A method and device for predicting power system faults taking environmental factors into consideration

Abstract

The embodiment of the present invention discloses a method and device for predicting faults in an electric power system taking environmental factors into consideration. The method comprises: based on historical meteorological data of the electric power system and historical fault data corresponding to the historical meteorological data, analyzing meteorological factors affecting fault prediction accuracy at different altitudes, mathematically modeling the meteorological factors to obtain weather correction factors and altitude correction factors, and constructing a component fault-meteorological probability model; performing regression analysis based on the component fault-meteorological probability model to construct a component fault-meteorological association model; obtaining a sample training set based on the component fault-meteorological association model; training the model using the sample training set through a reinforcement learning method to obtain a fault prediction model; and finally inputting real-time data into the fault prediction model, thereby realizing the prediction of the probability of faults in the electric power system by comprehensively considering different altitudes and meteorological conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid system fault prediction, and in particular to a method and device for predicting power grid system faults taking environmental factors into consideration. Background Art

[0002] With the needs of economic development, new power systems have gradually developed. In order to ensure the safe and stable operation of new power systems, the problems of fault diagnosis and fault prediction cannot be ignored. At present, the research on fault prediction at home and abroad mainly predicts the future operation status of power systems based on the obtained operation data. However, since different geographical meteorology will affect the safe and stable operation of the power grid, it is necessary to study the fault prediction considering different altitudes and meteorological conditions. Summary of the invention

[0003] The main purpose of the present invention is to provide a method and device for predicting power system faults taking environmental factors into consideration, which can solve the problem in the prior art that different altitudes and meteorological conditions are not taken into consideration when predicting power system faults.

[0004] To achieve the above object, the present invention provides a first aspect of a power system fault prediction method taking environmental factors into consideration, the method comprising:

[0005] Acquire historical meteorological data of the power system and historical fault data corresponding to the historical meteorological data, and analyze meteorological factors that affect the accuracy of power system fault prediction at different altitudes based on the historical meteorological data and the historical fault data;

[0006] Mathematically modeling the meteorological factors to obtain a weather correction factor and an altitude correction factor, wherein the weather correction factor is the correlation between different weather conditions and fault prediction accuracy, and the altitude correction factor is the correlation between different altitudes and fault prediction accuracy;

[0007] Construct a component failure-meteorological probability model based on weather correction factors and altitude correction factors;

[0008] Perform regression analysis based on the component failure-weather probability model to construct a component failure-weather correlation model;

[0009] A sample training set is obtained based on a component failure-weather correlation model, and a model is trained using the sample training set by a reinforcement learning method to obtain a fault prediction model;

[0010] The real-time data is input into the fault prediction model to predict the probability of a fault in the power system.

[0011] In the present technical scheme, the meteorological factors that affect the fault prediction accuracy of the power system at different altitudes are analyzed according to the historical meteorological data and the historical fault data corresponding to the historical meteorological data, and the meteorological factors are mathematically modeled to obtain the weather correction factor and the altitude correction factor. After the weather correction factor and the altitude correction factor are obtained, a component fault-meteorological correlation model is established according to the weather correction factor and the altitude correction factor. A sample training set is obtained based on the component fault-meteorological correlation model, and the model is trained using the sample training set by the reinforcement learning method to obtain a fault prediction model. By analyzing the meteorological factors that affect the fault prediction accuracy at different altitudes from the historical meteorological data of the power system and the historical fault data corresponding to the historical meteorological data, the meteorological factors are mathematically modeled to obtain the weather correction factor and the altitude correction factor, so that the component fault-meteorological correlation model obtained according to the weather correction factor and the altitude correction factor can combine different altitudes and meteorological conditions to obtain fault data, so as to realize the fault prediction model obtained by training the model using the sample training set, and can consider different altitudes and meteorological conditions to predict the probability of failure.

[0012] In combination with the first aspect, in one possible implementation, the formula of the weather correction factor is as follows:

[0013]

[0014] Among them, T mj represents the weather correction factor, is the gravitational coefficient of the weather before the fault on the fault prediction, is the gravitational coefficient of the weather after the fault on the fault prediction, G mj is the weather state coefficient, d is the duty cycle, U mj is the voltage rating;

[0015] The formula for the altitude correction factor is as follows:

[0016]

[0017] Among them, Γ H is the altitude correction factor, f is the power frequency, and H is the altitude;

[0018] The formula of the component failure-meteorological probability model is as follows:

[0019]

[0020] Among them, λ mj is the total number of failures of the mth component in a cycle, λ avg is the average number of failures within a component cycle, P jis the percentage of the duration of weather in state j in the statistical period. The weather is divided into three states: normal, severe, and disaster. H is the altitude correction factor, T mj is the weather correction factor.

