A method and system for analyzing reliability of power system

Abstract

The present invention discloses a method and system for analyzing the reliability of a power system, and relates to the technical field of power system reliability analysis, including collecting power system data and preprocessing it, and extracting features from the preprocessed data; constructing a first model based on the preprocessed data, and obtaining a first result according to the first model; constructing a second model according to real-time data and the first result, obtaining a second result, and optimizing the information gain sensitivity parameter in the first model using the second result; and determining the final system reliability index according to the optimized first model and the second model. The present invention establishes a comprehensive, accurate and adaptive power system reliability analysis framework by comprehensively using technologies such as the Internet of Things, advanced feature extraction algorithms, multidimensional fault pattern recognition and dynamic Monte Carlo simulation. It can capture system state changes in real time, deeply explore potential patterns in historical data, and effectively identify key influencing factors.

Description

Technical Field

[0001] The present invention relates to the technical field of power system reliability analysis, and in particular to a power system reliability analysis method and system. Background Art

[0002] As an important infrastructure in modern society, the reliability of the power system is directly related to the development of the national economy and social stability. With the continuous expansion of the scale and the increasing complexity of the power system, the traditional reliability analysis methods are increasingly showing their limitations. In recent years, with the rapid development of artificial intelligence and big data technology, data-driven power system reliability analysis methods have gradually become a research hotspot. These methods build more accurate system models by mining the information contained in massive historical data, thereby improving the accuracy and prediction ability of reliability assessment. However, the existing data-driven methods still have many shortcomings, such as the poor adaptability of the model, the difficulty in effectively handling the dynamically changing system state, and the insufficient sensitivity analysis of key parameters.

[0003] The existing power system reliability analysis technology has the following main problems: first, a single model is difficult to fully capture the complex dynamic characteristics of the system, resulting in limited accuracy and reliability of the analysis results; second, the static model cannot respond to the real-time changes in the system status in a timely manner, making the analysis results lag behind the actual situation; third, the traditional method does not have sufficient sensitivity analysis of model parameters, making it difficult to effectively identify and optimize key influencing factors; finally, the existing technology often ignores the organic combination of historical data and real-time data, and fails to fully utilize the rich information contained in multi-source heterogeneous data. These problems seriously restrict the effect and application scope of power system reliability analysis. Summary of the invention

[0004] In view of the problems existing in the existing system reliability analysis method, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is that a single model is difficult to fully capture the complex dynamic characteristics of the system, resulting in limited accuracy and reliability of the analysis results. The static model cannot respond to real-time changes in the system status in a timely manner, causing the analysis results to lag behind the actual situation.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a method for analyzing reliability of a power system, which comprises:

[0008] Collect power system data and preprocess it, and extract features from the preprocessed data;

[0009] Building a first model based on the feature-extracted data, and obtaining a first result according to the first model;

[0010] Constructing a second model according to the real-time data and the first result to obtain a second result, and optimizing the information gain sensitivity parameter in the first model using the second result;

[0011] The final system reliability index is determined according to the optimized first model and the second model.

[0012] As a preferred solution of the power system reliability analysis method of the present invention, the feature extraction extracts data points that have an impact on system reliability from the original signal according to importance through a self-compiler and mutual information, as shown in the following formula:

[0013] ,

[0014] Among them, F advanced is the high-level extracted feature, X is the original data matrix, W 1 and W 2 is the weight matrix of the autoencoder, b 1 and b 2 is the bias vector, σ is the sigmoid activation function, ReLU is the rectified linear unit activation function, m is the number of original features, and λ j is the attenuation factor of the jth feature, H(X j ) is the entropy of the jth feature, H(Y) is the entropy of the target variable Y, H(X j , Y) is X j The joint entropy of and Y.

[0015] As a preferred solution of the power system reliability analysis method of the present invention, wherein: the construction of the first model includes:

[0016] Construct a decision tree based on the information gain criterion and evaluate feature importance;

[0017] Ensemble the decision trees and ensure that each decision tree uses a random subset of features;

[0018] Create composite features through feature engineering and weighted fusion random forest model.

