Power distribution switch mechanism risk detection method and system

Abstract

The invention provides a power distribution switch mechanism risk detection method. The power distribution switch mechanism risk detection method comprises the steps of determining risk factors causing faults in a power distribution switch mechanism and risk indexes contained in the risk factors; based on an improved fuzzy analytic hierarchy process, constructing a fuzzy judgment matrix, and assigning and selecting the fuzzy judgment matrix meeting a predetermined condition in each risk factor and a corresponding initial weight vector; historical data of risk indexes are obtained, the initial weight vector of the selected fuzzy judgment matrix is combined, the correction amount of the initial weight vector of the selected fuzzy judgment matrix is obtained in the improved Bayesian neural network model, and the correction amount is further corrected to obtain a corresponding correction weight vector; and calculating the score of each risk factor according to the assignment of the selected fuzzy judgment matrix and the correction weight vector, and further calculating the score of the power distribution switch mechanism to determine the risk level. By implementing the method, the problems that the fault risk of the existing power distribution switch mechanism is difficult to identify and the health state cannot be quantitatively analyzed can be solved.

Description

Technical Field

[0001] The present invention relates to the field of power detection technology, and in particular to a method and system for detecting risks of a distribution switch mechanism. Background Art

[0002] Distribution switchgear is a key piece of electrical equipment in distribution networks, responsible for connecting, disconnecting, and protecting power. With the continuous growth of urban power grid loads and the advancement of smart grid construction, distribution systems are placing higher demands on the stability and intelligence of distribution switchgear.

[0003] Currently, distribution switch mechanisms, as core components, are directly related to the operational reliability and safety of distribution switchgear. However, traditional distribution switch mechanism operation and management rely primarily on manual inspections or scheduled maintenance. This approach has many shortcomings, including the difficulty in timely identifying potential failure risks and the inability to quantitatively analyze the switch mechanism's health status.

[0004] Therefore, it is necessary to provide a new distribution switch mechanism risk detection method that can solve the problems of difficulty in identifying fault risks of existing distribution switch mechanisms and inability to quantify and analyze health status. Summary of the Invention

[0005] The technical problem to be solved by the embodiments of the present invention is to provide a distribution switch mechanism risk detection method and system, which can solve the problems of difficulty in identifying fault risks of existing distribution switch mechanisms and inability to quantify and analyze health status.

[0006] In order to solve the above technical problems, an embodiment of the present invention provides a method for detecting risks of a distribution switch mechanism, the method comprising the following steps:

[0007] S1. Determine the risk factors that cause failures in the distribution switch mechanism and the risk indicators contained in each risk factor;

[0008] S2. Based on the improved fuzzy analytic hierarchy process, a fuzzy judgment matrix formed by its corresponding risk indicator is constructed for each risk factor and assigned a value, and the fuzzy judgment matrix of each risk factor after the assignment is screened to select a fuzzy judgment matrix that meets predetermined conditions in each risk factor, and further obtain an initial weight vector of the selected fuzzy judgment matrix in each risk factor; wherein the improved fuzzy analytic hierarchy process is constructed based on a preset fuzzy analytic hierarchy process, and the fuzzy entropy method is introduced into the fuzzy analytic hierarchy process to perform weighted correction on the fuzzy weights of each fuzzy judgment matrix;

[0009] S3. Obtain historical data of risk indicators contained in each risk factor, and combine them with the initial weight vector of the fuzzy judgment matrix selected in each risk factor, to obtain a correction amount of the initial weight vector of the fuzzy judgment matrix selected in each risk factor in the improved Bayesian neural network model, and further obtain a correction weight vector of the fuzzy judgment matrix selected in each risk factor based on the initial weight vector of the fuzzy judgment matrix selected in each risk factor; wherein the improved Bayesian neural network model is constructed based on a preset Bayesian neural network model, which introduces historical data and input samples into the input layer of the Bayesian neural network model, and designs a three-layer structure in the hidden layer of the Bayesian neural network model, and introduces predetermined connection weights and activation functions into each layer structure;

[0010] S4. Calculate the score of each risk factor based on the assignment of the selected fuzzy judgment matrix in each risk factor and its corresponding modified weight vector, and calculate the score of the distribution switch mechanism based on the score of each risk factor to determine the risk level of the distribution switch mechanism.