[0021] In combination with the first aspect, in one possible implementation method, the above-mentioned regression analysis is performed based on the component failure-meteorological probability model to construct a component failure-meteorological correlation model, including: establishing an overall regression model of the number of failures based on the component failure-meteorological probability model, and using the least squares method to perform parameter estimation on the regression coefficient of the overall regression model of the number of failures to obtain a partial regression coefficient; establishing a component failure-meteorological correlation model based on the overall regression model of the number of failures and the partial regression coefficient.

[0022] In combination with the first aspect, in a possible implementation, the above-mentioned overall regression model of the number of failures is:

[0023] C=BV+ε

[0024] Where B represents the partial regression coefficient; C is the number of component failures predicted by the component failure-meteorological probability model; V is the total number of failures of each component, and ε is the random error;

[0025] The component failure-weather correlation model is:

[0026]

[0027] in, represents the parameter estimate of the partial regression coefficient; n represents n elements, V n represents the total number of failures of the nth component, α 0 ,α 1 ,α 2 ,…,α n represents the partial regression coefficient.

[0028] In combination with the first aspect, in a possible implementation method, the above-mentioned sample training set based on the component failure-meteorological correlation model is used to train the component failure-meteorological correlation model through a reinforcement learning method to obtain a fault prediction model, including: using the component failure-meteorological correlation model to calculate fault data, and using the t-SNE dimensionality reduction visualization method to reduce the dimension of the fault data; using the mean shift algorithm to cluster the reduced-dimensional fault data to obtain an n-class training set; performing Adaboost training on the n-class training set to obtain m weak classifiers; calculating the weights of the m weak classifiers, and using the m weak classifiers and the weights to synthesize a strong classifier, and the strong classifier is the fault prediction model.

[0029] In combination with the first aspect, in a possible implementation method, the above-mentioned t-SNE dimension reduction visualization method is used to reduce the dimension of the fault data, and the formula is as follows:

[0030]

[0031] p ij Represents the data deviation in low dimension and is the fault data after dimension reduction, σ ij represents the amount of random error, ψ i represents a Gaussian distribution with variance σ.

[0032] In combination with the first aspect, in a possible implementation method, the mean shift algorithm is used to cluster the fault data after dimension reduction to obtain n types of training sets, including:

[0033] Given an initial point x 0 , set the step size, calculate the drift vector of the first step after the initial point

[0034]

[0035] Where: x 0 As a starting point, is the end point vector; represents the step size; σ is the variance of Gaussian distribution; μ ij Represents the similarity between low-dimensional space and high-dimensional space, judged by relative entropy, and the mathematical expression is

[0036] make judge Output the clustering results and obtain n types of training sets.

[0037] In combination with the first aspect, in a possible implementation manner, the m weak classifiers are:

[0038]

[0039] The strong classifier is:

[0040] y(x)=sign(m×f(x)).

[0041] In combination with the first aspect, in a possible implementation, after the component failure-weather correlation model is established according to the overall regression model of the number of failures and the partial regression coefficient, the following steps are included:

[0042] Based on the component failure-weather correlation model, the equation residual analysis is used to calculate the goodness of fit. When the goodness of fit is less than the goodness of fit threshold, the weather correction factor is re-determined and the weather correction factor is iteratively calculated.

[0043] To achieve the above object, the second aspect of the present invention provides a power system fault prediction device taking environmental factors into consideration, the device comprising:

[0044] Meteorological factor acquisition module: used to acquire historical meteorological data of the power system and historical fault data corresponding to the historical meteorological data, and analyze meteorological factors that affect the accuracy of power system fault prediction at different altitudes based on the historical meteorological data and historical fault data;

[0045] Mathematical modeling module: used to perform mathematical modeling on the meteorological factors and obtain weather correction factors and altitude correction factors;

[0046] Probabilistic model building module: used to build a component failure-meteorological probability model based on weather correction factors and altitude correction factors;

[0047] A correlation model building module: used for performing regression analysis based on the component failure-weather probability model to build a component failure-weather correlation model;

[0048] Prediction model building module: used to obtain a sample training set based on the component failure-meteorological correlation model, and train a model using the sample training set through a reinforcement learning method to obtain a fault prediction model;

[0049] Fault probability prediction module: used to input real-time data into the fault prediction model to predict the probability of failure of the power system.

