Power distribution network fault bidirectional research method combining correlation matrix and dynamic bayes

Abstract

This invention discloses a bidirectional fault assessment method for distribution networks combining correlation matrices and dynamic Bayesian methods. The method includes the following steps: constructing a correlation matrix based on the actual distribution network topology, and combining this with a fault information matrix to construct a fault assessment matrix; establishing fault criteria for both missed and false fault reports, thereby achieving top-down fault feeder identification; constructing a dynamic Bayesian network based on the actual distribution network topology and the "station-line-transformer-customer" relationship to infer the fault probability of each region; constructing membership functions and fuzzy inference rules to obtain a fuzzy inference system for adjusting fault probabilities; and finally inferring the most likely fault region, achieving bottom-up fault region assessment. Based on traditional matrix algorithms, this invention improves fault criteria by constructing a causal correlation matrix, effectively solving the feeder identification problem under conditions of missed and false fault reports, reducing computational load, and improving the accuracy of fault location.

Description

Technical Field

[0001] This invention relates to the field of distribution network fault detection, and in particular to a two-way fault assessment method for distribution networks that combines correlation matrix and dynamic Bayesian methods. Background Technology

[0002] As a crucial link connecting users and the power system, the safe operation of the distribution network is essential for the reliability of users' electricity supply. The topology of low-voltage distribution networks is complex and variable, making it difficult to obtain accurate topological information. Even when a fault signal is received, it is challenging to accurately locate the outage area. Furthermore, the massive amount of data generated by distribution secondary terminals, coupled with the complex operating environment and the potential for information loss during transmission, further complicates fault location assessment.

[0003] There is already a large body of literature on fault area assessment both domestically and internationally. Current feeder fault identification mostly adopts matrix calculation, which simplifies the calculation process to some extent, but the overall algorithm still has a certain degree of complexity, especially when dealing with large-scale power distribution networks, and the calculation cost is still relatively high. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a simple algorithm that is less affected by missed or false fault reports, combining correlation matrix and dynamic Bayesian methods for bidirectional fault assessment in distribution networks.

[0005] The technical solution of this invention to solve the above-mentioned technical problems is: a two-way fault assessment method for distribution networks combining correlation matrix and dynamic Bayesian methods, comprising the following steps:

[0006] Step 1: Construct an association matrix based on the actual topology of the distribution network, and combine it with the fault information matrix to build a fault assessment matrix;

[0007] Step 2: Establish fault criteria for both missed and false alarms of node fault information, thereby achieving top-down fault feeder identification;

[0008] Step 3: Construct a dynamic Bayesian network based on the actual topology of the distribution network and the "station-line-transformer-customer" relationship, and infer the fault probability of each area;

[0009] Step 4: Construct membership functions and fuzzy inference rules to obtain a fuzzy inference system for adjusting fault probabilities;

[0010] Step 5: Finally, deduce the most likely fault area to achieve bottom-up fault area assessment.

[0011] The above-mentioned method for bidirectional fault assessment of distribution networks combining correlation matrix and dynamic Bayesian methods, specifically step one, involves the following steps:

[0012] (1-1) When a fault occurs on a feeder in the distribution network, the fault indicator connected to the first end of the fault feeder will monitor whether a fault current exceeding the set value flows through the fault indicator and upload the monitored signal to the cloud platform. The fault indicator installed at the end of the fault feeder will not send a signal. The cloud platform generates a row vector matrix with the same number of columns as D′ as the fault information matrix G based on the uploaded information, where D′ is the causal correlation matrix between the feeder section and the switch node.

[0013] (1-2) The fault assessment matrix Q is obtained by multiplying the inverse matrices of G and D′ as follows:

[0014]

[0015] In the formula, when the element in the p-th row of Q is 1, it indicates that the p-th feeder has a fault; when the element in the p-th row of G is 0, it indicates that the p-th feeder has no fault. When the element in the q-th row of G is 1, it indicates that there is a fault current exceeding the setting value flowing through the fault indicator at the q-th node; when the element in the q-th row of G is 0, it indicates that there is no fault current exceeding the setting value flowing through the fault indicator at the q-th node. The element d in the p-th row and q-th column of D′... pq A value of 1 indicates that a fault in the p-th feeder affected the q-th fault indicator, d pq A value of 0 indicates that the p-th feeder is not associated with the q-th fault indicator.

[0016] The above-mentioned bidirectional fault assessment method for distribution networks combining correlation matrix and dynamic Bayesian methods, in step two, when node fault information is missed, it is manifested as the fault information uploaded by the feeder terminal unit (FTU) at that node being empty. When processing missed information, the following two assessment rules are constructed:

[0017] Rule 1: When the fault information of the nodes before and after a node is the same, the missed fault information is set to the same fault information as the nodes before and after it.