[0011] Wherein, the step S2 specifically includes:

[0012] S21. Based on the risk indicators contained in the same risk factor, the fuzzy judgment matrix corresponding to each risk factor is constructed respectively. The judgment results of the fuzzy membership of the importance between the indicators at the same level are obtained through expert review, and the fuzzy judgment matrix corresponding to each risk factor is assigned a value. Among them, the expression of the fuzzy judgment matrix of a certain risk factor is r ij is the risk indicator x i Relative to the risk indicator x j The importance of fuzzy membership; i, j are the row and column numbers of indicators respectively;

[0013] S22, through the formula Calculate the fuzzy weight of the fuzzy judgment matrix of each risk factor

[0014] S23, through the formula Fuzzy weight of the fuzzy judgment matrix for each risk factor Perform weighted fusion to obtain the fusion weight of the fuzzy judgment matrix of each risk factor Among them, λ∈[0,1] is the adjustment parameter; w′ i is the entropy weight obtained by the fuzzy entropy method, and its expression is

[0015] S24. Fusion weight of the fuzzy judgment matrix of each risk factor Calculate the maximum eigenvalue λ of the fuzzy judgment matrix of each risk factormax , and by the formula Calculate the consistency index CI of the fuzzy judgment matrix of each risk factor;

[0016] S25. Find the corresponding RI value of the fuzzy judgment matrix of each risk factor in the preset consistency index value table, and use the formula Calculate the consistency ratio CR of the fuzzy judgment matrix of each risk factor;

[0017] S26, determining whether the consistency ratio CR of the fuzzy judgment matrix of each risk factor is less than a preset ratio threshold; if yes, proceeding to the next step S27; if not, jumping to step S28;

[0018] S27, determine the conditions are met and output as the final selected fuzzy judgment matrix, and further calculate the fuzzy judgment matrix selected in each risk factor corresponding to the maximum eigenvalue λ max The initial weight vector Where n is the total number of risk indicators contained in the current risk factor; is the initial weight of the i-th risk indicator of the current risk factor;

[0019] S28 , after re-assigning the fuzzy judgment matrices whose consistency ratio CR is greater than the ratio threshold, return to step S22 .

[0020] Wherein, in the step S3, the input layer of the improved Bayesian neural network model is based on the input sample X of the input layer of the Bayesian neural network model as the initial weight vector of the fuzzy judgment matrix of each risk factor Combined with the historical data of risk indicators contained in each risk factor to form a joint input sample Among them, x i is the historical data value of the i-th risk indicator contained in the current risk factor;

[0021] The hidden layer of the improved Bayesian neural network model is based on the hidden layer of the Bayesian neural network model, and three new hidden layers are constructed; wherein each new hidden layer includes connection weights and ReLU activation function; kernel function K(x i ,x j ) reflects the risk indicator x i and risk indicator x j the degree of correlation between them;

[0022] The output result of the output layer of the improved Bayesian neural network model is the correction value ΔW=[δw1,δw2,...,δw n ]; where δ is the correction coefficient.

[0023] Among them, in the step S3, by formula Calculate the modified weight vector of the selected fuzzy judgment matrix in each risk factor

[0024] Wherein, the step S4 specifically includes:

[0025] After multiplying the value of the selected fuzzy judgment matrix in each risk factor by its corresponding modified weight vector, the accumulated sum is output as the score of each risk factor;

[0026] After multiplying the weight value pre-assigned to each risk factor by the score of each risk factor, the accumulated sum is output as the score of the distribution switch mechanism;

[0027] The obtained score of the distribution switch mechanism is compared with a plurality of preset level thresholds respectively, and the risk level of the distribution switch mechanism is obtained according to the comparison result.

[0028] The risk factors include equipment status, operating environment, operation and maintenance management status and historical fault status;

[0029] The risk indicators contained in the equipment status include the number of operations, contact resistance, temperature rise and spring fatigue;

[0030] The risk indicators of the operating environment include ambient temperature, ambient humidity, ambient pollution level, altitude and corrosive gas concentration;

[0031] The risk indicators included in the operation and maintenance management status include inspection frequency, maintenance cycle compliance rate, defect closure rate and operation and maintenance record integrity;

[0032] The risk indicators contained in the historical fault status include the number of faults, the diversity of fault types, the number of tripping events and the aging years of equipment.