[0050] The embodiment of the present invention has the following beneficial effects: in the technical solution, according to the historical meteorological data and the historical fault data corresponding to the historical meteorological data, the meteorological factors affecting the fault prediction accuracy of the power system at different altitudes are analyzed, and the meteorological factors are mathematically modeled to obtain the weather correction factor and the altitude correction factor, and according to the weather correction factor and the altitude correction factor, a component failure-meteorological probability model is constructed, and regression analysis is performed based on the component failure-meteorological probability model to construct a component failure-meteorological correlation model;

[0051] Based on the component failure-meteorological correlation model, a sample training set is obtained, and the model is trained using the sample training set through the reinforcement learning method to obtain a fault prediction model. By analyzing the meteorological factors that affect the fault prediction accuracy at different altitudes from the historical meteorological data and historical fault data of the power system, mathematical modeling of the meteorological factors is performed to obtain weather correction factors and altitude correction factors, so that the component failure-meteorological correlation model obtained according to the weather correction factor and the altitude correction factor can be combined with different altitudes and meteorological conditions to achieve the fault prediction model obtained by training the model using the sample training set, and can consider different altitudes and meteorological conditions to predict the probability of failure. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] in:

[0054] Figure 1 A schematic diagram of a flow chart of a method for predicting power system faults taking environmental factors into consideration in an embodiment of the present invention;

[0055] Figure 2 A schematic diagram of a method for establishing a component failure-weather correlation model in an embodiment of the present invention;

[0056] Figure 3 A schematic diagram of a flow chart of a method for verifying the effectiveness of a component failure-weather correlation model in an embodiment of the present invention;

[0057] Figure 4 A schematic diagram of a flow chart of a method for establishing a fault prediction model in an embodiment of the present invention;

[0058] Figure 5 It is a structural block diagram of a power system fault prediction device considering environmental factors in an embodiment of the present invention;

[0059] Figure 6 4 is a structural block diagram of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0060] 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.

[0061] This technical solution is applicable to scenarios where the probability of failure of a power system is predicted taking into account altitude and weather conditions.

[0062] The embodiment of the present invention provides a method for predicting power system faults taking environmental factors into consideration. Figure 1 , Figure 1 A flow chart of a method for predicting power system faults taking environmental factors into consideration is provided in an embodiment of the present invention, such as Figure 1 As shown, the specific steps are as follows:

[0063] Step S101: Acquire historical meteorological data of the power system and historical fault data corresponding to the historical meteorological data, and analyze meteorological factors that affect the accuracy of power system fault prediction at different altitudes based on the historical meteorological data and the historical fault data.

[0064] Step S102: mathematically model the meteorological factors to obtain weather correction factors and altitude correction factors.

[0065] Step S103: construct a component failure-meteorological probability model according to the weather correction factor and the altitude correction factor.

[0066] Step S104: Perform regression analysis based on the component failure-weather probability model to construct a component failure-weather correlation model.

[0067] Step S105, obtaining a sample training set based on the component failure-weather correlation model, and training a model using the sample training set by a reinforcement learning method to obtain a fault prediction model;

[0068] Step S106: input real-time data into the fault prediction model to predict the probability of a fault in the power system.

[0069] Among them, the weather correction factor is the correlation between different weather conditions and fault prediction accuracy; the altitude correction factor is the correlation between different altitudes and fault prediction accuracy; the fault-meteorological correlation model is the correlation between fault data and meteorological variables; the fault prediction model is used to predict the failure probability of the power system.

[0070] Since the present technical solution considers different altitudes and weather conditions to predict the failure probability of the power system, in order to better combine different altitudes and weather conditions when predicting the failure probability, the present embodiment obtains the correlation between different altitudes, weather conditions and failure prediction, and the above steps S101 and S102 can be executed.

