Power grid equipment risk assessment method based on multi-model fusion

Abstract

The invention relates to a power grid equipment risk assessment method based on multi-model fusion, and the method comprises the specific steps: collecting the historical operation data of power grid equipment, including the risk data of the power grid equipment during long-term operation, the influence data causing the long-term unstable operation of the power grid equipment, and the natural disaster data; a first machine learning model, a second machine learning model and a third machine learning model are constructed, and the first machine learning model and the second machine learning model are trained through risk data existing in long-term operation of the power grid equipment and influence data causing long-term unstable operation of the power grid equipment; and respectively outputting a first risk prediction result and a second risk prediction result through the trained first machine learning model and the second machine learning model.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment management, and in particular to a risk assessment method for power grid equipment based on multi-model fusion. Background Art

[0002] With the continuous expansion of the scale of power grids and the complexity of the equipment operating environment, how to scientifically and reasonably evaluate the risks of power grid equipment and optimize cost investment has become an important issue in the management of power companies. Equipment failure and aging are common problems in power systems. Traditional risk assessment methods focus on the analysis of a single factor of equipment, ignoring the impact of factors such as power grid operation risks, external environmental changes, and natural disasters on equipment performance. The current equipment risk assessment method cannot fully and accurately reflect the potential failure risks of equipment, resulting in certain decision-making blind spots in equipment management and resource allocation for power companies, and they cannot achieve the best cost investment and maintenance strategy.

[0003] At present, the risk assessment technology of power equipment mainly relies on the monitoring of historical equipment failure data, aging status and operating environment, but these methods usually have the following limitations: First, many traditional methods are based only on the operating data of the equipment itself, ignoring the influence of external factors such as grid load fluctuations and frequency changes; second, the assessment models are mostly based on empirical data or simplified statistical models, lacking the ability to comprehensively analyze multiple risk sources; finally, the existing technical solutions usually only consider the single maintenance or replacement decision of the equipment, but fail to comprehensively consider the cost investment, risk prediction and all-round optimization of equipment status. These limitations make the current risk assessment and resource allocation solutions unable to effectively meet the needs of the complex and dynamic operation of modern power grids. Summary of the invention

[0004] In order to solve the above problems existing in the prior art, the present invention proposes a power grid equipment risk assessment method based on multi-model fusion.

[0005] The technical solution of the present invention is as follows:

[0006] On the one hand, the present invention proposes a risk assessment method for power grid equipment based on multi-model fusion, and the specific steps include:

[0007] Collect historical operation data of power grid equipment, including risk data of long-term operation of power grid equipment, impact data of long-term unstable operation of power grid equipment, and natural disaster data;

[0008] Constructing a first machine learning model, a second machine learning model, and a third machine learning model, wherein the first machine learning model and the second machine learning model are respectively trained by risk data existing in the long-term operation of the power grid equipment and impact data causing the long-term unstable operation of the power grid equipment, and the first risk prediction result and the second risk prediction result are respectively outputted by the trained first machine learning model and the second machine learning model;

[0009] By integrating the Bayesian model averaging strategy with the hierarchical hybrid model method, effective information from natural disaster data is extracted, and then the key influencing factors are screened out based on the effective information of natural disaster data using the improved anti-entropy-grey correlation analysis method; a geological disaster risk warning model based on complex weather factors is constructed to predict the probability of geological disasters;

[0010] The third machine learning model is trained by using geological disaster probability data and key influencing factors; and the third risk prediction result is outputted by using the trained third machine learning model;

[0011] The coefficient of variation method is used to assign weights to the first risk prediction result, the second risk prediction result and the third risk prediction result to obtain the final equipment risk assessment result.

[0012] As a preferred implementation, the first machine learning model is a BP neural network model, the BP neural network model is trained using risk data existing in the long-term operation of the power grid equipment itself, and the first risk prediction result is output based on the trained BP neural network model;

[0013] The historical risk data of the long-term operation of the power grid equipment itself includes the equipment operating years, equipment load rate, equipment failure history, equipment historical operation and maintenance investment, and equipment load change frequency.

[0014] As a preferred implementation, the second machine learning model is a support vector machine model, the support vector machine model is trained using the impact data that causes long-term unstable operation of power grid equipment, and the second risk prediction result is output based on the trained support vector machine model;

[0015] The historical data affecting the long-term unstable operation of power grid equipment include voltage, current, load, power fluctuation, equipment status, and fault records.

[0016] As a preferred implementation, the steps of extracting effective information from natural disaster data by integrating the Bayesian model averaging strategy with the hierarchical hybrid model are specifically as follows:

[0017] Define the model space: First, define a model set that includes several mixed regression models, each with different fixed effect terms, random effect structures, or different covariance structures;

[0018] Specify prior distribution: Specify the optimal prior distribution for each model and parameters in the model;

[0019] Calculate the posterior probability: Use Bayes' theorem to calculate the posterior probability of each model through all the data of natural disaster influencing factors;

[0020] Model averaging: weight each model according to its posterior probability, and then take the weighted average of the model's predictions;

[0021] Inference and prediction: using the averaged model results to estimate parameters, test hypotheses, or predict future observations;

[0022] Sensitivity analysis: Analyze the impact of different prior choices on the results.

[0023] As a preferred implementation, the step of screening out key influencing factors from the effective information of natural disaster data using the improved anti-entropy-grey correlation analysis method is specifically as follows:

[0024] Data preprocessing: collecting effective data on power grid failures and natural disaster factors related to power grid failures, and performing necessary data cleaning, standardization or normalization processing;

[0025] Calculate the anti-entropy between the factors affecting natural disasters, and calculate the entropy value of each factor affecting natural disasters:

[0026] Through the anti-entropy formula conversion, the higher the anti-entropy value, the more certain and diverse the information provided by the natural disaster influencing factors, and the higher the weight is given.

[0027] The weights are determined, and the factors affecting natural disasters are weighted based on the anti-entropy value.