[0033] An embodiment of the present invention further provides a risk detection system for a power distribution switch mechanism, comprising:

[0034] A risk indicator hierarchical construction unit for determining risk factors causing failures in the distribution switch mechanism and the risk indicators contained in each risk factor;

[0035] a risk indicator initial weight analysis unit, configured to construct a fuzzy judgment matrix formed by the corresponding risk indicator for each risk factor based on an improved fuzzy analytic hierarchy process and assign values, and to screen the fuzzy judgment matrix of each risk factor after the assignment to select a fuzzy judgment matrix that satisfies a predetermined condition in each risk factor, and further obtain an initial weight vector of the selected fuzzy judgment matrix in each risk factor; wherein the improved fuzzy analytic hierarchy process is constructed based on a preset fuzzy analytic hierarchy process, and the fuzzy entropy method is introduced into the fuzzy analytic hierarchy process to perform weighted correction on the fuzzy weights of each fuzzy judgment matrix;

[0036] a risk indicator weight correction unit, for obtaining historical data of risk indicators contained in each risk factor, and combining the initial weight vector of the fuzzy judgment matrix selected in each risk factor, to obtain a correction amount of the initial weight vector of the fuzzy judgment matrix selected in each risk factor in the improved Bayesian neural network model, and further obtaining a correction weight vector of the fuzzy judgment matrix selected in each risk factor based on the initial weight vector of the fuzzy judgment matrix selected in each risk factor; wherein the improved Bayesian neural network model is constructed based on a preset Bayesian neural network model, which introduces historical data and input samples into the input layer of the Bayesian neural network model, and designs a three-layer structure in the hidden layer of the Bayesian neural network model and introduces predetermined connection weights and activation functions into each layer structure;

[0037] The risk level detection unit is used to calculate the score of each risk factor based on the assignment of the fuzzy judgment matrix selected in each risk factor and its corresponding modified weight vector, and calculate the score of the distribution switch mechanism based on the score of each risk factor to determine the risk level of the distribution switch mechanism.

[0038] The risk factors include equipment status, operating environment, operation and maintenance management status and historical fault status;

[0039] The risk indicators contained in the equipment status include the number of operations, contact resistance, temperature rise and spring fatigue;

[0040] The risk indicators of the operating environment include ambient temperature, ambient humidity, ambient pollution level, altitude and corrosive gas concentration;

[0041] The risk indicators included in the operation and maintenance management status include inspection frequency, maintenance cycle compliance rate, defect closure rate and operation and maintenance record integrity;

[0042] The risk indicators contained in the historical fault status include the number of faults, the diversity of fault types, the number of tripping events and the aging years of equipment.

[0043] The implementation of the embodiments of the present invention has the following beneficial effects:

[0044] The present invention effectively integrates expert experience with historical data, constructs initial weights through the fuzzy hierarchical analysis method, and fully considers the hierarchical relationship between risk indicators and the fuzziness of subjective judgment, making the indicator system construction more scientific and the structure more reasonable. The initial weights are further corrected through the improved Bayesian neural network model, which improves the stability and prediction reliability of the model in the case of insufficient samples or large data noise. It can effectively detect the risks of distribution switch mechanisms in actual operation, thereby solving the problems of difficulty in identifying fault risks of existing distribution switch mechanisms and the inability to quantify and analyze health status. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, without paying any creative work, other drawings obtained based on these drawings still fall within the scope of the present invention.

[0046] Figure 1 A flow chart of a method for detecting risk of a power distribution switch mechanism provided by an embodiment of the present invention;

[0047] Figure 2 A schematic structural diagram of an improved Bayesian neural network model in a risk detection method for a distribution switch mechanism provided by an embodiment of the present invention;

[0048] Figure 3 A schematic structural diagram of a distribution switch mechanism risk detection system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be described in further detail below with reference to the accompanying drawings.

[0050] like Figure 1 FIG. 1 is a diagram showing a method for detecting risk of a power distribution switch mechanism according to an embodiment of the present invention, the method comprising the following steps:

[0051] Step S1: determining risk factors causing failures in a distribution switch mechanism and risk indicators contained in each risk factor;

[0052] The specific process is to collect historical operation data of distribution switch mechanisms and corresponding risk level samples, and to form a risk assessment data set for distribution switch mechanisms in order to build a risk indicator system.