[0071] In this embodiment, the meteorological factors that affect the accuracy of power system fault prediction at different altitudes are first analyzed. In order to improve the effectiveness of the correlation between different altitudes, weather conditions and fault prediction, this embodiment analyzes the correlation between different altitudes and weather conditions and fault prediction from the historical meteorological data recorded during the actual operation of the power system and the historical fault data corresponding to the historical meteorological data. Specifically, the historical meteorological data of the power system and the historical fault data corresponding to the historical meteorological data are obtained, wherein the historical meteorological data may include system flow data, voltage data, and weather data at the time of the fault, etc. The historical fault data is the number of faults of the power system components. In this embodiment, considering that the length of use of the power system will also have a certain impact on the failure of the components, in order to reduce the error caused by the length of use of the power system, the historical meteorological data and historical fault data can be obtained once every preset time to obtain the historical meteorological data and historical fault data of the power system in the recent time.

[0072] After obtaining the historical meteorological data and historical fault data of the power system, the meteorological factors that affect the accuracy of power system fault prediction at different altitudes are analyzed from the historical meteorological data and historical fault data. The meteorological factors include altitude factors and weather factors. In order to quantify the relationship between meteorological factors and fault prediction, this embodiment mathematically models the meteorological factors to obtain weather correction factors and altitude correction factors.

[0073] Specifically, the weather correction factor can be expressed by the following formula:

[0074]

[0075] Among them, T mj is the weather condition influencing factor, is the influence coefficient of weather before fault on fault prediction, is the influence coefficient of weather after the fault on fault prediction, G mj is the weather state coefficient, and its value is determined according to the historical data fitting results. d is the duty cycle, which can be taken as 0.75. mj The voltage rating of the power system.

[0076] in, It can be calculated by the following formula:

[0077]

[0078] Where: G mj (t) represents the weather condition coefficient at each moment in a period T, and the weather condition is recorded every 1 minute.

[0079] The altitude correction factor can be expressed as follows:

[0080]

[0081] Among them, Γ H is the altitude correction factor, f is the power frequency, and H is the altitude.

[0082] After the weather correction factor and the altitude correction factor are obtained, a component failure-meteorological probability model is established according to the weather correction factor and the altitude correction factor, that is, the above step S103 is executed.

[0083] Among them, the component failure-meteorological probability model is used to predict the failure probability of each component.

[0084] In this embodiment, the predicted fault data is obtained by using the weather correction factor and the altitude correction factor, and the predicted fault data is processed to obtain a component failure-meteorological probability model. The component failure-meteorological probability model can be expressed by the following formula:

[0085]

[0086] Among them, λ mj is the total number of failures of the mth component in a cycle, λ avg is the average number of failures within a component cycle, P j is the percentage of the duration of weather in state j in the statistical period. According to the IEEE standard guidelines, the weather is divided into three states: normal, severe, and disaster. H is the altitude correction factor, T mj It is the factor affecting weather conditions.

[0087] The average number of failures within a component cycle can be expressed by the following formula:

[0088]

[0089] Where: avg is the average failure rate correction model. a, b, c, and d correspond to four weather conditions, namely strong wind, heavy fog, thunderstorm, and snowfall.

[0090] After the component failure-weather probability model is obtained, a regression analysis is performed based on the component failure-weather probability model to construct a component failure-weather correlation model, that is, step S104 is executed.

[0091] Among them, the component failure-meteorological correlation model is the correlation between failure data and meteorological variables.

[0092] Specifically, refer to Figure 2 As shown, Figure 2FIG. 1 is a flow chart of a method for establishing a component failure-weather correlation model provided in this embodiment. Figure 2 As shown, the specific steps are as follows:

[0093] Step S1041, establishing a failure frequency overall regression model based on the component failure-meteorological probability model, and using the least squares method to perform parameter estimation on the regression coefficient of the failure frequency overall regression model to obtain a partial regression coefficient;

[0094] Step S1042: establishing a component failure-weather correlation model according to the overall regression model of the number of failures and the partial regression coefficients.

[0095] In this embodiment, the overall regression model of the number of failures can be expressed by the following formula:

[0096] C=E(C|V)+ε

[0097] Where: C is the number of component failures predicted by the component failure-meteorological probability model, V is the total number of failures of each component, E(CV) is the conditional mean of C with respect to V, and ε is the random error.