[0028] The reference sequence and comparison sequence are constructed, power grid failure is selected as the reference sequence, and various natural disaster influencing factors are selected as the comparison sequence to build the grey system model.

[0029] Calculate the association coefficient and association degree, apply the weighted difference sequence, and calculate the association coefficient between each comparison sequence and the reference sequence.

[0030] The grey correlation degree of each comparison sequence is calculated to obtain a correlation matrix; the natural disaster influencing factors whose correlation degree meets the preset threshold are selected as key influencing factors using the correlation matrix.

[0031] As a preferred implementation, the steps for constructing the geological disaster risk warning model based on complex weather elements are specifically as follows:

[0032] Firstly, by establishing a geological model and conducting numerical simulation, the deformation and movement characteristics of geological disasters are analyzed, and the displacement of geological disasters under different rainfall conditions is obtained as an intensity indicator. Then, historical disasters and previous rainfall conditions are counted, and combined with the results of regional geological disaster susceptibility zoning, a short-term meteorological warning criterion for geological disasters in power grid equipment is established. The geological model is trained using historical disasters and previous rainfall data in the region to obtain a geological disaster risk warning model based on complex weather elements.

[0033] As a preferred implementation, the third neural network prediction model is an improved neural network based on a long short-term memory network combined with a snow melting optimization algorithm. The combination method of the long short-term memory network and the snow melting optimization algorithm is specifically as follows:

[0034] The snow melt optimization algorithm is used to optimize the hyperparameters in the long short-term memory network model. The snow melt optimization algorithm generates molecules at random positions by initializing the positions of molecules, and divides the molecules into two sub-populations using a dual-population mechanism. The number and position of each sub-population are continuously optimized.

[0035] As a preferred implementation, the steps of using the coefficient of variation method to respectively weight the first risk prediction result, the second risk prediction result and the third risk prediction result to obtain the final equipment risk assessment result are specifically as follows:

[0036] R total =W1·R1+W2·R2+W3·R3

[0037] In the formula, R tatal is the final equipment risk assessment result; R1, R2, and R3 are the risk prediction results of the first machine learning model, the second machine learning model, and the third machine learning model respectively;

[0038] On the other hand, the present invention proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a method for risk assessment of power grid equipment based on multi-model fusion as described in any embodiment of the present invention is implemented.

[0039] On the other hand, the present invention proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a risk assessment method for power grid equipment based on multi-model fusion as described in any embodiment of the present invention.

[0040] The present invention has the following beneficial effects:

[0041] 1. The present invention not only takes into account the failure rate and aging condition of the equipment itself, but also integrates multiple risk factors such as grid load fluctuations, environmental changes and disaster impacts, thereby providing a more accurate equipment risk prediction.

[0042] 2. The present invention establishes a dynamic optimization algorithm through the coefficient of variation method, realizes the reasonable allocation of equipment cost investment, ensures that high-risk equipment is repaired and replaced in time, and effectively saves resource investment in low-risk equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0046] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0047] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0048] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0049] Embodiment 1:

[0050] See also Figure 1 , a risk assessment method for power grid equipment based on multi-model fusion, the specific steps include:

[0051] Collect historical operation data of power grid equipment, including risk data of long-term operation of power grid equipment, impact data of long-term unstable operation of power grid equipment, and natural disaster data;

[0052] In this embodiment, the power grid equipment is classified according to the operation risk, and the historical operation data is collected respectively;

[0053] Equipment operation risk: This refers to equipment risks caused by factors such as equipment aging, design defects, and insufficient maintenance, which usually occur during the long-term operation of the equipment.

[0054] Equipment risks caused by grid operation risks: equipment failure or performance degradation caused by unstable operation of the grid itself, such as voltage fluctuations, current fluctuations, load fluctuations and other factors.

[0055] Equipment risk caused by natural disasters: refers to the impact of natural disasters (such as lightning strikes, floods, typhoons, earthquakes, etc.) on equipment, resulting in equipment failure or damage.

[0056] Constructing a first machine learning model, a second machine learning model, and a third machine learning model, wherein the first machine learning model and the second machine learning model are respectively trained by risk data existing in the long-term operation of the power grid equipment and impact data causing the long-term unstable operation of the power grid equipment, and the first risk prediction result and the second risk prediction result are respectively outputted by the trained first machine learning model and the second machine learning model;

[0057] By integrating the Bayesian model averaging strategy with the hierarchical hybrid model method, effective information from natural disaster data is extracted, and then the key influencing factors are screened out based on the effective information of natural disaster data using the improved anti-entropy-grey correlation analysis method; a geological disaster risk warning model based on complex weather factors is constructed to predict the probability of geological disasters;

[0058] The third machine learning model is trained by using geological disaster probability data and key influencing factors; and the third risk prediction result is outputted by using the trained third machine learning model;

[0059] The coefficient of variation method is used to assign weights to the first risk prediction result, the second risk prediction result and the third risk prediction result to obtain the final equipment risk assessment result.

[0060] As a preferred implementation of this embodiment, the first machine learning model is a BP neural network model, the BP neural network model is trained using risk data existing in the long-term operation of the power grid equipment itself, and the first risk prediction result is output based on the trained BP neural network model;

[0061] The historical risk data of the long-term operation of the power grid equipment itself includes the equipment operating years, equipment load rate, equipment failure history, equipment historical operation and maintenance investment, and equipment load change frequency.

[0062] In this embodiment, the system identifies the main indicators that affect the equipment operation risk, including equipment operation years, equipment load rate, equipment failure history, equipment historical operation and maintenance investment, equipment load change frequency, etc. These indicators work together on the reliability and failure rate of the equipment. Through the comprehensive evaluation of these factors, the operation risk of the equipment can be effectively predicted, thereby providing a basis for formulating scientific maintenance and management strategies to ensure the stability and safety of power grid equipment.