[0018] Rule 2: When the fault information of a node is different from that of its preceding and following adjacent nodes, the node is removed as an invalid node and is not considered in either D or G. The analysis is then performed again.

[0019] The above-mentioned bidirectional fault assessment method for distribution networks combining correlation matrix and dynamic Bayesian methods, in step two, when a fault information is falsely reported, is manifested by a reversal of the fault information uploaded by the FTU at that node, i.e., the actual fault information is 1 but is reported as 0, or the actual fault information is 0 but is reported as 1. The following two assessment rules are constructed:

[0020] Rule 3: After uploading false fault information, if the analysis results show that multiple sections of the same line have faults, the error is corrected by comparing the adjacent information of the nodes. If the fault information of a certain node is different from the fault information uploaded by the adjacent nodes, then the fault information of that node has been falsely reported. The falsely reported information is modified to be the same as the information of the adjacent nodes. If the value of an element in the analysis matrix is ​​-1, then the value is set to 0.

[0021] Rule 4: If the analysis result is incorrect after uploading false alarm information, it is necessary to collect the fault information again.

[0022] The above-mentioned method for bidirectional fault assessment of distribution networks combining correlation matrix and dynamic Bayesian methods, specifically step three, involves the following steps:

[0023] (3-1) Integrate the fault data from all sides of the distribution network, including the station, line, transformer and customer, perform data cleaning and integration, construct a complete dataset containing fault information, and divide it into training set and test set;

[0024] (3-2) Based on the power outage events of users in the transformer area, and combined with the dependency relationship and time sequence characteristics of the fault data, a dynamic Bayesian network DBN is constructed. The DBN includes six key nodes: the proportion of power outages of users in the transformer area, the preliminary judgment of event types, the complaint telephone numbers of users in the transformer area, the event information of low-voltage branch terminal, the information of transformer terminal and the final judgment event.

[0025] (3-3) Learn the node parameters of DBN by maximum likelihood estimation and calculate the probability of failure in different regions of the test set.

[0026] The above-mentioned bidirectional fault assessment method for distribution networks, combining correlation matrix and dynamic Bayesian methods, is applicable to the i-th node x in a distribution Bayesian network. i x i The data contains m power distribution events. Assume that all events related to node x have been obtained. i The distribution events of the relevant nodes are E′={e1,e2,···,e n}, then x i The kth power distribution event e k The conditional probabilities of occurrence are as follows:

[0027]

[0028] T represents the number of time slices. Represents the power distribution event e in the t-th time slice. k Par represents the probability of a power distribution event occurring, P B For the transition probability;

[0029] When a fault occurs in the distribution network, leading to a power outage, a fault event chain can be represented as F(r) = {t}.i│ x i,1 ,xi , 2,…x i,j …x i,n}, t i x represents i Fault records, x i,n x represents i The corresponding nth distribution event, the meaning of this formula is that the fault event F(r) occurring on the line is due to numerous distribution events x i,n The occurrence or non-occurrence of the fault event F(r) is the cause; using Bayes' theorem, the probability P(F) corresponding to the fault event F(r) occurring on the line at this time is inferred. r The calculation formula is the chain rule shown below:

[0030] P(F r )=P(x1)P(x2|x1)...P(x i |x1x2...x i-1 )

[0031] P(x1) represents the probability of node x1 occurring, and P(x2|x1) represents the probability of x2 occurring given that x1 has occurred.

[0032] For a network with T time slices, the distribution events in the event chain include: power outage event A (transformer area user), preliminary assessment of power outage type B, user complaint call C, low-voltage branch terminal event D, transformer terminal event E, and fault event F; where fault events {F = 1, l = 1, 2, 3, 4} represent user events, meter box events, low-voltage branch events, and transformer area events, respectively. Within time slice T, the posterior probability of line faults caused by these distribution events is further expressed by the following formula:

[0033]

[0034] In the formula, P(F=l) T Let represent the probability that fault event l occurs in the line during the T-th time slice, (·). 1,T It represents the observations of the set of elements contained in a certain feature over time slices 1 to T.

[0035] The above-mentioned method for bidirectional fault assessment of distribution networks combining correlation matrix and dynamic Bayesian methods, specifically step four, involves the following steps:

[0036] (4-1) Based on the complete dataset constructed in step (3-1), the interval of the input membership function is determined by the fuzzy clustering algorithm, and fuzzy inference rules are formulated; according to the probability of different fault events, the output membership function is divided into four levels: low, medium, high and very high, and fuzzy weights are assigned to each level respectively.