[0053] At this point, risk factors that may cause failures in the distribution switch mechanism include, but are not limited to, equipment status, operating environment, operation and maintenance management status, and historical fault status. Risk indicators for equipment status include, but are not limited to, the number of operations, contact resistance, temperature rise, and spring fatigue; risk indicators for the operating environment include, but are not limited to, ambient temperature, humidity, pollution level, altitude, and corrosive gas concentration; risk indicators for operation and maintenance management status include, but are not limited to, inspection frequency, maintenance cycle compliance rate, defect closure rate, and the completeness of operation and maintenance records; and risk indicators for historical fault status include, but are not limited to, the number of failures, the diversity of fault types, the number of tripping events, and equipment aging.

[0054] Step S2: Based on the improved fuzzy analytic hierarchy process, a fuzzy judgment matrix formed by its corresponding risk indicator is constructed for each risk factor and assigned a value, and the fuzzy judgment matrix of each risk factor after the assignment is screened to select a fuzzy judgment matrix that meets a predetermined condition in each risk factor, and further obtain an initial weight vector of the selected fuzzy judgment matrix in each risk factor; wherein the improved fuzzy analytic hierarchy process is constructed based on the preset fuzzy analytic hierarchy process, and the fuzzy entropy method is introduced into the fuzzy analytic hierarchy process to perform weighted correction on the fuzzy weights of each fuzzy judgment matrix;

[0055] The specific process is to use the improved fuzzy analytic hierarchy process (FAHP) to construct the fuzzy judgment matrix between the risk indicators at each level, so as to obtain the corresponding initial weight values, specifically:

[0056] Step S21: Based on the risk indicators contained in the same risk factor, construct the fuzzy judgment matrix corresponding to each risk factor, and obtain the judgment result of the fuzzy membership of the importance between the indicators at the same level through expert review, and assign values ​​to the fuzzy judgment matrix corresponding to each risk factor; wherein, the expression of the fuzzy judgment matrix of a certain risk factor is: r ij is the risk indicator x i Relative to the risk indicator x j The importance of fuzzy membership; i and j are the row and column numbers of indicators respectively.

[0057] Step S22: by formula Calculate the fuzzy weight of the fuzzy judgment matrix of each risk factor

[0058] Step S23, by formula Fuzzy weight of the fuzzy judgment matrix for each risk factor Perform weighted fusion to obtain the fusion weight of the fuzzy judgment matrix of each risk factor Among them, λ∈[0,1] is the adjustment parameter; wi ′ is the entropy weight obtained by the fuzzy entropy method, and its expression is It should be noted that step S23 is to correct the fuzzy weights by introducing the fuzzy entropy method in consideration of the large deviation and insufficient accuracy of the judgment of industry experts. This improves the objectivity and stability of the weights.

[0059] Step S24: Fusion weight of the fuzzy judgment matrix of each risk factor Calculate the maximum eigenvalue λ of the fuzzy judgment matrix of each risk factor max , and by the formula Calculate the consistency index CI of the fuzzy judgment matrix of each risk factor.

[0060] Step S25: Find the corresponding RI value of the fuzzy judgment matrix of each risk factor in the preset consistency index value table, and calculate the value of the RI value by the formula Calculate the consistency ratio CR of the fuzzy judgment matrix of each risk factor.

[0061] In one embodiment, the preset consistency index value table is shown in Table 1 below:

[0062] Table 1

[0063] n 1 2 3 4 5 6 7 8 9 RI 0 0 0.52 0.89 1.12 1.24 1.36 1.41 1.46

[0064] When n=1 or n=2, CR=0. When n≥3, the formula To calculate the consistency ratio CR.

[0065] Step S26: determine whether the consistency ratio CR of the fuzzy judgment matrix of each risk factor is less than a preset ratio threshold; if yes, execute the next step S27; if not, jump to step S28.

[0066] In one embodiment, the preset ratio threshold is 0.1. When CR<0.1, the consistency is considered acceptable and the next step S27 is executed; otherwise, the process jumps to step S28 and re-assigns values ​​for appropriate adjustments.

[0067] Step S27: determine whether the conditions are met and output the final selected fuzzy judgment matrix, and further calculate the fuzzy judgment matrix selected in each risk factor corresponding to the maximum eigenvalue λ max The initial weight vector Where n is the total number of risk indicators contained in the current risk factor; is the initial weight of the i-th risk indicator of the current risk factor.