[0063] Specific steps for building and training the BP neural network prediction model:

[0064] The BP neural network is trained using the training data set to obtain a BP neural network prediction model; and a prediction module is used to input the real-time data into the BP neural network prediction model to obtain a probability prediction result of the equipment failure.

[0065] The algorithm steps of BP neural network include: initializing the network, selecting random numbers in the interval (-1,1) to assign values ​​to each connection weight of the network, and setting the error and the maximum number of iterations of the network;

[0066] Among them, the parameters of the training network model are specifically:

[0067] The output of the middle hidden layer is:

[0068] y j =f1(net j );

[0069]

[0070] Where j = 1, 2, ..., m,

[0071] The output of the output layer is:

[0072] O k =f2(net k );

[0073]

[0074] Where j = 1, 2, ..., l,

[0075] The transfer functions of functions f1 and f2 are both Sigmoid functions:

[0076]

[0077] Calculate the error function of each layer based on the expected output and the actual output:

[0078]

[0079] For the input layer, the error function is:

[0080]

[0081] Use error back propagation to calculate the error and adjust the network weights:

[0082]

[0083] The error signal is derived as:

[0084]

[0085] By continuously adjusting the weights, the error is continuously reduced:

[0086]

[0087] Calculate the global error:

[0088]

[0089] The algorithm ends when the global error reaches the initial setting accuracy or when the number of training times reaches the maximum number of iterations set initially; otherwise, the error functions of each layer continue to be calculated, the network weights are adjusted, and the global error is calculated.

[0090] As a preferred implementation of this embodiment, the second machine learning model is a support vector machine model, the support vector machine model is trained using the impact data that causes long-term unstable operation of the power grid equipment, and the second risk prediction result is output based on the trained support vector machine model;

[0091] The historical data affecting the long-term unstable operation of power grid equipment include voltage, current, load, power fluctuation, equipment status, and fault records.

[0092] In this embodiment, the construction process of the support vector machine risk prediction model is as follows:

[0093] Given a sample set {(x i ,y i ),i=1,2,…,l}, where x i ∈R n ,y i∈{+1,-1}. The principle of support vector machine for solving nonlinear function optimization is to solve the high-dimensional feature space by mapping the nonlinearity into the high-dimensional space. First, a decision function (classification hyperplane) should be constructed to classify the sample data. The classification hyperplane can maximize the blank area on both sides of the hyperplane while ensuring the classification accuracy, so as to achieve the optimal classification of linearly separable problems. Linear separability means that sample points belonging to different categories are separated by one or several straight lines.

[0094] If there is a classification hyperplane:

[0095] WX+b=0;

[0096] Satisfy the constraints:

[0097] y i (WX+b)-1≥0,i=1,2,…,l;

[0098] The training set is said to be linearly separable. Where W is the weight vector, X is the input vector, and b∈R is the bias.

[0099] Let p represent the distance between the hyperplane and the nearest sample. The purpose of the support vector machine is to find a hyperplane with the largest separation margin, which is the optimal hyperplane.

[0100] Therefore, we need to determine w and b when p is maximum. The prediction function obtained is:

[0101] Y = f(x) = sgn(WX+b);

[0102] Its generalization ability is the best. Among them, sgn(·) is the sign function. That is, the classification error rate of the original sample should be minimized.

[0103] From analytic geometry, we know that the distance from any point in the sample space to the optimal hyperplane is:

[0104]

[0105] Simplified:

[0106] g(x)=r||W0||=W0X+b0;

[0107] Normalize the function so that |g(x0)|=1. This is the support vector, which is closest to the classification decision surface and plays a leading role in SVM.

[0108] So the algebraic distance from the support vector to the optimal hyperplane is

[0109]

[0110] Therefore, it can be considered that maximizing the separation margin is equivalent to minimizing the norm of the weight vector ‖W0‖. Therefore, the classification hyperplane that minimizes ‖W0‖ is the optimal hyperplane.

[0111] The classification machine is an algorithm that transforms the optimization problem of solving the optimal classification surface by the maximum margin method into its dual problem, thereby solving the original classification problem by solving the relatively simple dual problem.

[0112] Using the Lagrange optimization method and Wofle duality theory, the above problem can be transformed into its dual problem, that is, maximizing the universal function:

[0113]

[0114] Among them, α i is the Lagrange multiplier corresponding to sample i.

[0115] Solving the above problem, the optimal classification function is:

[0116]

[0117] Among them, nsv is the number of support vectors, and b is the classification threshold, which can be obtained by taking the median of any pair of support vectors in the two categories.

[0118] By introducing the relaxation term ξ i It can realize the generalized classification surface and solve the problem that the training samples are linearly inseparable, that is, it compromises the minimum misclassified samples and the maximum classification interval. For nonlinear problems, it can be transformed into a linear problem in a high-dimensional space through nonlinear transformation, and then the optimal classification surface is sought in this high-dimensional space. Only the inner product operation between samples (x i ·x j ) is involved, so only the inner product operation needs to be performed in the high-dimensional space, and the inner product operation can be realized by the function in the original space. According to the Hilbert-Schmidt principle, as long as the kernel function K(x i ·x j ) satisfies Mercer's condition, which corresponds to the inner product in a certain exchange space.

[0119] Therefore, the kernel function K(x i ·x j ) instead of the inner product in the formula, we can realize linear classification after some nonlinear transformation, and the optimal classification function becomes:

[0120]

[0121] In short, SVM maps the input vector to a high-dimensional feature space through some pre-selected nonlinear mapping, and constructs the optimal classification hyperplane in this feature space. In form, the SVM classification function is similar to a neural network, and the output is a linear combination of intermediate nodes, and each intermediate node corresponds to a support vector, as shown in the figure below. The number of intermediate nodes in a neural network must be selected based on experience or comparative experiments, and different logics will have a great impact on network performance; while the number of intermediate nodes of SVM is automatically determined by calculation.