[0037] (4-2) Based on the probability of power outage events occurring in each region obtained from the reasoning in step (3-3), construct the output membership function to characterize the degree of fault;

[0038] (4-3) Determine whether the power outage probabilities of each area obtained in step (3-3) need to be corrected. If correction is required, the power outage probabilities are used as inputs to the fuzzy inference system, and the defuzzified values ​​are obtained through the defuzzification process. The defuzzified values ​​are used to correct the original maximum probability state, thereby adjusting and optimizing the power outage probabilities.

[0039] (4-4) Compare the results obtained from the test set with the actual recorded data to evaluate the performance of the model in the assessment of power outage areas under conditions of complete and incomplete information.

[0040] The beneficial effects of this invention are as follows:

[0041] 1. This invention, based on traditional matrix algorithms, constructs a causal correlation matrix by combining the causal relationship between faulty feeders and nodes, thus improving fault criteria and effectively solving the feeder identification problem under conditions of missed and false fault reports. This method not only reduces computational load but also improves the accuracy of fault location, and can adapt to complex power distribution network operating environments.

[0042] 2. This invention constructs a power outage area assessment model based on dynamic fuzzy Bayesian networks, which incorporates the causal relationship between user power outage events and various areas of the distribution network into the analysis, and integrates power distribution data from the user side and the line side, effectively solving the problem of power outage area assessment under conditions of incomplete information and reducing the dependence on data integrity.

[0043] 3. This invention combines fuzzy theory and actual fault data to design a membership function and inference rules. By evaluating whether the probability obtained from the inference needs to be corrected, the probability that needs to be corrected is used as input for fuzzification processing, and then defuzzification is used to obtain the defuzzified value, which ultimately realizes the correction of the probability of the power outage area and improves the accuracy of the power outage area assessment.

[0044] 4. The method proposed in this invention can achieve a high accuracy rate in judging the power outage area even under conditions of incomplete fault information, and is superior to traditional methods.

[0045] 5. This invention can not only identify faulty feeders, but also conduct in-depth analysis of the causes of the faults. Through a two-way analysis strategy, it provides more comprehensive fault assessment results, which greatly improves the efficiency of operation and maintenance personnel in inspecting and maintaining faulty areas, shortens fault handling time, and ensures the stable operation of the distribution network. Attached Figure Description

[0046] Figure 1 This is a flowchart of the present invention.

[0047] Figure 2 This is a schematic diagram of a simple power distribution network topology.

[0048] Figure 3 This is a flowchart for analyzing cases where fault information is missed.

[0049] Figure 4 This is a flowchart for analyzing false alarms in fault information.

[0050] Figure 5 This is a schematic diagram of the dynamic fuzzy Bayes principle.

[0051] Figure 6 This is a schematic diagram of a dynamic fuzzy Bayesian structure.

[0052] Figure 7 This is a schematic diagram of a typical medium- and low-voltage power distribution network.

[0053] Figure 8 This is a schematic diagram of a dynamic fuzzy Bayesian network structure.

[0054] Figure 9 This is a schematic diagram of the actual dynamic fuzzy Bayesian network structure.

[0055] Figure 10 This is a schematic diagram of a hierarchical fuzzy inference system.

[0056] Figure 11 This is a 3D curve graph with fuzzy rules.

[0057] Figure 12 Schematic diagram of a single-source, multi-T open-loop distribution network structure

[0058] Figure 13 This is a performance comparison chart between the method of this invention and the traditional method.

[0059] Figure 14 This is a performance analysis diagram of the method of the present invention under conditions of incomplete information. Detailed Implementation

[0060] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0061] like Figure 1 As shown, a bidirectional fault assessment method for distribution networks combining correlation matrix and dynamic Bayesian methods includes the following steps:

[0062] Step 1: Construct an association matrix based on the actual topology of the distribution network, and combine it with the fault information matrix to construct a fault assessment matrix.

[0063] Step one requires constructing an correlation matrix based on the actual topology of the distribution network, so as to... Figure 2For example, the specific steps of step one are as follows:

[0064] (1-1) When a fault occurs on a feeder in the distribution network, the fault indicator connected to the first end of the fault feeder will monitor whether a fault current exceeding the set value flows through the fault indicator and upload the monitored signal to the cloud platform. The fault indicator installed at the end of the fault feeder will not send a signal. The cloud platform generates a row vector matrix with the same number of columns as D′ as the fault information matrix G based on the uploaded information, where D′ is the causal correlation matrix between the feeder section and the switch node.

