Specifically, the Markov chain Monte Carlo method takes the current sequence and source-load fluctuation index of all nodes at the same acquisition time, as well as the coupling fluctuation index of all different nodes at the same acquisition time, as inputs, and outputs a posterior joint distribution sample set of the node's current state vector. Based on this, the power supply reliability of the distribution network is calculated, realizing a quantitative assessment of the power supply reliability of the power grid in the scenario of distributed source-load access. The current state vector of the node is composed of all current data of the node at the acquisition time and within a first preset time period before the acquisition time.

[0059] Specifically, this application uses the current sequences of all nodes as observation data and the source-load fluctuation index of the nodes and the coupling fluctuation index of different nodes as prior constraint information. Based on Bayes' theorem, it integrates the likelihood function and the prior distribution to complete the normalization process and construct the joint posterior distribution of the current state vector of the nodes. At the same time, it relies on the Metropolis-Hastings algorithm to complete sampling and implicitly captures the nonlinear correlation characteristics between nodes through the state transition process, effectively overcoming the shortcomings of traditional Monte Carlo simulation that ignores the spatial correlation of nodes and is difficult to embed physical constraints.

[0060] However, the existing MCMC algorithm has obvious defects: when there are strong coupling fluctuation characteristics between distribution network nodes, the node independent proposal distribution adopted by the algorithm cannot adapt to the joint fluctuation law of nodes, resulting in slow convergence speed of Markov chain in strong coupling dimension, requiring a huge number of iterations to converge to a stationary distribution, and extremely low computational efficiency.

[0061] To address the aforementioned issues, an adaptive optimization improvement is performed on the proposal distribution covariance matrix of the Metropolis-Hastings algorithm based on the obtained coupled fluctuation index.

[0062] First, arrange all current data of the node at the acquisition time and within a first preset time before the acquisition time in sequence to obtain the current state vector of the node at the acquisition time.

[0063] Based on all current data from all nodes at the acquisition time and within a first preset time prior to the acquisition time, as well as the coupling fluctuation index of different nodes at the acquisition time, the proposed distribution covariance matrix of the Metropolis-Hastings algorithm is constructed. Specifically; in, When analyzing the power supply reliability of the distribution network at the time of data acquisition, the proposal distribution covariance matrix of the Metropolis-Hastings algorithm represents the [missing information - likely a specific value or term]. Line 1 Column elements; and They are nodes and nodes The standard deviation of all current data at the time of acquisition and within a first preset time before the time of acquisition; For nodes and nodes The coupling fluctuation index at the time of acquisition; For the Kroneckerdelta function, when The value is 1 if the condition is met, and 0 otherwise. The value of the minimum positive number is a preset value, and the minimum positive number should be greater than or equal to 1. and less than or equal to In this embodiment, the value of the smallest positive number is... This ensures that all elements in the covariance matrix are always positive.

[0064] For the off-diagonal elements in the proposed distribution covariance matrix, i.e. The corresponding Kroneckerdelta function is set to 0. The off-diagonal covariance is linearly controlled by the coupling fluctuation index of different nodes at the acquisition time. When the coupling fluctuation index of different nodes at the acquisition time is greater than 0, the corresponding covariance is increased, promoting joint state shift in the same direction. When the coupling fluctuation index of different nodes at the acquisition time is less than 0, the corresponding covariance is decreased, even making the covariance negative, guiding the node state to shift in the opposite direction. This calculation method allows the proposal distribution covariance matrix to adaptively represent the true coupling strength and direction between nodes, accelerating state space exploration efficiency in strongly coupled scenarios, improving sampling performance and chain convergence speed. Furthermore, the stronger the coupling between nodes and the larger the corresponding coupling fluctuation index, the larger the covariance will be configured in the corresponding dimension of the proposal distribution.

[0065] For the diagonal elements in the proposed distribution covariance matrix, i.e. ,at this time, The corresponding value is 0, with the standard deviation of the node's own historical current data used as the benchmark value for the diagonal elements of the covariance matrix.

[0066] The eigenvalues ​​of the proposed distribution covariance matrix are solved. When there are eigenvalues ​​less than 0, the smallest constant that makes all the eigenvalues ​​of the proposed distribution covariance matrix greater than or equal to zero is added to all diagonal elements of the proposed distribution covariance matrix to ensure that the proposed distribution covariance matrix satisfies the positive semidefinite condition, so that the Markov chain sampling can be executed smoothly.