[0098] Among them, use B = [α 0 ,α 1 ,α 2 ,…,α n ] T Represents the partial regression coefficient in the overall regression model, and the overall regression model of the number of failures is equivalent to a matrix form:

[0099] C=BV+ε

[0100] Among them, use B = [α 0 ,α 1 ,α 2 ,…,α n ] T It is expressed as the partial regression coefficient in the overall regression model of the number of failures; C is the number of component failures predicted by the component failure-meteorological probability model; V is the total number of failures of each component, and ε is the random error.

[0101] In this embodiment, the regression coefficient can be evaluated by the least square method to obtain the partial regression coefficient Specifically:

[0102]

[0103] in, It represents the parameter estimate of the partial regression coefficient; T is the matrix form of the linear regression equation, and its mathematical expression is T=BW+ε; W is the meteorological condition observation sample; W' is the transpose of W.

[0104] The λ in the component failure-meteorological probability model mj has a functional relationship with V, so V can be calculated according to λ mj The final component failure-weather correlation model can be expressed as follows:

[0105]

[0106] n represents n components, V n represents the total number of failures of the nth component, α 0 ,α 1 ,α 2 ,…,α n represents the partial regression coefficient.

[0107] In addition, in order to ensure the validity of the component failure-weather correlation model, in this embodiment, the component failure-weather correlation model is verified for validity. Figure 3 , Figure 3 A schematic diagram of a flow chart of a method for verifying the effectiveness of a component failure-weather correlation model provided in this embodiment, such as Figure 3 As shown, the specific steps are as follows:

[0108] Step S10421: Based on the component failure-weather correlation model, the equation residual analysis is used to calculate the goodness of fit. When the goodness of fit is less than the goodness of fit threshold, the weather correction factor is re-determined and the weather correction factor is iteratively calculated.

[0109] In this embodiment, the goodness of fit is calculated using equation residual analysis, specifically:

[0110] R 2 =SSR / (SSR+SSE)

[0111] Among them, R 2 is the percentage of the variance to the dependent variable variance, indicating the goodness of fit. The greater the goodness of fit, the better the fitting effect, and the more accurate the prediction results of the component failure-weather correlation model.

[0112] In this embodiment, a goodness of fit threshold is set, and the validity of the component failure-meteorological correlation model is determined by comparing the calculated goodness of fit with the goodness of fit threshold. Specifically, when the goodness of fit is greater than the goodness of fit threshold, the validity of the component failure-meteorological correlation model is determined to be excellent, and the component failure-meteorological correlation model can directly output the prediction result. When the goodness of fit is less than the goodness of fit threshold, the validity of the component failure-meteorological correlation model is determined to be poor, which means that the current component failure-meteorological correlation model cannot meet the requirements and the component failure-meteorological correlation model needs to be re-determined.

[0113] In this embodiment, by re-determining the weather correction factor The value is used to re-establish the component failure-weather correlation model, specifically the weather correction factor The weather correction factor is iteratively calculated until the goodness of fit calculated by the component failure-meteorological correlation model established based on the re-determined weather correction factor and the altitude correction factor is greater than the goodness of fit threshold, then the weather correction factor is stopped. The values ​​are iteratively calculated to obtain the final component failure-weather correlation model, that is, returning to execute step S103 of constructing the component failure-weather probability model according to the weather correction factor and the altitude correction factor.

[0114] For example, the goodness of fit threshold is set to 0.85. When the goodness of fit is greater than 0.85, the effectiveness of the component failure-weather correlation model is excellent, and the prediction results of the component failure-weather correlation model are directly output; when the goodness of fit is less than 0.85, the effectiveness of the component failure-weather correlation model is poor, and the weather correction factor is adjusted. The values ​​are iteratively calculated, and finally the prediction result of the component failure-meteorological correlation model output is obtained when the goodness of fit meets the goodness of fit threshold.

[0115] After the component failure-meteorological correlation model is obtained, a fault prediction model is obtained according to the component failure-meteorological correlation model, that is, the above step S105 is executed, a sample training set is obtained based on the component failure-meteorological correlation model, and a model is trained using the sample training set through a reinforcement learning method to obtain a fault prediction model.

[0116] Among them, the fault prediction model is used to predict the failure probability of the power system.