[0065] (1-2) The fault assessment matrix Q is obtained by multiplying the inverse matrices of G and D′ as follows:

[0066]

[0067] In the formula, when the element in the p-th row of Q is 1, it indicates that the p-th feeder has a fault; when the element in the p-th row of G is 0, it indicates that the p-th feeder has no fault. When the element in the q-th row of G is 1, it indicates that there is a fault current exceeding the setting value flowing through the fault indicator at the q-th node; when the element in the q-th row of G is 0, it indicates that there is no fault current exceeding the setting value flowing through the fault indicator at the q-th node. The element d in the p-th row and q-th column of D′... pq A value of 1 indicates that a fault in the p-th feeder affected the q-th fault indicator, d pq A value of 0 indicates that the p-th feeder is not associated with the q-th fault indicator.

[0068] Step 2: Establish fault criteria for both missed and false alarms of node fault information to achieve top-down fault feeder identification.

[0069] Figure 3 and Figure 4 The flowcharts are for analyzing fault information under the conditions of missed and false alarms. When a node fault information is missed, it is manifested by the fact that the fault information uploaded by the feeder terminal unit (FTU) at that node is empty. Figure 2 Taking the information reported by the FTU as G = [1*1 0 0 0] as an example, the following two judgment rules are constructed when processing missed reports, and the process is as follows: Figure 3 As shown.

[0070] Rule 1: When the fault information of the nodes before and after a node is the same, the missed fault information is set to the same fault information as the nodes before and after it; Figure 2 The FTU reported information missing at node 2: For example, G = [1*1 0 0 0]. The missing fault information in G can be set to 1, and the judgment result is that the fault is in feeder section (3), and the judgment result is correct.

[0071] Rule 2: When the fault information of a node's preceding and following adjacent nodes differs, the node is considered invalid and removed from both D and G analyses; a new assessment is then conducted. Figure 2 The FTU reported information missing at node 3: For example, G = [1 1*0 0 0]. After removing node 3 from the network and reconstructing G and D, we get Q = [0 1 0 0 0]. At this time, the judgment result is that the feeder section (2) is faulty, and the judgment result is correct.

[0072] When a fault is falsely reported, it manifests as a reversal of the fault information uploaded by the FTU at that node; that is, the fault information is actually 1 but reported as 0, or the fault information is actually 0 but reported as 1. The following two judgment rules are established, and the process is as follows: Figure 4 As shown:

[0073] Rule 3: After uploading false fault information, if the analysis results show multiple section faults on the same line, error correction is performed based on comparison of adjacent node information. If the fault information of a node differs from the fault information uploaded by its immediate and adjacent nodes, then the fault information of that node is considered a false alarm. The false alarm information is corrected to be the same as that of its adjacent nodes. If an element in the analysis matrix has a value of -1, then that value is set to 0. When fault information is falsely reported, it may result in an element in the analysis matrix having a value of -1. In this case, -1 has the same meaning as 0, so -1 can be set to 0. Figure 2 The FTU reports false alarm information at node 5: For example, we get Q = [0 1 1 0 1 0], and the judgment result is that faults occur in sections (1), (3), and (5). Since the probability of multiple faults occurring simultaneously in multiple sections of a line is low, the fault information matrix needs to be adjusted. Since the fault information of the two nodes adjacent to node 5 is 0, and since the judgment result is a multi-section fault, the fault information of node 5 is modified to 0.

[0074] Rule 4: If the analysis result is incorrect after uploading false alarm information, it is necessary to collect the fault information again. Figure 2 The FTU reports false alarm information at node 3: For example, Q = [0 1 0 0 00] is obtained, and the judgment result is that the section (2) is faulty, and the FTU on the node needs to re-report the fault information.

[0075] Step 3: Construct a dynamic Bayesian network based on the actual topology of the distribution network and the "station-line-transformer-customer" relationship, and infer the fault probability of each area.

[0076] Traditional Bayesian networks primarily rely on explicit probability distributions to describe causal relationships and conditional dependencies between events. However, in practical power distribution networks, due to complex operating environments, uncertain equipment states, and the influence of external factors, the probability of events often exhibits significant fuzziness and uncertainty. To address this issue, the theory of fuzzy Bayesian networks in power distribution introduces fuzzy sets and membership functions to defuzzify uncertain information, thus combining fuzzy theory with Bayesian networks. This allows the network to enhance its reasoning ability regarding event probabilities under fuzzy and uncertain conditions. Its basic principle is as follows: Figure 5 As shown.