[0067] Using the Markov chain Monte Carlo method and the Metropolis-Hastings algorithm, the current sequence of all nodes at the same acquisition time, the source-load fluctuation index, and the coupling fluctuation index of all different nodes at the same acquisition time are used as inputs to construct the proposed distribution covariance matrix of the Metropolis-Hastings algorithm, and obtain the posterior joint distribution sample set of the current state vector of all nodes at the same acquisition time.

[0068] The posterior joint distribution sample set is a large number of node current state combinations that conform to the coupling law, physical constraints, and source-load fluctuation correlation of the real power grid.

[0069] It is important to understand that for each state combination in the posterior joint distribution sample set, there is a normalized sample value corresponding to each node.

[0070] Therefore, for each state combination in the posterior joint distribution sample set, the sum of the product of the normalized sample value corresponding to the node and the preset current range and the preset current minimum value is recorded as the predicted current of the node at the corresponding acquisition time; when the predicted current of the node at the corresponding acquisition time is greater than 1.2 times the rated current of the node, it is determined that the current limit of the corresponding state combination has been exceeded; the probability of the current limit exceeding of all state combinations in the posterior joint distribution sample set is recorded as the current limit exceeding probability of the node at the corresponding acquisition time.

[0071] The preset current range is the difference between the maximum and minimum values ​​of the current data collected at the node location during the acquisition time and within a first preset time period prior to the acquisition time. The preset minimum current value is the minimum value of the current data collected at the node location during the acquisition time and within a first preset time period prior to the acquisition time.

[0072] The two nodes corresponding to the maximum values ​​of the coupling fluctuation index of all different nodes at the acquisition time are denoted as the critical node pair at the acquisition time. The critical node pair is the node pair with the strongest coupling and the largest influence range at the acquisition time.

[0073] Based on the key node pairs at the time of data acquisition and the current over-limit probability of all nodes at the time of data acquisition, a power distribution network reliability coupling analysis report is output.

[0074] This completes the coupled analysis of power supply reliability in the distribution network.

Claims

1. A coupled analysis method for power supply reliability in distribution networks with distributed source-load access, characterized in that, The method includes the following steps: Collect current data for each node of each feeder in the distribution network and establish the current sequence of the node at the time of collection. Based on the difference in current data of nodes at the acquisition time and within a second preset time before the acquisition time, a current feature vector of the node at the acquisition time is constructed. The current feature vectors of all nodes at the acquisition time are clustered. Based on the clustering results, the fluctuation severity coefficient and the degree of similar anomalies of the node at the acquisition time are determined. Combining the randomness of the current sequence of the node at the acquisition time, the source load fluctuation index of the node at the acquisition time is calculated. Based on the correlation of current data of different nodes at the acquisition time and the second preset time before the acquisition time, the current correlation coefficient of different nodes at the acquisition time is calculated. Based on the mutual information value of the current sequence of different nodes at the acquisition time and the degree of similar anomalies of the nodes at the acquisition time, the weighted implicit coupling strength of different nodes at the acquisition time is calculated. Combining the source-load fluctuation index and the current correlation coefficient, the coupling fluctuation index of different nodes at the acquisition time is obtained. Based on the current sequence, source-load fluctuation index, and coupling fluctuation index of different nodes at the acquisition time of all nodes, a proposed distribution covariance matrix is ​​constructed to obtain the posterior joint distribution sample set of all nodes at the acquisition time. Combined with the coupling fluctuation index, the power supply reliability coupling analysis of the distribution network is realized.

2. The method for coupled analysis of power supply reliability in distribution networks oriented towards distributed source-load access as described in claim 1, characterized in that, The method for constructing the current feature vector is as follows: Calculate the root mean square and peak-to-valley difference of all current data at the acquisition time and within the second preset time period before the acquisition time, and denot them as the root mean square and peak-to-valley difference of current characteristics at the acquisition time, respectively. Arrange the mean, variance, skewness, and kurtosis of the root mean square of current characteristics at all acquisition times and within the second preset time period before the acquisition time, as well as the mean, variance, skewness, and kurtosis of the peak-to-valley difference of current characteristics, in order to obtain the current feature vector at the acquisition time.