[0068] Step S28 , after re-assigning the fuzzy judgment matrices whose consistency ratio CR is greater than the ratio threshold, return to step S22 .

[0069] Step S3, obtaining historical data of risk indicators contained in each risk factor, and combining the initial weight vector of the fuzzy judgment matrix selected in each risk factor, in the improved Bayesian neural network model, to obtain a correction amount of the initial weight vector of the fuzzy judgment matrix selected in each risk factor, and further for each risk factor The initial weight vector of the fuzzy judgment matrix selected is obtained to obtain a correction weight vector of the fuzzy judgment matrix selected in each risk factor; wherein, the improved Bayesian neural network model is constructed based on a preset Bayesian neural network model, which introduces historical data and input samples into the input layer of the Bayesian neural network model, and designs a three-layer structure in the hidden layer of the Bayesian neural network model and introduces predetermined connection weights and activation functions into each layer structure;

[0070] The specific process is as follows: the traditional Bayesian neural network model consists of three parts: input layer, hidden layer, and output layer; among them, the input layer is mainly used to receive the initial weight vectors of each fuzzy judgment matrix obtained by the improved fuzzy hierarchical analysis method, that is, the input sample X is the initial weight vector of the fuzzy judgment matrix of each risk factor The hidden layer is mainly used for feature extraction and calculation; the output layer is used to output the corrected weight vector.

[0071] In order to further improve the generalization ability of the model, the input layer of the Bayesian neural network model is improved. Specifically, the input sample X of the input layer of the Bayesian neural network model is the initial weight vector of the fuzzy judgment matrix of each risk factor. Combined with the historical data of risk indicators contained in each risk factor to form a joint input sample Among them, x i is the historical data value of the i-th risk indicator contained in the current risk factor.

[0072] At the same time, considering that there are physical constraints and logical relationships between some risk indicators (for example, "mechanical aging" is strongly correlated with "number of actions"), they need to be integrated into the Bayesian neural network model as prior constraints. Therefore, the hidden layer of the Bayesian neural network model is improved. Specifically, three new hidden layers are constructed based on the hidden layer of the Bayesian neural network model. Each new hidden layer includes connection weights. and ReLU activation function; kernel function K(x i ,x j ) reflects the risk indicator x i and risk indicator x jIn addition, to solve the problems of increasing model parameters and increasing training time cost during Bayesian deep neural network training, the Monte Carlo dropout technique is used to calculate the neuron output of the next layer using different dropout masks in each propagation.

[0073] Finally, the output result of the output layer of the Bayesian neural network model is improved to the correction value ΔW=[δw1,δw2,...,δw n ]; where δ is the correction coefficient.

[0074] Based on the above content, the improved structure of the traditional Bayesian neural network model, such as Figure 2 shown.

[0075] It should be noted that the Bayesian neural network model training loss function proposed in the embodiment of the present invention is:

[0076] L total =L MSE (θ)+λ·L reg (θ)

[0077]

[0078] L reg (θ)=D KL [q(A|θ)||P(A)]-E q(W∣θ) [log P(D|A)]

[0079] Among them, L total Represents the overall loss function, L MSE (θ) represents the data loss function, L eeg (θ) represents the regularization loss function, Represents the difference between the model input weight parameters and the sample set labels. KL [q(A|θ)||P(A)] represents the KL divergence between the approximate posterior distribution q(A|θ) of the model weight parameter and the prior distribution P(A), E q(W∣θ) [logP(D|A)] is the expected log-likelihood, which represents the expected likelihood of the model for the training data given the weight distribution, taking into account the uncertainty of the weights. D represents the training data.

[0080] It can be seen that by combining the historical data of the risk indicators contained in each risk factor with the initial weight vector of the fuzzy judgment matrix selected in each risk factor, the correction amount ΔW = [δw1, δw2, ..., δw n ], further through the formula The modified weight vector of the selected fuzzy judgment matrix in each risk factor can be calculated

[0081] Step S4: Calculate the score of each risk factor based on the assignment of the fuzzy judgment matrix selected in each risk factor and its corresponding modified weight vector, and calculate the score of the distribution switch mechanism based on the score of each risk factor to determine the risk level of the distribution switch mechanism.