[0077] Fuzzy theory, through the quantification of fuzzy information, enables systems to still derive reasonable and effective inferences when faced with fuzzy inputs. Furthermore, fuzzy theory can transform the probability information calculated in Bayesian networks into membership degrees, and combine this with fuzzy inference systems and corresponding inference rules to analyze and judge fault states. A schematic diagram of fault inference is shown below. Figure 6 As shown.

[0078] A typical medium- and low-voltage distribution network structure on a feeder is as follows: Figure 7 As shown. Figure 7 This includes medium-voltage lines, medium-voltage branch line switches, low-voltage lines of the distribution area, low-voltage user side, meter boxes, and distribution terminals such as DTUs, FTUs, and smart meters on the lines. When a power outage occurs on the low-voltage user side of the distribution area due to a fault in the distribution network, the smart meter at the user's location will automatically collect the outage information. As a terminal device, the smart meter can not only monitor the user's power consumption status in real time, but also transmit data to a concentrator or data acquisition terminal (DTU). At the same time, other distribution terminal devices such as feeder terminal units (FTUs) and remote terminal units (RTUs) will also generate or collect event data related to the fault. However, due to the complexity of the distribution network operating environment, there may be problems such as missed reports and false reports during data collection, making it difficult to accurately locate the fault point from a single data source. Therefore, it is necessary to integrate multiple data sources and combine them with a large amount of historical data for in-depth analysis to improve the accuracy and reliability of fault diagnosis.

[0079] Based on this, construct as Figure 8 The dynamic Bayesian network shown combines various event data in the power distribution network to achieve probabilistic reasoning between events, and can infer the probability of other related events occurring when a certain event occurs.

[0080] The specific steps of step three are as follows:

[0081] (3-1) Integrate the fault data from all sides of the distribution network, including the station, line, transformer and customer, clean and integrate the data, construct a complete dataset containing fault information, and divide it into training set and test set.

[0082] (3-2) Based on the power outage events of users in the transformer area, and combined with the dependency relationship and time sequence characteristics between fault data, a dynamic Bayesian network DBN is constructed. The DBN includes six key nodes: the proportion of power outages of users in the transformer area, the preliminary judgment of event types, the complaint telephone numbers of users in the transformer area, the event information of low-voltage branch terminal, the information of transformer terminal and the final judgment event, as shown in Table 1.

[0083] Table 1 Bayesian Network Node Information Table

[0084]

[0085] Dynamic Bayesian networks can be further represented as follows: Figure 9 As shown. Figure 9 In this system, the network uses power outage events within the distribution area as the basis for analysis and combines the time sequence relationship between distribution terminal information and data to infer the status of other nodes, achieving a bottom-up comprehensive assessment of the fault area and solving the problem of locating the power outage area.

[0086] (3-3) Learn the node parameters of DBN by maximum likelihood estimation and calculate the probability of failure in different regions of the test set.

[0087] For the i-th node x in the power distribution Bayesian network i x i The data contains m power distribution events. Assume that all events related to node x have been obtained. i The distribution events of the relevant nodes are E′={e1,e2,···,e n}, then x i The kth power distribution event e k The conditional probabilities of occurrence are as follows:

[0088]

[0089] T represents the number of time slices. Represents the power distribution event e in the t-th time slice. k Par represents the probability of a power distribution event occurring, P B For the transition probability;

[0090] When a fault occurs in the distribution network, leading to a power outage, a fault event chain can be represented as F(r) = {t}. i│ x i,1 ,x i,2 ,…x i,j …x i,n}, ti x represents i Fault records, x i,n x represents i The corresponding nth distribution event, the meaning of this formula is that the fault event F(r) occurring on the line is due to numerous distribution events x i,n The occurrence or non-occurrence of the fault event F(r) is the cause; using Bayes' theorem, the probability P(F) corresponding to the fault event F(r) occurring on the line at this time is inferred. r The calculation formula is the chain rule shown below:

[0091] P(F r )=P(x1)P(x2|x1)...P(x i |x1x2...x i-1 )

[0092] P(x1) represents the probability of node x1 occurring, and P(x2|x1) represents the probability of x2 occurring given that x1 has occurred.

[0093] For a network with T time slices, the distribution events in the event chain include: power outage event A (transformer area user), preliminary assessment of power outage type B, user complaint call C, low-voltage branch terminal event D, transformer terminal event E, and fault event F; where fault events {F = 1, l = 1, 2, 3, 4} represent user events, meter box events, low-voltage branch events, and transformer area events, respectively. Within time slice T, the posterior probability of line faults caused by these distribution events is further expressed by the following formula:

[0094]

[0095] In the formula, P(F=l) T Let represent the probability that fault event l occurs in the line during the T-th time slice, (·). 1,T It represents the observations of the set of elements contained in a certain feature over time slices 1 to T.