[0019] As a preferred solution of the power system reliability analysis method of the present invention, the decision tree uses information gain as the splitting criterion and selects the most discriminative feature partition data. The information gain is calculated as shown in the following formula:

[0020] ,

[0021] Among them, H(T) is the entropy of node T, a is the feature, Tv is the subset when feature a takes value v, values(a) is all possible values ​​of feature a, and IG() is the information gain of the decision tree; is the entropy value of the subset when the feature value is v;

[0022] The decision tree ensemble is represented by the following formula:

[0023] ,

[0024] Among them, f b (x) is a single decision tree, B is the number of decision trees, L b is the set of leaf nodes of decision tree b, c l is the predicted value of leaf node l, I() is the indicator function, R l is the feature space region corresponding to leaf node l; Is a single decision tree for the input feature x The prediction output function of

[0025] The feature engineering is shown in the following formula:

[0026] ,

[0027] Among them, φ k is the feature conversion function, h k is an aggregate function; In X is the original feature vector that needs to be transformed by feature engineering;

[0028] The decision tree and the feature engineering are integrated to form a first model, as shown in the following formula:

[0029] MDFM ( X ) = a ⋅ 1 B ∑ b = 1 B [ β ⋅ ∑ i = 1 N w i ⋅ f b ( x i ) + ( 1 + β ) ⋅ exp( − γ ⋅ IG b ) ∑ j = 1 M exp( − γ ⋅ IG j ) ] + ( 1 − a ) ⋅ g k ( X ) ,

[0030] Among them, MDFM(X) is the output of the multi-dimensional fault mode model, X is the input feature vector, α is the model fusion parameter, B is the number of decision trees in the random forest, β is the internal balance parameter of the decision tree, N is the number of features, and w i is the weight of the i-th feature, f b (x i ) is the prediction result of the bth decision tree for the i-th feature, γ is the information gain sensitivity parameter, IG b is the information gain of the bth decision tree, M is the total number of all decision trees, K is the number of composite features after feature engineering, and g k (X) is the function of the kth composite feature; It is the contribution measure of the jth feature to data classification.

[0031] As a preferred solution of the power system reliability analysis method of the present invention, the construction of the second model includes:

[0032] Establishing a basic simulation model and integrating the real-time data and the first result;

[0033] Generate fault scenarios based on current system status and load changes using dynamic sampling algorithms;

[0034] The fault scenarios are simulated simultaneously, and the reliability index under each fault scenario is calculated.

[0035] As a preferred solution of the power system reliability analysis method of the present invention, the dynamic sampling algorithm includes:

[0036] Construct a weighted average framework as shown below:

[0037] ;

[0038] For each scenario, determine the reliability of all critical components , as shown below:

[0039] ;

[0040] Analyze system reliability over time for each scenario , as shown below:

[0041] ;

[0042] in, R sys is the system reliability index, N is the number of dynamically generated fault scenarios, M is the number of components in the system, ω i is the weight of the i-th scenario, P f,j is the failure probability of the jth component, obtained from the failure mode model, I c,j is the importance coefficient of component j, λ j is the failure rate of component j, t i is the simulation time of the i-th scenario; is the failure rate, is the failure rate of the jth component, is the simulation time of the ith scenario, e is the base of natural logarithms.

[0043] As a preferred solution of the power system reliability analysis method of the present invention, wherein: the parameter of the first model optimized by using the second result is an information gain sensitivity parameter, comprising the following steps:

[0044] Define the search range of the information gain sensitivity parameter and discretize it with a fixed step size;

[0045] For each information gain sensitivity parameter, updating the first model and calculating the result;

[0046] generating failure scenarios through a second model and calculating average system reliability;

[0047] Compare different information gain sensitivity parameter values ​​and select the information gain sensitivity parameter with the highest average system reliability;

[0048] Based on the information gain sensitivity parameter with the highest system reliability, the step size is reduced and the optimization steps are repeated to determine the optimal information gain sensitivity parameter;

[0049] The optimal information gain sensitivity parameter is the information gain sensitivity value that maximizes the system reliability index under the current system state.