[0117] In this embodiment, a sample training set is obtained based on the component failure-weather correlation model, and a model is trained using the sample training set to obtain a fault prediction model. In this embodiment, in order to improve the effectiveness of the acquired fault prediction model, before obtaining the fault prediction model based on the sample training set, the sample training set is processed, referring to Figure 4 As shown, Figure 4 A flowchart of a method for establishing a fault prediction model provided by an embodiment of the present invention is shown in FIG. Figure 4 As shown, the specific steps are as follows:

[0118] Step S1051, using the component failure-meteorological correlation model to calculate the fault data, and using the t-SNE dimension reduction visualization method to perform dimension reduction processing on the fault data;

[0119] Step S1052: clustering the fault data after dimension reduction using a mean shift algorithm to obtain n types of training sets;

[0120] Step S1053, perform Adaboost training on the n-type training set to obtain m weak classifiers;

[0121] Step S1054: Calculate the weights of the m weak classifiers, and use the m weak classifiers and the weights to synthesize a strong classifier, where the strong classifier is the fault prediction model.

[0122] Since the obtained fault data is high-dimensional data, in order to improve the data processing efficiency, the t-SNE method is used to reduce the dimension of the sample training set for visualization. Specifically:

[0123]

[0124] Where: p ij is the data deviation in low dimension, σ ij is the random error, ψ i is a Gaussian distribution with a variance of σ. Since the sample training set contains fault data of multiple power system components, ψ i It can be understood as a Gaussian distribution with a variance of σ for the i-th element.

[0125] Different variances σ correspond to different Gaussian distributions, and we need to find the optimal σ ij :

[0126]

[0127] Secondly, the fault data is visualized and the similarity of each pair of adjacent moments in low dimension is calculated, i.e., ψ i With ψ i+1 The similarity between:

[0128]

[0129] Among them, ij represents the similarity of each pair of adjacent moments in low dimension; σ represents the variance of Gaussian distribution; ψ i is a Gaussian distribution with variance σ; μ ij Represents the similarity between low-dimensional space and high-dimensional space, judged by relative entropy, and the mathematical expression is

[0130] In order to improve the model training effect, the drift mean algorithm is used to perform cluster analysis on the sample training set after dimension reduction visualization processing to obtain n types of sample training sets, specifically:

[0131] Given an initial point x 0 , set the step size, calculate the drift vector of the first step after the initial value point

[0132]

[0133] Among them, x 0 As a starting point, is the end point vector; represents the step size; σ is the variance of Gaussian distribution; μ ij Represents the similarity between low-dimensional space and high-dimensional space, judged by relative entropy, and the mathematical expression is

[0134] make judge Output the clustering results and obtain n types of training sets.

[0135] Initialize data:

[0136] G 1 =(η 11 ,…,η 1i ,…,η 1N )

[0137] Where N is the number of training sets, η 1i is the weight of the i-th group of samples after the first iteration.

[0138] After obtaining the n-class sample training set, perform Adaboost training on the n-class training set to obtain m weak classifiers:

[0139]

[0140] Calculate the weight λ corresponding to each classifier m , using m weak classifiers and corresponding weights λ m By synthesizing a strong classifier, we can get the fault prediction model:

[0141] y(x)=sign(m×f(x))

[0142] Specifically, m weak classifiers are input to synthesize a strong classifier, which is the fault prediction model. The product of the predicted results of each weak classifier and its corresponding weight is the final prediction result of each classifier. Then, the final prediction results of all classifiers are added together to obtain the final prediction result, which is the prediction result of the strong classifier, that is, the predicted fault probability of the fault prediction model. In this embodiment, the fault probability is predicted by inputting the real-time data of the power system into the fault prediction model.

[0143] The above describes a method for predicting power system faults taking environmental factors into consideration provided by the present embodiment. To better implement the above method, the following describes a device 50 for predicting power system faults taking environmental factors into consideration. The device includes:

[0144] Meteorological factor acquisition module 501: used to acquire historical meteorological data of the power system and historical fault data corresponding to the historical meteorological data, and analyze meteorological factors that affect the accuracy of power system fault prediction at different altitudes based on the historical meteorological data and the historical fault data;

[0145] Mathematical modeling module 502: used to perform mathematical modeling on the meteorological factors to obtain weather correction factors and altitude correction factors;

[0146] Probabilistic model building module 503: used to build a component failure-meteorological probability model according to the weather correction factor and the altitude correction factor;

[0147] A correlation model building module 504 is used to perform regression analysis based on the component failure-weather probability model to build a component failure-weather correlation model;

[0148] Prediction model building module 505: used to use the fault data obtained based on the component fault-weather correlation model as a sample training set, and train the model using the sample training set through a reinforcement learning method to obtain a fault prediction model;

[0149] Fault probability prediction module 506: used to input real-time data into the fault prediction model to predict the probability of faults occurring in the power system.