[0122] Vapnik generalized the results obtained in SVM classification theory by introducing the ε-insensitive loss function, making it applicable to function fitting. With this generalization, the SVM method becomes a general method for expressing functions in high-dimensional space. It can be used for various function estimation problems, including density estimation problems and solving linear operator equations. i and The following optimization problem is constructed:

[0123]

[0124] Among them, the constant C is the penalty coefficient.

[0125] The optimization problem in the dual space is:

[0126]

[0127] Among them, α i , is the Lagrange multiplier.

[0128] Solve this problem and get the optimal Lagrange multiplier α i , So we get the fitting function:

[0129]

[0130] As a preferred implementation of this embodiment, the steps of extracting effective information from natural disaster data by integrating the Bayesian model averaging strategy and the hierarchical hybrid model are specifically as follows:

[0131] Define the model space: First, define a model set that includes several mixed regression models, each with different fixed effect terms, random effect structures, or different covariance structures;

[0132] Specify prior distribution: Specify the optimal prior distribution for each model and parameters in the model;

[0133] Calculate the posterior probability: Use Bayes' theorem to calculate the posterior probability of each model through all the data of natural disaster influencing factors;

[0134] Model averaging: weight each model according to its posterior probability, and then take the weighted average of the model's predictions;

[0135] Inference and prediction: using the averaged model results to estimate parameters, test hypotheses, or predict future observations;

[0136] Sensitivity analysis: Analyze the impact of different prior choices on the results.

[0137] In this embodiment, Bayesian Model Averaging (BMA) is a statistical method that handles model uncertainty by ensembling multiple models on a given data set. Instead of selecting a single best model, BMA assigns a posterior probability to each model and then weights the model's predictions based on these probabilities to obtain the final prediction or inference result. This approach recognizes that in the real world, due to data limitations or model simplifications, no single model can perfectly capture the process of generating all data.

[0138] Mixed regression models (such as mixed linear models, mixed effects models, etc.) are a collection of various types of regression models that can be used to capture complex structures in the data, such as the combination of fixed effects and random effects, or heterogeneity in the data. Mixed models are particularly suitable for data analysis with hierarchical structures or clustering effects.

[0139] To combine Bayesian model averaging with a mixture regression model, you can take the following steps:

[0140] Define the model space: First, identify your model set, which may include different mixed regression models, each with different fixed effects terms, random effects structures, or different covariance structures.

[0141] Specify prior distributions: Specify appropriate prior distributions for each model and parameter in the model. This step is critical to Bayesian analysis because it encodes our beliefs about the model parameters before we see the data.

[0142] Calculate the posterior probability: Using Bayes' theorem, calculate the posterior probability of each model using the data. This involves calculating the likelihood function and prior probability of the model, as well as a normalization constant to ensure that the sum of the posterior probabilities of all models is 1.

[0143] Model Averaging: Each model is weighted according to its posterior probability, and then the predictions of the models are weighted averaged. This means that the more likely models will contribute more to their final predictions.

[0144] Inference and prediction: Use the averaged model results to estimate parameters, test hypotheses, or predict future observations. This step takes into account all information about model uncertainty and is usually more robust than predictions based on a single model.

[0145] Sensitivity analysis: Analyze the impact of different prior choices on the results to ensure the robustness of the results.

[0146] As a preferred implementation of this embodiment, the step of screening out key influencing factors from the effective information of natural disaster data using the improved anti-entropy-grey correlation analysis method is specifically as follows:

[0147] Data preprocessing: collecting effective data on power grid failures and natural disaster factors related to power grid failures, and performing necessary data cleaning, standardization or normalization processing;

[0148] Calculate the anti-entropy between the factors affecting natural disasters, and calculate the entropy value of each factor affecting natural disasters:

[0149] Through the anti-entropy formula conversion, the higher the anti-entropy value, the more certain and diverse the information provided by the natural disaster influencing factors, and the higher the weight is given.

[0150] The weights are determined, and the factors affecting natural disasters are weighted based on the anti-entropy value.

[0151] The reference sequence and comparison sequence are constructed, power grid failure is selected as the reference sequence, and various natural disaster influencing factors are selected as the comparison sequence to build the grey system model.

[0152] Calculate the association coefficient and association degree, apply the weighted difference sequence, and calculate the association coefficient between each comparison sequence and the reference sequence.

[0153] The grey correlation degree of each comparison sequence is calculated to obtain a correlation matrix; the natural disaster influencing factors whose correlation degree meets the preset threshold are selected as key influencing factors using the correlation matrix.

[0154] In this embodiment, an improved anti-entropy-grey correlation analysis method is introduced to quantitatively analyze the correlation between equipment failure and natural disaster indicators. The improved anti-entropy-grey correlation analysis method is a method that aims to improve the accuracy and objectivity of traditional grey correlation analysis, especially in the scenario of dealing with the correlation analysis between equipment failure and its multiple indicators. The following is an overview of the basic concepts of the method and its implementation steps:

[0155] Reasons for the introduction of anti-entropy thinking

[0156] Enhanced objectivity: Traditional grey correlation analysis uses equal weighting, that is, all indicators are considered equally important, which may lead to the neglect of information on key influencing factors. The anti-entropy method quantifies the amount of information on influencing factors and automatically adjusts the weights of influencing factors, so that the analysis results can better reflect the actual situation.

[0157] Revealing intrinsic associations: Anti-entropy can highlight those influencing factors with large variations and rich information, thereby giving them more attention in association analysis and helping to discover more subtle association patterns.

[0158] Reduce bias: When the correlation between influencing factors is complex and scattered, anti-entropy-grey correlation analysis can effectively avoid the analysis bias caused by local correlation and improve the overall accuracy of the analysis.