[0096] Step 4: Construct the membership function and fuzzy inference rules to obtain the fuzzy inference system used to adjust the failure probability.

[0097] The specific steps of step four are as follows:

[0098] (4-1) Based on the complete dataset constructed in step (3-1), the interval of the input membership function is determined using a fuzzy clustering algorithm, and fuzzy inference rules are formulated. According to the probability of different fault events, the output membership function is divided into four levels: low (L), medium (M), high (H), and extremely high (E), and assigned fuzzy weights of 0.1, 0.3, 0.6, and 1 respectively. For example, a transformer substation power outage is usually accompanied by a large-scale user power outage, therefore its corresponding risk level is extremely high. The constructed fuzzy rules are shown in Table 2. Table 2 shows the fuzzy rules between different features and... Figure 10 The first layer of the fuzzy inference system generates rules corresponding to the risk level, while the other layers generate corresponding fuzzy rules based on the relationship between the risk level and the features. The three-dimensional curve of the fuzzy rules is shown below. Figure 11 As shown.

[0099] Table 2 Fuzzy Inference Rules

[0100]

[0101]

[0102] (4-2) Based on the probability of power outage events occurring in each region obtained from the reasoning in step (3-3), construct the output membership function to characterize the degree of fault.

[0103] (4-3) Determine whether the power outage probabilities of each area obtained in step (3-3) need to be corrected. If correction is required, the power outage probabilities are used as inputs to the fuzzy inference system, and the defuzzified values ​​are obtained through the defuzzification process. The defuzzified values ​​are used to correct the original maximum probability state, thereby adjusting and optimizing the power outage probabilities.

[0104] (4-4) Compare the results obtained from the test set with the actual recorded data to evaluate the performance of the model in the assessment of power outage areas under conditions of complete and incomplete information.

[0105] Step 5: Finally, deduce the most likely fault area to achieve bottom-up fault area assessment.

[0106] Example

[0107] (1) Top-down fault feeder identification

[0108] Using Python programming methods Figure 12 The analysis focuses on a single-source, multi-T open-loop distribution network. This distribution system comprises 12 switching nodes and 12 feeder sections, of which L1 to L2 are... 12Each feeder section and each switching node is equipped with an FTU (Fault Transfer Unit) to monitor for fault current flow in real time. Considering practical application scenarios, this invention sets up six typical fault scenarios—normal fault information, missed alarms, and false alarms—to verify the accuracy and feasibility of the improved matrix algorithm. The test results are shown in Table 3.

[0109] Table 3 Fault Feeder Identification Results

[0110]

[0111] The results in the table above show that the proposed matrix algorithm can correctly identify the faulty feeder segment in both single-segment and multi-segment fault scenarios, with both false and false alarms. The faulty segments L8 and L... 10 Analysis is performed when a fault occurs simultaneously and the fault information of node S4 is falsely reported.

[0112] When section L8 and L 10 When a fault occurs, and node S4 reports a false fault report, the fault information matrix is ​​as follows: The constructed causal relationship matrix D′ is shown in the following equation.

[0113]

[0114] The fault judgment matrix Q can be obtained as Q = [0 0 1 0 0 0 0 1 0 1 0 0]. In the judgment result, multiple segment faults are found. According to the judgment criteria, the fault information matrix needs to be checked. It is found that the fault information uploaded by node 4 and its two adjacent nodes are different. Therefore, the fault information uploaded by node 4 is set to 0, and then the judgment is re-examined to obtain the result.

[0115] (2) Bottom-up fault area analysis

[0116] To verify the performance of the fault region assessment model proposed in this invention under conditions of complete information, the training and test sets were divided in an 8:2 ratio. The same set of data was tested using traditional Bayesian networks, dynamic Bayesian networks, and the method of this invention, respectively. In this invention, the Receiver Operating Characteristic (ROC) curve is used to evaluate the assessment performance. By plotting the ROC curve, the classification performance of the assessment model under different thresholds can be visually displayed. In the comparison process, in addition to the ROC curve, three key indicators were also introduced: Standard Error (SE), Confidence Interval (CI), and Accuracy (A%), to comprehensively evaluate the model's performance. The results are shown in Table 4 and... Figure 13 As shown.

[0117] Table 4 Comparison of Evaluation Indicators

[0118]

[0119] Depend on Figure 13 As shown in Table 4, compared to directly using the Bayesian network method, the method of the present invention improves performance on AUROC, SE, and CI by 10.63%, 35.96%, and 7.95%, respectively; compared to the dynamic Bayesian network method, the method of the present invention improves performance on AUROC, SE, and CI by 7.57%, 19.78%, and 6.46%, respectively. In terms of A%, the method of the present invention improves performance on AUROC by 10.7% and on dynamic Bayesian networks by 5.9%, respectively.