[0150] In one possible implementation, the above-mentioned correlation model construction module 504 is specifically used to: establish an overall regression model of the number of failures based on the component failure-meteorological probability model, and use the least squares method to perform parameter estimation on the regression coefficient of the overall regression model of the number of failures to obtain the partial regression coefficient; according to the overall regression model of the number of failures and the partial regression coefficient, establish a component failure-meteorological correlation model.

[0151] In one possible implementation, the prediction model construction module 505 is specifically used to: use the component failure-meteorological correlation model to calculate the fault data, and use the t-SNE dimension reduction visualization method to reduce the dimension of the fault data; use the mean shift algorithm to cluster the reduced fault data to obtain n-type training sets; perform Adaboost training on the n-type training sets to obtain m weak classifiers; calculate the weights of the m weak classifiers, and use the m weak classifiers and the weights to synthesize a strong classifier, and the strong classifier is the fault prediction model.

[0152] In the above-mentioned device, the meteorological factors affecting the fault prediction accuracy of the power system at different altitudes are analyzed according to the historical meteorological data and the historical fault data corresponding to the historical meteorological data, and the meteorological factors are mathematically modeled to obtain the weather correction factor and the altitude correction factor. After the weather correction factor and the altitude correction factor are obtained, a component fault-meteorological correlation model is established according to the weather correction factor and the altitude correction factor. A sample training set is obtained based on the component fault-meteorological correlation model, and the model is trained using the sample training set by the reinforcement learning method to obtain a fault prediction model. By analyzing the meteorological factors affecting the fault prediction accuracy at different altitudes from the historical meteorological data of the power system and the historical fault data corresponding to the historical meteorological data, the meteorological factors are mathematically modeled to obtain the weather correction factor and the altitude correction factor, so that the component fault-meteorological correlation model obtained according to the weather correction factor and the altitude correction factor can combine different altitudes and meteorological conditions to obtain fault data, so as to realize the fault prediction model obtained by training the model using the sample training set, and can consider different altitudes and meteorological conditions to predict the probability of failure.

[0153] Figure 6 FIG. 1 shows an internal structure diagram of a computer device in an embodiment. The computer device may be a terminal or a server. Figure 6 As shown, the computer device includes a processor, a memory and a network interface connected via a system bus. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor can implement the age recognition method. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor can implement the age recognition method. Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0154] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0155] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0156] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for predicting power system faults taking environmental factors into consideration, characterized in that: The method comprises: Acquire historical meteorological data of the power system and historical fault data corresponding to the historical meteorological data, and analyze meteorological factors that affect the accuracy of power system fault prediction at different altitudes based on the historical meteorological data and the historical fault data; Performing mathematical modeling on the meteorological factors to obtain weather correction factors and altitude correction factors; Construct a component failure-meteorological probability model based on weather correction factors and altitude correction factors; Perform regression analysis based on the component failure-weather probability model to construct a component failure-weather correlation model; The fault data is calculated using the component fault-meteorological correlation model, and the t-SNE dimension reduction visualization method is used to reduce the dimension of the fault data; The mean shift algorithm is used to cluster the fault data after dimension reduction to obtain the Q-class training set; Perform Adaboost training on the Q-class training set to obtain m weak classifiers; Calculate the weights of m weak classifiers, and use the m weak classifiers and the weights to synthesize a strong classifier, where the strong classifier is a fault prediction model; Inputting real-time data into the fault prediction model to predict the probability of a fault in the power system; The formula for the weather correction factor is as follows: in, represents the weather correction factor, is the gravitational coefficient of the weather before the fault on the fault prediction, is the gravitational coefficient of the weather after the fault on the fault prediction, is the weather state coefficient, is the duty cycle, is the voltage rating; The formula for the altitude correction factor is as follows: in, is the altitude correction factor, is the power frequency, is the altitude; The formula of the component failure-meteorological probability model is as follows: in, For a cycle The total number of component failures, is the average number of failures within a component cycle, for The duration of the weather status accounts for the percentage of the statistical period. The weather is divided into three states: normal, severe, and disaster. is the altitude correction factor, is the weather correction factor; The step of performing regression analysis based on the component failure-weather probability model to construct a component failure-weather correlation model includes: Based on the component failure-meteorological probability model, a failure number overall regression model is established, and the least square method is used to estimate the regression coefficients of the failure number overall regression model to obtain partial regression coefficients. Establishing a component failure-weather correlation model based on the overall regression model of the number of failures and the partial regression coefficients; The overall regression model of the number of failures is: Where B represents the partial regression coefficient; The number of component failures predicted by the component failure-meteorological probability model; is the total number of failures of each component, is a random error; The component failure-weather correlation model is: in, represents the parameter estimate of the partial regression coefficient; n represents n elements, represents the total number of failures of the nth component, represents the partial regression coefficient.