[0082] The specific process is as follows: first, the value of the selected fuzzy judgment matrix in each risk factor is multiplied by its corresponding modified weight vector, and the accumulated sum is output as the score of each risk factor; second, the weight value pre-assigned to each risk factor is multiplied by the score of each risk factor, and the accumulated sum is output as the score of the distribution switch mechanism; finally, the obtained score of the distribution switch mechanism is compared with multiple preset level thresholds, and the risk level of the distribution switch mechanism is obtained based on the comparison results.

[0083] It should be noted that calculating the weights of the risk indicators at each level in step S4 and comprehensively evaluating the risk level of the distribution switch mechanism from the bottom level to the upper level is a common technical means in this field, and the specific contents will not be repeated here.

[0084] like Figure 3 FIG. 1 is a diagram showing a risk detection system for a power distribution switch mechanism according to an embodiment of the present invention, comprising:

[0085] A risk indicator hierarchical construction unit 110 is used to determine the risk factors causing failures in the distribution switch mechanism and the risk indicators contained in each risk factor;

[0086] The risk indicator initial weight analysis unit 120 is used to construct a fuzzy judgment matrix formed by the corresponding risk indicator for each risk factor based on the improved fuzzy analytic hierarchy process and assign values, and to screen the fuzzy judgment matrix of each risk factor after the assignment to select the fuzzy judgment matrix that meets the predetermined conditions in each risk factor, and further obtain the initial weight vector of the selected fuzzy judgment matrix in each risk factor; wherein the improved fuzzy analytic hierarchy process is constructed based on the preset fuzzy analytic hierarchy process, and the fuzzy entropy method is introduced into the fuzzy analytic hierarchy process to perform weighted correction on the fuzzy weights of each fuzzy judgment matrix;

[0087] The risk indicator weight correction unit 130 is used to obtain historical data of risk indicators contained in each risk factor, and combine the initial weight vector of the fuzzy judgment matrix selected in each risk factor to obtain a correction amount of the initial weight vector of the fuzzy judgment matrix selected in each risk factor in the improved Bayesian neural network model, and further obtain the correction weight vector of the fuzzy judgment matrix selected in each risk factor based on the initial weight vector of the fuzzy judgment matrix selected in each risk factor; wherein the improved Bayesian neural network model is constructed based on a preset Bayesian neural network model, which introduces historical data and input samples into the input layer of the Bayesian neural network model, and designs a three-layer structure in the hidden layer of the Bayesian neural network model, and introduces predetermined connection weights and activation functions into each layer structure;

[0088] The risk level detection unit 140 is used to calculate the score of each risk factor based on the assignment of the fuzzy judgment matrix selected in each risk factor and its corresponding modified weight vector, and calculate the score of the distribution switch mechanism based on the score of each risk factor to determine the risk level of the distribution switch mechanism.

[0089] The risk factors include equipment status, operating environment, operation and maintenance management status and historical fault status;

[0090] The risk indicators contained in the equipment status include the number of operations, contact resistance, temperature rise and spring fatigue;

[0091] The risk indicators of the operating environment include ambient temperature, ambient humidity, ambient pollution level, altitude and corrosive gas concentration;

[0092] The risk indicators included in the operation and maintenance management status include inspection frequency, maintenance cycle compliance rate, defect closure rate and operation and maintenance record integrity;

[0093] The risk indicators contained in the historical fault status include the number of faults, the diversity of fault types, the number of tripping events and the aging years of equipment.

[0094] The implementation of the embodiments of the present invention has the following beneficial effects:

[0095] The present invention effectively integrates expert experience with historical data, constructs initial weights through the fuzzy hierarchical analysis method, and fully considers the hierarchical relationship between risk indicators and the fuzziness of subjective judgment, making the indicator system construction more scientific and the structure more reasonable. The initial weights are further corrected through the improved Bayesian neural network model, which improves the stability and prediction reliability of the model in the case of insufficient samples or large data noise. It can effectively detect the risks of distribution switch mechanisms in actual operation, thereby solving the problems of difficulty in identifying fault risks of existing distribution switch mechanisms and the inability to quantify and analyze health status.

[0096] It is worth noting that in the above system embodiment, the various system modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the various functional modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0097] Those skilled in the art will understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, and the program can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc.