3. The method for coupled analysis of power supply reliability in distribution networks oriented towards distributed source-load access as described in claim 1, characterized in that, The method for determining the severity coefficient of fluctuation is as follows: When the mean of all data in the current feature vector of the cluster center of the cluster is less than the first preset threshold and the variance of all data is less than the second preset threshold, the fluctuation severity coefficient of the node in the cluster at the corresponding collection time is assigned the first preset parameter value. When the variance of all data in the current feature vector of the cluster center of the cluster is greater than or equal to the second preset threshold, and the absolute value of the skewness of all data is less than the third preset threshold, the fluctuation severity coefficient of the node in the cluster at the corresponding acquisition time is assigned the second preset parameter value. When the absolute value of the skewness of all data in the current feature vector of the cluster center of the cluster is greater than or equal to the third preset threshold, and the kurtosis of all data is greater than the fourth preset threshold, the fluctuation severity coefficient of the node in the cluster at the corresponding acquisition time is assigned the third preset parameter value. The fluctuation severity coefficient of the nodes in the remaining clusters at the corresponding collection time is assigned the value of the fourth preset parameter. Among them, the first preset parameter value is less than the second preset parameter value, the second preset parameter value is less than the fourth preset parameter value, and the fourth preset parameter value is less than the third preset parameter value.

4. The method for coupled analysis of power supply reliability in distribution networks oriented towards distributed source-load access as described in claim 1, characterized in that, The degree of similar anomaly is the normalized value of the Euclidean distance from the node to the cluster center within the corresponding cluster.

5. The method for coupled analysis of power supply reliability in distribution networks oriented towards distributed source-load access as described in claim 1, characterized in that, The calculation method for the source load fluctuation index is as follows: Obtain the normalized permutation entropy value of the current sequence of the node at the acquisition time. The product of the sum of the same anomaly degree of the node at the acquisition time and the number 1, the fluctuation severity coefficient, and the normalized permutation entropy value of the current sequence of the node at the acquisition time is recorded as the source load fluctuation index of the node at the acquisition time.

6. The method for coupled analysis of power supply reliability in distribution networks oriented towards distributed source-load access as described in claim 1, characterized in that, The current correlation coefficient is the Pearson correlation coefficient of all current data from two different nodes at the same acquisition time and within a second preset time period before the same acquisition time.

7. The method for coupled analysis of power supply reliability in distribution networks oriented towards distributed source-load access as described in claim 1, characterized in that, The method for calculating the weighted implicit coupling strength is as follows: Calculate the mutual information value of the current sequences of two different nodes at the same acquisition time; The difference between the number 1 and the degree of similar anomaly of the node at the time of data collection is recorded as the first difference of the node at the time of data collection. The product of the first differences of two different nodes at the same time of data collection is recorded as the confidence level of the two different nodes at the same time of data collection. The product of the confidence level of two different nodes at the same acquisition time and the mutual information value of the current sequence is denoted as the weighted implicit coupling strength of the two different nodes at the same acquisition time.

8. The method for coupled analysis of power supply reliability in distribution networks oriented towards distributed source-load access as described in claim 1, characterized in that, The method for obtaining the coupling fluctuation index is as follows: The arithmetic square root of the product of the source-load fluctuation indices of two different nodes at the acquisition time, and the product of the weighted implicit coupling strength and current correlation coefficient of two different nodes at the acquisition time, are denoted as the coupling fluctuation index of two different nodes at the acquisition time.

9. The method for coupled analysis of power supply reliability in distribution networks oriented towards distributed source-load access as described in claim 1, characterized in that, The proposed formula for calculating the elements in the distribution covariance matrix is: in, The first element in the covariance matrix of the proposal distribution of the Metropolis-Hastings algorithm represents the... Line number Column elements; and They are nodes and nodes The standard deviation of all current data at the time of acquisition and within a first preset time before the time of acquisition; For nodes and nodes The coupling fluctuation index at the time of acquisition; For the Kroneckerdelta function, when The value is 1 if it is true, and 0 otherwise. It is a preset minimum positive number.

10. The method for coupled analysis of power supply reliability in distribution networks oriented towards distributed source-load access as described in claim 1, characterized in that, The specific methods for achieving coupled analysis of power supply reliability in the distribution network by combining the coupling fluctuation index are as follows: For each state combination in the posterior joint distribution sample set, the sum of the product of the normalized sample value corresponding to the node and the preset current range and the preset current minimum value is recorded as the predicted current of the node at the corresponding acquisition time. When the predicted current of the node at the corresponding acquisition time is greater than 1.2 times the rated current of the node, it is determined that the current limit of the corresponding state combination has been exceeded. The probability of the current limit exceeding of all state combinations in the posterior joint distribution sample set is recorded as the probability of the current limit exceeding of the node at the corresponding acquisition time. The two nodes corresponding to the maximum values ​​of the coupling fluctuation index of all different nodes at the acquisition time are denoted as the key node pair at the acquisition time. Based on the key node pairs at the time of data acquisition and the current over-limit probability of all nodes at the time of data acquisition, a power distribution network reliability coupling analysis report is output.

Images