2. The method according to claim 1, characterized in that The t-SNE dimension reduction visualization method is used to reduce the dimension of the fault data. The formula is as follows: Indicates the data deviation in low dimension, represents the amount of random error, The variance is Gaussian distribution.

3. The method according to claim 1, characterized in that The mean shift algorithm is used to cluster the fault data after dimension reduction to obtain a Q-class training set, including: Given an initial point , set the step size, calculate the drift vector of the first step after the initial point : Where: is the initial point, is the end point vector; represents the step length; is the variance of Gaussian distribution; Represents the similarity between low-dimensional space and high-dimensional space, judged by relative entropy, and the mathematical expression is ; make ,judge , output the clustering results and obtain the Q-class training set.

4. The method according to claim 1, characterized in that After the component failure-weather correlation model is established according to the overall regression model of the number of failures and the partial regression coefficient, the method further comprises: Based on the component failure-weather correlation model, the equation residual analysis is used to calculate the goodness of fit. When the goodness of fit is less than the goodness of fit threshold, the weather correction factor is re-determined and the weather correction factor is iteratively calculated.

5. A power system fault prediction device taking environmental factors into consideration, characterized in that: The device comprises: Meteorological factor acquisition module: used to acquire historical meteorological data of the power system and historical fault data corresponding to the historical meteorological data, and analyze meteorological factors that affect the accuracy of power system fault prediction at different altitudes based on the historical meteorological data and historical fault data; Mathematical modeling module: used to perform mathematical modeling on the meteorological factors to obtain weather correction factors and altitude correction factors; the formula of the weather correction factor is as follows: in, represents the weather correction factor, is the gravitational coefficient of the weather before the fault on the fault prediction, is the gravitational coefficient of the weather after the fault on the fault prediction, is the weather state coefficient, is the duty cycle, is the voltage rating; The formula for the altitude correction factor is as follows: in, is the altitude correction factor, is the power frequency, is the altitude; Probabilistic model building module: used to build a component failure-meteorological probability model based on the weather correction factor and the altitude correction factor; the formula of the component failure-meteorological probability model is as follows: in, For a cycle The total number of component failures, is the average number of failures within a component cycle, for The duration of the weather status accounts for the percentage of the statistical period. The weather is divided into three states: normal, severe, and disaster. is the altitude correction factor, is the weather correction factor; Correlation model building module: used to perform regression analysis based on the component failure-meteorological probability model to build a component failure-meteorological correlation model; establish a failure number overall regression model based on the component failure-meteorological probability model, and use the least squares method to perform parameter estimation on the regression coefficient of the failure number overall regression model to obtain a partial regression coefficient; Establishing a component failure-weather correlation model based on the overall regression model of the number of failures and the partial regression coefficients; The overall regression model of the number of failures is: Where B represents the partial regression coefficient; The number of component failures predicted by the component failure-meteorological probability model; is the total number of failures of each component, is a random error; The component failure-weather correlation model is: in, represents the parameter estimate of the partial regression coefficient; n represents n elements, represents the total number of failures of the nth component, represents the partial regression coefficient; Prediction model building module: used to calculate fault data using the component fault-meteorological correlation model, and use the t-SNE dimension reduction visualization method to reduce the dimension of the fault data; use the mean shift algorithm to cluster the fault data after dimension reduction to obtain a Q-class training set; perform Adaboost training on the Q-class training set to obtain m weak classifiers; calculate the weights of the m weak classifiers, and use the m weak classifiers and weights to synthesize a strong classifier, which is a fault prediction model; Fault probability prediction module: used to input real-time data into the fault prediction model to predict the probability of failure of the power system.

Images