[0120] To verify the performance of the fault region assessment model proposed in this invention under conditions of incomplete information, a grouped experiment was designed. First, a dynamic Bayesian network model was trained using complete training data. Then, a portion of terminal data was randomly deleted from the complete test set to generate an incomplete dataset, simulating data loss in real-world applications and thus evaluating the model's assessment capability under incomplete information conditions. The experiment was designed with four groups: complete information, 10% information loss, 20% information loss, and 30% information loss. The results are shown in Table 5. Figure 14 As shown.

[0121] Table 5 Comparison of Evaluation Indicators

[0122]

[0123] Depend on Figure 14 As shown in Table 5, when the degree of information loss is 10% and 20%, the accuracy of the judgment of the method of the present invention reaches 83.59% and 80.24%, respectively.

[0124] In summary, the method proposed in this invention, based on traditional matrix algorithms, constructs a causal correlation matrix by combining the causal relationship between faulty feeders and nodes, thus improving the fault judgment criteria and effectively solving the feeder identification problem under conditions of missed and false fault reports. This method not only reduces the computational load but also improves the accuracy of fault location and can adapt to complex distribution network operating environments. By combining fuzzy theory and actual fault data, this invention designs a membership function and inference rules. By evaluating whether the probabilities obtained from the inference need to be corrected, the probabilities that need to be corrected are used as input for fuzzification processing, and then defuzzification is performed to obtain the defuzzified value, ultimately realizing the correction of the probability of the power outage area and improving the accuracy of fault area assessment.

Claims

1. A bidirectional fault assessment method for distribution networks combining correlation matrix and dynamic Bayesian methods, characterized in that, Includes the following steps: Step 1: Construct an association matrix based on the actual topology of the distribution network, and combine it with the fault information matrix to build a fault assessment matrix; Step 2: Establish fault criteria for both missed and false alarms of node fault information, thereby achieving top-down fault feeder identification; Step 3: Construct a dynamic Bayesian network based on the actual topology of the distribution network and the "station-line-transformer-customer" relationship, and infer the fault probability of each area; The specific steps of step three are as follows: (3-1) Integrate the fault data from all sides of the distribution network "station-line-transformer-customer", perform data cleaning and integration, construct a complete dataset containing fault information, and divide it into training set and test set; (3-2) Based on the power outage events of users in the transformer area, and combined with the dependency relationship and time series characteristics between fault data, a dynamic Bayesian network DBN is constructed. The DBN includes six key nodes: the proportion of power outages of users in the transformer area, the preliminary judgment of event types, the complaint telephone numbers of users in the transformer area, the event information of low-voltage branch terminal, the information of transformer terminal and the final judgment of events. (3-3) Learn the node parameters of DBN by maximum likelihood estimation and calculate the probability of failure in different regions of the test set; Step 4: Construct membership functions and fuzzy inference rules to obtain a fuzzy inference system for adjusting fault probabilities; Step 5: Finally, deduce the most likely fault area to achieve bottom-up fault area assessment.

2. The bidirectional fault assessment method for distribution networks combining correlation matrix and dynamic Bayesian methods according to claim 1, characterized in that, The specific steps of step one are as follows: (1-1) When a fault occurs on a feeder in the distribution network, the fault indicator connected to the beginning of the faulty feeder will monitor whether a fault current exceeding the set value flows through the fault indicator and upload the monitored signal to the cloud platform. The fault indicator installed at the end of the faulty feeder will not emit a signal. The cloud platform generates a signal based on the uploaded information. Row vector matrices with the same number of columns are used as fault information matrices G , This is the causal relationship matrix between the feeder section and the switch node; (1-2) will G , The fault assessment matrix is ​​obtained by multiplying the inverse matrix. Q as follows: ； In the formula, Q The Middle p When the element of a row is 1, it indicates that the first row is... p The feeder line failed, the first p When the element of a row is 0, it indicates that the first element is 0. p No faults occurred in the feeder line; G The first in q When the row element is 1, it indicates the first row. q A fault current exceeding the set value is flowing through the fault indicator on the first node. q When the row element is 0, it indicates the first row. q No fault current exceeding the set value flowed through the fault indicator on any node; The Middle p Line 1 q Column elements d pq When it is 1, it means the first p When the feeder line failed, it affected the first q One fault indicator, d pq When it is 0, it means the first time. p feeder and the first q The fault indicators are unrelated.