[0159] In this embodiment, the basic steps of improving the anti-entropy-grey correlation analysis method are as follows:

[0160] Data preprocessing: collecting effective data on power grid failures and natural disaster factors related to power grid failures, and performing necessary data cleaning, standardization or normalization processing;

[0161] Calculate the anti-entropy between the factors affecting natural disasters, and calculate the entropy value of each factor affecting natural disasters:

[0162] Through the anti-entropy formula conversion, the higher the anti-entropy value, the more certain and diverse the information provided by the natural disaster influencing factors, and the higher the weight is given.

[0163] The weights are determined, and the factors affecting natural disasters are weighted based on the anti-entropy value.

[0164] The reference sequence and comparison sequence are constructed, power grid failure is selected as the reference sequence, and various natural disaster influencing factors are selected as the comparison sequence to build the grey system model.

[0165] Calculate the association coefficient and association degree, apply the weighted difference sequence, and calculate the association coefficient between each comparison sequence and the reference sequence.

[0166] The grey correlation degree of each comparison sequence is calculated to obtain a correlation matrix; the natural disaster influencing factors whose correlation degree meets the preset threshold are selected as key influencing factors using the correlation matrix.

[0167] In order to avoid the problem of weight imbalance caused by the excessive sensitivity of influencing factors in the entropy weight method, the anti-entropy value method is used to objectively weight the influencing factors. The specific steps are as follows.

[0168] 1) Construct a matrix for evaluating the impact factors of natural disasters. Select n natural disaster impact factors, k regions, and m years of data (replace letters), and establish the corresponding initial matrix, x ijis the value of the jth natural disaster influencing factor of the ith object, and the matrix X of natural disaster influencing factors is evaluated. m for

[0169]

[0170] 2) Calculate the anti-entropy of each natural disaster influencing factor. Define the anti-entropy h′ as

[0171]

[0172] Where: p j is the entropy value of the jth natural disaster influencing factor, 0≤p j ≤1, and

[0173] The anti-entropy h′ of each natural disaster influencing factor is:

[0174]

[0175] Where: c ij is the entropy value of the jth natural disaster influencing factor under the i-th object,

[0176] 3) Normalization. The anti-entropy of each natural disaster influencing factor is further normalized to obtain:

[0177]

[0178] Where: d j is the objective weight of the jth natural disaster influencing factor.

[0179] In order to scientifically and reasonably evaluate the impact of various natural disaster factors on distribution network failures, the natural disaster factors are combined and weighted as follows:

[0180]

[0181] Where: w j is the combined weight for evaluating the impact factor j of natural disasters; c j ,d j are the objective weight and subjective weight of the jth natural disaster influencing factor. (Objective weight calculated based on entropy value)

[0182] The fuzzy mathematics comprehensive evaluation method is adopted and the ascending semi-trapezoidal distribution function is used as the membership function to fit and score the various natural disaster influencing factors.

[0183] In order to further quantify the correlation between faults and their natural disaster influencing factors, the improved anti-entropy-grey correlation analysis method is used to determine the correlation level between the two. Traditional grey correlation analysis establishes a correlation matrix by obtaining a differential sequence, and generally adopts equal weighting. This equal weighting method easily conceals the characteristic relationship between various natural disaster influencing factors and ignores the rich information contained in the natural disaster influencing factors themselves. Since the influencing factors of natural disasters come from multiple aspects and levels, when the correlation relationship is relatively discrete, a trend of local correlation is likely to occur, resulting in analysis deviations. Therefore, anti-entropy is introduced on the basis of the entropy method to overcome the existing subjective problems. The specific steps are as follows. The calculation formula for the correlation between sequences in the traditional grey correlation analysis method is:

[0184]

[0185] Where: γ j is the correlation between the jth natural disaster influencing factor and other natural disaster influencing factors; ζ rj In order to characterize the correlation coefficient between the natural disaster influencing factor j and other natural disaster influencing factors r, the fixed weight value is 1 / n.

[0186] The residual degree of the jth natural disaster influencing factor is defined as

[0187] T j =1-e j ;

[0188] Where: T j is the residual degree of the jth natural disaster influencing factor; e j is the ratio of the anti-entropy of the jth natural disaster influencing factor to the maximum entropy, that is, the relative entropy. The specific expression is:

[0189]

[0190] Where: E j 、E max are the anti-entropy and maximum entropy of the jth natural disaster influencing factor. j The larger the value is, the more obvious the characteristics of the natural disaster influencing factors are and the higher the importance is.

[0191] The proportional weight of other natural disaster influencing factors r is set to b(r), and the specific calculation formula is:

[0192]

[0193] As a preferred implementation of this embodiment, the steps for constructing the geological disaster risk warning model based on complex weather elements are specifically as follows:

[0194] Firstly, by establishing a geological model and conducting numerical simulation, the deformation and movement characteristics of geological disasters are analyzed, and the displacement of geological disasters under different rainfall conditions is obtained as an intensity indicator. Then, historical disasters and previous rainfall conditions are counted, and combined with the results of regional geological disaster susceptibility zoning, a short-term meteorological warning criterion for geological disasters in power grid equipment is established. The geological model is trained using historical disasters and previous rainfall data in the region to obtain a geological disaster risk warning model based on complex weather elements.

[0195] In this embodiment, the third machine learning model takes the probability of geological disasters and key influencing indicators as input parameters to construct a power grid fault prediction model based on an improved neural network.

[0196] The duration of the rainfall event that induces geological disasters is recorded as D, the cumulative rainfall during this period is recorded as E, and the ratio of E to D is the average rainfall intensity. The principle is to depict the previous rainfall E and the duration condition D of the geological disaster in double logarithmic coordinates, and carry out fitting analysis on the data points of the rainfall event that induces geological disasters.

[0197] E=c+α×D β ;

[0198] E is the cumulative rainfall (mm); D is the rainfall duration, α, β, c are statistical parameters.

[0199] However, the process of rainfall events inducing geological disasters is not uniform. It is generally believed that only the rainfall on the day of the geological disaster has a full effect on the occurrence of geological disasters, and the rest is considered to be only partially effective for geological disasters. Therefore, effective rainfall is usually used to characterize the threshold index of rainfall-induced geological disasters. This patent mainly uses the power exponential formula to calculate the corresponding effective rainfall.