[0050] In a second aspect, an embodiment of the present invention provides a power system reliability analysis system, which includes:

[0051] The data acquisition module is used to collect and pre-process the power system data and extract features from the pre-processed data;

[0052] A first model building module, used to build a first model based on the preprocessed data, and determine a first result according to the first model;

[0053] A second model building module, used to build a second model according to the real-time data and the first result to obtain a second result;

[0054] The optimization module is used to optimize the parameters in the first model using the second result, and determine the final system reliability index according to the optimized first model and the second model.

[0055] In a third aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, any step of the above-mentioned power system reliability analysis method is implemented.

[0056] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the above-mentioned power system reliability analysis method is implemented.

[0057] The beneficial effect of the present invention is to establish a comprehensive, accurate and adaptive power system reliability analysis framework by comprehensively using technologies such as Internet of Things technology, advanced feature extraction algorithm, multidimensional fault pattern recognition and dynamic Monte Carlo simulation. It can not only capture system state changes in real time, but also deeply explore potential patterns in historical data and effectively identify key influencing factors. Through the feature extraction method combining autoencoder and mutual information, the system can adapt to the nonlinear and time-varying characteristics of the power system and improve data utilization efficiency. The introduction of the multidimensional fault pattern recognition model, combined with random forest and feature engineering, enhances the system's understanding and prediction capabilities of complex fault modes. Dynamic Monte Carlo simulation further optimizes the accuracy of reliability assessment, while the dynamic adjustment mechanism of information gain sensitivity parameters ensures that the model can be continuously optimized and adapt to the dynamic changes of the system. This multi-level, adaptive analysis method significantly improves the accuracy, real-time and robustness of power system reliability assessment, and provides a powerful tool for operation management and decision support of power systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. 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 creative work. Among them:

[0059] Figure 1 Flowchart of the power system reliability analysis method.

[0060] Figure 2 Optimization flow chart of power system reliability analysis method. DETAILED DESCRIPTION

[0061] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0062] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0063] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0064] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0065] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0066] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0067] Example 1

[0068] Reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a method and system for analyzing reliability of a power system, including:

[0069] In view of the existing power system reliability analysis technology, the present invention proposes a new power system reliability analysis method and system. This method constructs a multi-level model, organically combines historical data and real-time data, and introduces an information gain sensitivity parameter optimization mechanism to effectively solve the problems of poor model adaptability, insufficient dynamic response capability, and insufficient parameter sensitivity analysis in the existing technology. The details are as follows:

[0070] It should be noted that, in this embodiment, the first model is a multidimensional failure mode model, the second model is a model for system reliability calculation based on the Monte Carlo method, the first result is the output of the multidimensional failure mode model, and the second result is a system reliability index.

[0071] IoT devices and smart sensor networks deployed at key nodes of the power system continuously collect data at a high frequency. These devices include not only traditional voltage and current sensors, but also environmental parameter monitors such as temperature, humidity, vibration, as well as smart meters and phasor measurement units (PMUs), and transmit the data in real time to the central data processing center.

[0072] In the data center, data quality checks are first performed to identify and mark outliers, noise, and missing data. For detected anomalies, the algorithm intelligently interpolates or corrects based on historical patterns and data from adjacent sensors. Next, the algorithm applies advanced feature extraction techniques to extract key features from the original signal. These features are ranked by importance, automatically identifying and filtering out data points that have a significant impact on system reliability.