3. The bidirectional fault assessment method for distribution networks combining correlation matrix and dynamic Bayesian methods according to claim 1, characterized in that, In step two, when a node fault information is missed, it is manifested as the fault information uploaded by the feeder terminal unit (FTU) at that node being empty. When processing the missed information, the following two judgment rules are constructed: Rule 1: When the fault information of the nodes before and after a node is the same, the missed fault information is set to the same fault information as the nodes before and after it. Rule 2: When the fault information of a node's preceding and following adjacent nodes differs, the node is considered invalid and removed from the list. D , G The node will not be considered in the analysis, and a new assessment will be conducted.

4. The bidirectional fault assessment method for distribution networks combining correlation matrix and dynamic Bayesian methods according to claim 3, characterized in that, In step two, when a fault information is falsely reported, it manifests as a reversal of the fault information uploaded by the FTU at that node, i.e., the actual fault information is 1 but is reported as 0, or the actual fault information is 0 but is reported as 1. The following two judgment rules are constructed: Rule 3: After uploading false alarm fault information, if the analysis results show that multiple sections of the same line have faults, the error is corrected by comparing the adjacent information of the nodes. If the fault information of a node is different from the fault information uploaded by its neighboring nodes, then the fault information of that node is a false alarm. The false alarm information is modified to be the same as that of the neighboring nodes. If an element in the judgment matrix has a value of -1, then that value is set to 0. Rule 4: If the analysis result is incorrect after uploading false alarm information, it is necessary to collect the fault information again.

5. The bidirectional fault assessment method for distribution networks combining correlation matrix and dynamic Bayesian methods according to claim 4, characterized in that, In step (3-3), for the i-th node in the power distribution Bayesian network... x i , x i The data contains m power distribution events. Assume all events related to the nodes have been obtained. x i Distribution events at relevant nodes E ′={ e 1 , e 2 ,···, e n },So x i The k A power distribution incident e k The conditional probabilities of occurrence are as follows: ； T Indicates the number of time slices. Indicates the first t Power distribution events in a time slice e k , Par Indicates the probability of a power distribution event occurring. P B For the transition probability; When a fault occurs in the distribution network, leading to a power outage, a fault event chain is represented as follows: F ( r )={ t i│ x i,1 , x i,2, … x i,j … x i,n }, t i x represents i Fault records, x i,n express x i The corresponding number n A power distribution event, the meaning of which is a fault event occurring on the line. F ( r This is due to numerous power distribution incidents. x i,n The occurrence or non-occurrence of the fault event leads to the inference of the fault event on the line at this time using Bayes' theorem. F ( r The probability corresponding to ) P ( F r The calculation formula is the chain rule shown below: ； Represents a node The probability of occurrence express In the event of The probability of occurrence; For containing T In a time-slice network, the power distribution events in the event chain include power outage events for users in the transformer area. A Preliminary assessment of the type of power outage B User complaint hotline in the district C Low-voltage branch line terminal events D Terminal incidents in the distribution area E Fault events F Among them, the fault events { F = l , l =1,2,3,4} represent user events, meter box events, low-voltage branch line events, and transformer area events, respectively, within the time slice. T Within this context, the posterior probability of line faults caused by these distribution events is further expressed by the following formula: ； In the formula, P ( F = l ) T Indicates the first T A fault event occurred in the circuit within a certain time slice. l The probability of (·) 1,T This represents the set of elements contained in a certain characteristic factor in time slice 1 to... T The observed values.

6. The bidirectional fault assessment method for distribution networks combining correlation matrix and dynamic Bayesian methods according to claim 1, characterized in that, The specific steps of step four are as follows: (4-1) Based on the complete dataset constructed in step (3-1), the interval of the input membership function is determined by the fuzzy clustering algorithm, and the fuzzy inference rules are formulated; according to the probability of different fault events, the output membership function is divided into four levels: low, medium, high and very high, and fuzzy weights are assigned to them respectively. (4-2) Based on the probability of a power outage event occurring in each region obtained from the reasoning in step (3-3), construct an output membership function to characterize the degree of fault; (4-3) Determine whether the power outage probabilities of each area obtained in step (3-3) need to be corrected. If correction is required, the power outage probabilities are used as inputs to the fuzzy inference system, and the defuzzified values ​​are obtained through the defuzzification process. The defuzzified values ​​are used to correct the original maximum probability state, thereby adjusting and optimizing the power outage probabilities. (4-4) Compare the results obtained from the test set with the actual recorded data to evaluate the performance of the model in the assessment of power outage areas under conditions of complete information and incomplete information.

Images