[0098] The above disclosure is merely a preferred embodiment of the present invention and certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for detecting risk of a power distribution switch mechanism, characterized in that: The method comprises the following steps: S1. Determine the risk factors that cause failures in the distribution switch mechanism and the risk indicators contained in each risk factor; S2. Based on the improved fuzzy analytic hierarchy process, a fuzzy judgment matrix formed by its corresponding risk indicator is constructed for each risk factor and assigned a value, and the fuzzy judgment matrix of each risk factor after the assignment is screened to select a fuzzy judgment matrix that meets predetermined conditions in each risk factor, and further obtain an initial weight vector of the selected fuzzy judgment matrix in each risk factor; wherein the improved fuzzy analytic hierarchy process is constructed based on a preset fuzzy analytic hierarchy process, and the fuzzy entropy method is introduced into the fuzzy analytic hierarchy process to perform weighted correction on the fuzzy weights of each fuzzy judgment matrix; S3. Obtain historical data of risk indicators contained in each risk factor, and combine them with the initial weight vector of the fuzzy judgment matrix selected in each risk factor, to obtain a correction amount of the initial weight vector of the fuzzy judgment matrix selected in each risk factor in the improved Bayesian neural network model, and further obtain a correction weight vector of the fuzzy judgment matrix selected in each risk factor based on the initial weight vector of the fuzzy judgment matrix selected in each risk factor; wherein the improved Bayesian neural network model is constructed based on a preset Bayesian neural network model, which introduces historical data and input samples into the input layer of the Bayesian neural network model, and designs a three-layer structure in the hidden layer of the Bayesian neural network model, and introduces predetermined connection weights and activation functions into each layer structure; S4. Calculate the score of each risk factor based on the assignment of the selected fuzzy judgment matrix in each risk factor and its corresponding modified weight vector, and calculate the score of the distribution switch mechanism based on the score of each risk factor to determine the risk level of the distribution switch mechanism.

2. The risk detection method for a power distribution switch mechanism according to claim 1, wherein: The step S2 specifically includes: S21. Based on the risk indicators contained in the same risk factor, the fuzzy judgment matrix corresponding to each risk factor is constructed respectively. The judgment results of the fuzzy membership of the importance between the indicators at the same level are obtained through expert review, and the fuzzy judgment matrix corresponding to each risk factor is assigned a value. Among them, the expression of the fuzzy judgment matrix of a certain risk factor is r ij is the risk indicator x i Relative to the risk indicator x j The importance of fuzzy membership; i, j are the row and column numbers of indicators respectively; S22, through the formula Calculate the fuzzy weight of the fuzzy judgment matrix of each risk factor S23, through the formula Fuzzy weight of the fuzzy judgment matrix for each risk factor Perform weighted fusion to obtain the fusion weight of the fuzzy judgment matrix of each risk factor Among them, λ∈[0,1] is the adjustment parameter; w i ′ is the entropy weight obtained by the fuzzy entropy method, and its expression is S24. Fusion weight of the fuzzy judgment matrix of each risk factor Calculate the maximum eigenvalue λ of the fuzzy judgment matrix of each risk factor max , and by the formula Calculate the consistency index CI of the fuzzy judgment matrix of each risk factor; S25. Find the corresponding RI value of the fuzzy judgment matrix of each risk factor in the preset consistency index value table, and use the formula Calculate the consistency ratio CR of the fuzzy judgment matrix of each risk factor; S26, determining whether the consistency ratio CR of the fuzzy judgment matrix of each risk factor is less than a preset ratio threshold; if yes, proceeding to the next step S27; if not, jumping to step S28; S27, determine the conditions are met and output as the final selected fuzzy judgment matrix, and further calculate the fuzzy judgment matrix selected in each risk factor corresponding to the maximum eigenvalue λ max The initial weight vector Where n is the total number of risk indicators contained in the current risk factor; is the initial weight of the i-th risk indicator of the current risk factor; S28 , after re-assigning the fuzzy judgment matrices whose consistency ratio CR is greater than the ratio threshold, return to step S22 .

3. The risk detection method for a power distribution switch mechanism according to claim 2, wherein: In step S3, the input layer of the improved Bayesian neural network model is based on the input sample X of the input layer of the Bayesian neural network model as the initial weight vector of the fuzzy judgment matrix of each risk factor. Combined with the historical data of risk indicators contained in each risk factor to form a joint input sample Among them, x i is the historical data value of the i-th risk indicator contained in the current risk factor; The hidden layer of the improved Bayesian neural network model is based on the hidden layer of the Bayesian neural network model, and three new hidden layers are constructed; wherein each new hidden layer includes connection weights and ReLU activation function; kernel function K(x i ,x j ) reflects the risk indicator x i and risk indicator x j the degree of correlation between them; The output result of the output layer of the improved Bayesian neural network model is the correction value ΔW=[δw1,δw2,...,δw n ]; where δ is the correction coefficient.