[0200] R c =R0+αR1+α 2 R2+...+α n R n ;

[0201] R c is the effective rainfall, R0 is the rainfall on the day of the geological disaster, and R n is the rainfall n days before the disaster. α is the attenuation coefficient, and n is the number of days.

[0202] The risk calculation method for geological disasters at the base of transmission line towers is as follows.

[0203] R = H × V;

[0204] Among them, R is the geological disaster risk, H is the hazard probability analysis, and V is the vulnerability analysis.

[0205] Based on Bayesian theory, H is obtained by combining the ED rainfall threshold model and the rainfall threshold exceedance probability calculation. The time probability of rainfall-induced geological disasters is divided into two parts, namely the probability of a certain rainfall condition occurring and the probability of geological disasters occurring under the rainfall condition. First, the rainfall corresponding to different extreme value recurrence conditions in a certain area is determined by statistical methods; then, the rainfall threshold for inducing geological disasters in each local area is obtained by statistical analysis, and the probability of geological disasters occurring under rainfall conditions exceeding the threshold is analyzed by statistical analysis of historical geological disaster events; the time probability of geological disasters is obtained by combining the two.

[0206] Assume that the rainfall event is A and the geological disaster event is B, P[A] and P[B|A] are the probability of the rainfall event and the geological disaster under the rainfall conditions, respectively. The time probability of rainfall-induced geological disasters in Bayesian conditional probability calculation is P(AB)

[0207] It is expressed as:

[0208] P(AB)=P[A]·P[B|A];

[0209] In the exceedance probability calculation determined based on the ED effective rainfall threshold model, after obtaining the geological disaster threshold of each research unit, combined with its historical rainfall records, the number of rainfall events that exceeded the threshold rainfall conditions in history, as well as the number of geological disasters induced by such events, are analyzed. The two can be combined to obtain the probability of geological disasters P[B|A]. The probability distribution of geological disasters over the same long period of time can be evaluated using binomial distribution and Poisson distribution. When the time length is a discrete time interval of one year, the average annual probability of geological disasters is related to the time period T of the occurrence of geological disasters under this condition. The annual probability of geological disasters can be expressed as

[0210]

[0211] T is the occurrence period of geological disaster events, and λ is the frequency of geological disasters. This method is also applicable to calculating the probability of geological disasters in a given time period. Let P0 be the probability of no geological disasters occurring in a given time period (within 1 year). Then P 0t is the probability that no geological disaster occurs within t years, so the probability of geological disasters occurring in many years can be expressed as:

[0212]

[0213] Where e is a natural constant and λ is the frequency of geological disasters under certain rainfall conditions. Therefore, the probability of one or more geological disasters occurring in the next t years is:

[0214] P = 1-e-λt;

[0215] The risk probability V of qualified disasters is mainly calculated through the Monte Carlo simulation method. Based on the statistical sampling theory, the Monte Carlo method studies the numerical calculation method of generating random variables through computers, which is also called random simulation method or statistical experimental method.

[0216] First, establish a function to represent the stable state of geological hazards:

[0217]

[0218] Each parameter is a random variable that controls the stability of geological disasters, which is determined by the rock and soil properties of geological disasters, deformation mechanism and other factors. The above parameters are defined as rainfall, single maximum rainfall, rainfall duration, rock and soil type. It is generally believed that these parameters have a certain range of values, and they are random within the range of values. The reasons for this randomness include the inhomogeneity of geological bodies, changes in external environmental factors, measurement errors, etc.

[0219] Fs represents the stability coefficient of geological disasters. When Fs≤1, geological disasters are considered to be in an unstable state. In N groups of random numbers, there are M times when Fs≤1 occurs. When N is large enough, it is considered that the frequency of Fs≤1 is close to the probability. The dangerous probability of geological disasters is expressed as follows:

[0220]

[0221] When N is large enough, its mean μ F and standard deviation σ F They are:

[0222]

[0223] The risk probability of geological disasters is:

[0224] P f =1-Φ(β).

[0225] As a preferred implementation of this embodiment, the third neural network prediction model is an improved neural network based on a long short-term memory network combined with a snow melting optimization algorithm. The combination method of the long short-term memory network and the snow melting optimization algorithm is specifically as follows:

[0226] The snow melt optimization algorithm is used to optimize the hyperparameters in the long short-term memory network model. The snow melt optimization algorithm generates molecules at random positions by initializing the positions of molecules, and divides the molecules into two sub-populations using a dual-population mechanism. The number and position of each sub-population are continuously optimized.

[0227] In this embodiment, the traditional neural network has the problem of gradient vanishing due to the input at any previous time, which makes it difficult for the neural network to converge and the accuracy is not high. The LSTM neural network built in this paper is designed to effectively transmit and express information in long time series, and will not cause useful information from a long time ago to be forgotten.

[0228] The input of the forget gate is the hidden state h of the previous moment t-1 and the current input x t , output f t Used to control which information in memory cells should be forgotten at a certain moment, f is the bias vector of the forget gate. The relationship between input and output is as follows:

[0229] f t =σ(W f [h t-1 ,x t ]+b f );

[0230] In the LSTM neural network, the input of the input gate is the hidden state h of the previous moment. t-1 and the current input x t , output i t Used to determine the information in the current input that should be added to the memory cell, b i is the bias vector of the input gate. The relationship between its input and output is:

[0231] i t =σ(W i [h t-1 ,x t ]+b i );

[0232] In addition, the input of the memory cell is the hidden state h of the previous moment t-1 and the current input x t , the output is the current cell C t Used to update memory cells, b c is the bias vector of the output gate.

[0233]

[0234] Where:

[0235] ——External output information of the memory unit module at time t.

[0236] When using the SAO algorithm to optimize LSTM parameters, the algorithm is first initialized to determine the number of molecule clusters N and the dimension of the problem dim. Then, a batch of N rows and dim columns of molecules are randomly generated.

[0237]

[0238] Where:

[0239] Z——position information matrix of all molecules;

[0240] Z min , Z max ——the upper and lower limits of the molecular position;

[0241] Z rand ——Random numbers, used to randomly generate different molecular positions;

[0242] Z 1,1 ——The position of the first molecule in the first dimension.

[0243] The snow melting optimization algorithm used in this paper uses a dual population mechanism to divide the molecular population P into two sub-populations P a and P b The number of molecules in the population is N a and N b , the expression is:

[0244]

[0245] N b (t+1)=N b (t)-1.

[0246] As a preferred implementation of this embodiment, the steps of using the coefficient of variation method to respectively weight the first risk prediction result, the second risk prediction result and the third risk prediction result to obtain the final equipment risk assessment result are specifically as follows:

[0247] R total =W1·R1+W2·R2+W3·R3

[0248] In the formula, R total is the final equipment risk assessment result; R1, R2, and R3 are the risk prediction results of the first machine learning model, the second machine learning model, and the third machine learning model respectively;

[0249] In this embodiment, the coefficient of variation method is a statistical indicator commonly used in statistics to measure data differences. This method weights each indicator according to the degree of variation of its observed values ​​on all evaluated objects. The basic principle of the coefficient of variation method is that the greater the degree of variation, the greater the impact of the indicator on the comprehensive evaluation, and the weight reflects the size of the indicator's ability to distinguish. However, it cannot reflect the independence of the indicator and the evaluator's understanding of the importance of the indicator, and lacks comprehensiveness and scientificity.

[0250] The specific method is as follows:

[0251] (1) Weight calculation

[0252] Calculate the standard deviation σ of each indicator k , represents the absolute variation degree of each indicator, and its calculation formula is:

[0253]

[0254] Calculate the coefficient of variation c of each indicator k , which reflects the relative variation of each indicator, and its calculation formula is:

[0255]

[0256] The coefficient of variation of each indicator is normalized to obtain the weight of each indicator:

[0257]

[0258] Then the weight vector of each indicator is obtained as

[0259] w”=(w″1,w″2,…,w″ m );

[0260] (2) Comprehensive risk settlement

[0261] The comprehensive risk is a weighted average based on the predicted results of each risk and the corresponding weight. The predicted result of each risk can be a probability value. The calculation formula for the comprehensive risk is:

[0262] R total =W1·R1+W2·R2+W3·R3;

[0263] In the formula, R total is the final equipment risk assessment result; R1, R2, and R3 are the risk prediction results of the first machine learning model, the second machine learning model, and the third machine learning model respectively;

[0264] (3) Invest in scientific evaluation

[0265] Based on the comprehensive risk assessment results, the power company can decide to make reasonable maintenance and replacement investments for the equipment. Generally, higher-risk equipment requires more attention and investment to reduce the probability of failure and ensure the safety and stability of the power grid.

[0266] High-risk equipment: Equipment with higher comprehensive risk assessment results requires more maintenance, inspection, care or replacement to avoid sudden equipment failures and ensure the stable operation of the power grid.

[0267] Low-risk equipment: For equipment with a low comprehensive risk assessment result, maintenance investment can be appropriately reduced, and upgrades can be delayed to save costs.

[0268] By using the coefficient of variation method to calculate risk weights and combining them with the comprehensive risk assessment results, we can scientifically assess the risk of equipment failure and determine reasonable cost investment accordingly. High-risk equipment requires more investment to prevent failures, while the investment in low-risk equipment can be moderately reduced, thereby achieving optimal resource allocation and improving the safety and economic benefits of the power grid.

[0269] Embodiment 2:

[0270] A power grid equipment risk assessment system based on multi-model fusion, comprising:

[0271] The data collection module collects historical operation data of power grid equipment, including risk data of long-term operation of power grid equipment, impact data of long-term unstable operation of power grid equipment, and natural disaster data;

[0272] The self-risk prediction module builds a BP neural network risk prediction model, uses the historical data of risks that exist in the long-term operation of power grid equipment to train the BP neural network, obtains the BP neural network risk prediction model, and predicts the risks that exist in the long-term operation of power grid equipment itself;

[0273] The long-term unstable operation risk prediction module builds a support vector machine risk prediction model. After cleaning and standardizing the historical data that affects the long-term unstable operation of power grid equipment, it inputs the support vector machine model for training, selects the optimal kernel function and adjusts the hyperparameters; and uses the trained support vector machine risk prediction model to predict the risks caused by the long-term unstable operation of power grid equipment.

[0274] The natural disaster risk prediction module, by integrating the Bayesian model averaging strategy and the hierarchical hybrid model method, selects the key influencing factors from all the natural disaster influencing factors that affect the stable operation of power grid equipment; constructs a geological disaster risk warning model based on complex weather elements to predict the probability of geological disasters, and constructs an improved neural network equipment failure prediction model based on the long short-term memory network combined with the snow melting optimization algorithm. The improved neural network equipment failure prediction model is trained using natural disaster probability data and key influencing factor data; the trained improved neural network equipment failure prediction model is used to predict the risks of power grid equipment caused by natural disasters;

[0275] The comprehensive risk prediction module uses the coefficient of variation method to assign weights to the BP neural network risk prediction model, the support vector machine risk prediction model and the improved neural network equipment failure prediction model to comprehensively obtain the final equipment risk assessment results.

[0276] Embodiment three:

[0277] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, a method for risk assessment of power grid equipment based on multi-model fusion as described in any embodiment of the present invention is implemented.

[0278] Embodiment 4:

[0279] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements a risk assessment method for power grid equipment based on multi-model fusion as described in any embodiment of the present invention.