4. The risk detection method for a power distribution switch mechanism according to claim 3, wherein: In step S3, the formula Calculate the modified weight vector of the selected fuzzy judgment matrix in each risk factor 5. The risk detection method for a power distribution switch mechanism according to claim 4, characterized in that: The step S4 specifically includes: After multiplying the value of the selected fuzzy judgment matrix in each risk factor by its corresponding modified weight vector, the accumulated sum is output as the score of each risk factor; After multiplying the weight value pre-assigned to each risk factor by the score of each risk factor, the accumulated sum is output as the score of the distribution switch mechanism; The obtained score of the distribution switch mechanism is compared with a plurality of preset level thresholds respectively, and the risk level of the distribution switch mechanism is obtained according to the comparison result.

6. The risk detection method for a power distribution switch mechanism according to claim 5, characterized in that: The risk factors include equipment status, operating environment, operation and maintenance management status and historical fault status; The risk indicators contained in the equipment status include the number of operations, contact resistance, temperature rise and spring fatigue; The risk indicators of the operating environment include ambient temperature, ambient humidity, ambient pollution level, altitude and corrosive gas concentration; The risk indicators included in the operation and maintenance management status include inspection frequency, maintenance cycle compliance rate, defect closure rate and operation and maintenance record integrity; The risk indicators contained in the historical fault status include the number of faults, the diversity of fault types, the number of tripping events and the aging years of equipment.

7. A distribution switch mechanism risk detection system, characterized in that: include: A risk indicator hierarchical construction unit for determining risk factors causing failures in the distribution switch mechanism and the risk indicators contained in each risk factor; a risk indicator initial weight analysis unit, configured to construct a fuzzy judgment matrix formed by the corresponding risk indicator for each risk factor based on an improved fuzzy analytic hierarchy process and assign values, and to screen the fuzzy judgment matrix of each risk factor after the assignment to select a fuzzy judgment matrix that satisfies a predetermined condition in each risk factor, and further obtain an initial weight vector of the selected fuzzy judgment matrix in each risk factor; wherein the improved fuzzy analytic hierarchy process is constructed based on a preset fuzzy analytic hierarchy process, and the fuzzy entropy method is introduced into the fuzzy analytic hierarchy process to perform weighted correction on the fuzzy weights of each fuzzy judgment matrix; a risk indicator weight correction unit, for obtaining historical data of risk indicators contained in each risk factor, and combining the initial weight vector of the fuzzy judgment matrix selected in each risk factor, to obtain a correction amount of the initial weight vector of the fuzzy judgment matrix selected in each risk factor in the improved Bayesian neural network model, and further obtaining a correction weight vector of the fuzzy judgment matrix selected in each risk factor based on the initial weight vector of the fuzzy judgment matrix selected in each risk factor; wherein the improved Bayesian neural network model is constructed based on a preset Bayesian neural network model, which introduces historical data and input samples into the input layer of the Bayesian neural network model, and designs a three-layer structure in the hidden layer of the Bayesian neural network model and introduces predetermined connection weights and activation functions into each layer structure; The risk level detection unit is used to calculate the score of each risk factor based on the assignment of the fuzzy judgment matrix selected in each risk factor and its corresponding modified weight vector, and calculate the score of the distribution switch mechanism based on the score of each risk factor to determine the risk level of the distribution switch mechanism.

8. The risk detection system for distribution switch mechanism according to claim 7, characterized in that: The risk factors include equipment status, operating environment, operation and maintenance management status and historical fault status; The risk indicators contained in the equipment status include the number of operations, contact resistance, temperature rise and spring fatigue; The risk indicators of the operating environment include ambient temperature, ambient humidity, ambient pollution level, altitude and corrosive gas concentration; The risk indicators included in the operation and maintenance management status include inspection frequency, maintenance cycle compliance rate, defect closure rate and operation and maintenance record integrity; The risk indicators contained in the historical fault status include the number of faults, the diversity of fault types, the number of tripping events and the aging years of equipment.

Images