[0280] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A risk assessment method for power grid equipment based on multi-model fusion, characterized in that: The specific steps include: Collect historical operation data of power grid equipment, including risk data of long-term operation of power grid equipment, impact data of long-term unstable operation of power grid equipment, and natural disaster data; Constructing a first machine learning model, a second machine learning model, and a third machine learning model, wherein the first machine learning model and the second machine learning model are respectively trained by risk data existing in the long-term operation of the power grid equipment and impact data causing the long-term unstable operation of the power grid equipment, and the first risk prediction result and the second risk prediction result are respectively outputted by the trained first machine learning model and the second machine learning model; By integrating the Bayesian model averaging strategy with the hierarchical hybrid model method, effective information from natural disaster data is extracted, and then the key influencing factors are screened out based on the effective information of natural disaster data using the improved anti-entropy-grey correlation analysis method; a geological disaster risk warning model based on complex weather factors is constructed to predict the probability of geological disasters; The third machine learning model is trained by using geological disaster probability data and key influencing factors; and the third risk prediction result is outputted using the trained third machine learning model; The coefficient of variation method is used to assign weights to the first risk prediction result, the second risk prediction result and the third risk prediction result to obtain the final equipment risk assessment result.

2. A method for risk assessment of power grid equipment based on multi-model fusion according to claim 1, characterized in that: The first machine learning model is a BP neural network model, which uses the risk data of the long-term operation of the power grid equipment to train the BP neural network model, and outputs a first risk prediction result based on the trained BP neural network model; The historical risk data of the long-term operation of the power grid equipment itself includes the equipment operating years, equipment load rate, equipment failure history, equipment historical operation and maintenance investment, and equipment load change frequency.

3. A method for risk assessment of power grid equipment based on multi-model fusion according to claim 1, characterized in that: The second machine learning model is a support vector machine model, which is trained using the impact data that causes long-term unstable operation of power grid equipment, and outputs a second risk prediction result based on the trained support vector machine model; The historical data affecting the long-term unstable operation of power grid equipment include voltage, current, load, power fluctuation, equipment status, and fault records.

4. The method for risk assessment of power grid equipment based on multi-model fusion according to claim 1 is characterized in that: The steps of extracting effective information from natural disaster data by integrating the Bayesian model averaging strategy and the hierarchical hybrid model are specifically as follows: Define the model space: First, define a model set that includes several mixed regression models, each with different fixed effect terms, random effect structures, or different covariance structures; Specify prior distribution: Specify the optimal prior distribution for each model and parameters in the model; Calculate the posterior probability: Use Bayes' theorem to calculate the posterior probability of each model through all the data of natural disaster influencing factors; Model averaging: weight each model according to its posterior probability, and then take the weighted average of the model's predictions; Inference and prediction: using the averaged model results to estimate parameters, test hypotheses, or predict future observations; Sensitivity analysis: Analyze the impact of different prior choices on the results.

5. The method for risk assessment of power grid equipment based on multi-model fusion according to claim 1 is characterized in that: The steps of screening out key influencing factors from the effective information of natural disaster data using the improved anti-entropy-grey correlation analysis method are as follows: Data preprocessing: collecting effective data on power grid failures and natural disaster factors related to power grid failures, and performing necessary data cleaning, standardization or normalization processing; Calculate the anti-entropy between the factors affecting natural disasters, and calculate the entropy value of each factor affecting natural disasters: Through the anti-entropy formula conversion, the higher the anti-entropy value, the more certain and diverse the information provided by the natural disaster influencing factors, and the higher the weight is given. The weights are determined, and the factors affecting natural disasters are weighted based on the anti-entropy value. The reference sequence and comparison sequence are constructed, power grid failure is selected as the reference sequence, and various natural disaster influencing factors are selected as the comparison sequence to build the grey system model. Calculate the association coefficient and association degree, apply the weighted difference sequence, and calculate the association coefficient between each comparison sequence and the reference sequence. The grey correlation degree of each comparison sequence is calculated to obtain a correlation matrix; the natural disaster influencing factors whose correlation degree meets the preset threshold are selected as key influencing factors using the correlation matrix.

6. A method for risk assessment of power grid equipment based on multi-model fusion according to claim 1, characterized in that: The steps for constructing the geological disaster risk warning model based on complex weather factors are specifically as follows: Firstly, by establishing a geological model and conducting numerical simulation, the deformation and movement characteristics of geological disasters are analyzed, and the displacement of geological disasters under different rainfall conditions is obtained as an intensity indicator. Then, historical disasters and previous rainfall conditions are counted, and combined with the results of regional geological disaster susceptibility zoning, a short-term meteorological warning criterion for geological disasters in power grid equipment is established. The geological model is trained using historical disasters and previous rainfall data in the region to obtain a geological disaster risk warning model based on complex weather elements.

7. A method for risk assessment of power grid equipment based on multi-model fusion according to claim 1, characterized in that: The third neural network prediction model is an improved neural network based on a long short-term memory network combined with a snow melting optimization algorithm. The combination method of the long short-term memory network and the snow melting optimization algorithm is specifically as follows: The snow melt optimization algorithm is used to optimize the hyperparameters in the long short-term memory network model. The snow melt optimization algorithm generates molecules at random positions by initializing the positions of molecules, and divides the molecules into two sub-populations using a dual-population mechanism. The number and position of each sub-population are continuously optimized.

8. The method for risk assessment of power grid equipment based on multi-model fusion according to claim 1 is characterized in that: The steps of using the coefficient of variation method to respectively weight the first risk prediction result, the second risk prediction result and the third risk prediction result to obtain the final equipment risk assessment result are specifically as follows: R total =W1·R1+W2·R2+W3·R3 In the formula, R total is the final equipment risk assessment result; R1, R2, and R3 are the risk prediction results of the first machine learning model, the second machine learning model, and the third machine learning model, respectively.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, a power grid equipment risk assessment method based on multi-model fusion as described in any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a risk assessment method for power grid equipment based on multi-model fusion as described in any one of claims 1 to 8 is implemented.

Images