[0073] Preferably, in order to screen data points that have a significant impact on system reliability, the advanced feature extraction technology uses autoencoders as the basic feature extraction technology, considering that power system data usually has high dimensions, nonlinearity, and complex spatiotemporal correlations. It can effectively learn these complex features, especially when processing large-scale sensor data. For example, it can capture the potential relationship between voltage fluctuations, load changes, and equipment status, as follows:

[0074] ,

[0075] Further considering the importance of each feature, for example, the temperature of some key equipment is more important than the ambient humidity. Mutual information is used to quantify the correlation between each feature and the system to help identify the most important features, as follows:

[0076] ,

[0077] Only mutual information is added to quantify the correlation of each feature, but it is actually uncontrollable, that is, the importance of different parameters at different time scales is different, so it is necessary to adjust the mutual information for the data at a specific time to extract the most important features. At the same time, when processing different types of features, adjustments can also be made, which can be expressed as follows:

[0078] ,

[0079] Finally, while retaining the complex features extracted by the autoencoder, dynamically adjusting the impact of each feature based on its importance means that in power systems, this feature extraction technology is better adapted to different operating conditions and failure modes. For example, during normal operation, it pays more attention to long-term trends; when an anomaly is detected, it quickly turns its attention to the key parameters that cause the failure:

[0080] ,

[0081] Expands to:

[0082] ,

[0083] Among them, F advanced is the high-level extracted feature, X is the original data matrix, W 1 and W 2 is the weight matrix of the autoencoder, b 1 and b 2 is the bias vector, σ is the sigmoid activation function, ReLU is the rectified linear unit activation function, m is the number of original features, λ_j is the attenuation factor of the jth feature, H(X j ) is the entropy of the jth feature, H(Y) is the entropy of the target variable Y, H(X j ,Y) is X j The joint entropy of and Y.

[0084] For example, in one feasible embodiment, as follows:

[0085] Construct an autoencoder model and train the autoencoder. Calculate the entropy of each original feature and the entropy of the target variable. In this embodiment, the entropy of the target variable is the entropy of the system reliability index. Synchronously calculate the joint entropy of the original feature and the reliability index. Calculate the mutual information through the results, initialize the attenuation factor of each feature, and calculate the importance factor.

[0086] Initial feature extraction is performed through the trained autoencoder, and then adjusted based on the feature importance. After verification, feature extraction is performed on the newly collected data.

[0087] Furthermore, data is collected from the historical operation records of the power system, including equipment fault records, operating parameters, environmental conditions, etc. The collected historical data is processed through the above data preprocessing method, and systematically classified and labeled, including multiple dimensions such as fault type, fault severity, fault duration, and fault impact range. Each fault event is assigned a multi-dimensional label to form structured fault event data.

[0088] The fault time data is then subjected to the above-mentioned feature extraction method to combine feature vectors, and a multi-dimensional fault pattern recognition model is constructed based on these feature vectors.

[0089] Furthermore, in the multi-dimensional fault mode of the power system, not all features contribute equally to fault prediction. For example, the operating parameters of some equipment may be more indicative of potential faults than environmental factors. The importance of features is evaluated by the random forest algorithm to identify the most critical fault indicators. The specific formula is as follows:

[0090] ,

[0091] Where T is the number of decision trees, ΔI(X i , t) is the information gain of feature Xi in tree t.

[0092] A decision tree is constructed based on the information gain criterion. The information gain is calculated for each node. The information gain is used as the splitting criterion. The most discriminative features are selected at each step to divide the data. This can clearly show the decision paths that lead to different types of failures. The specific formula is as follows:

[0093] ,

[0094] Among them, H(T) is the entropy of node T, a is the feature, T v is the subset when feature a takes value v, and values(a) is all possible values ​​of feature a; It is the entropy value of the subset when the feature value is v.

[0095] Further building multiple decision trees and integrating their results, each tree uses a randomly selected subset of features. This approach increases the robustness of the model, especially when facing new or rare failure modes, and can more comprehensively simulate various possible failure scenarios:

[0096] ,

[0097] Among them, f b (x) is a single decision tree, B is the number of decision trees, L b is the set of leaf nodes of decision tree b, c l is the predicted value of leaf node l, I() is the indicator function, R l is the feature space region corresponding to leaf node l; Is a single decision tree for the input feature x The predicted output function.

[0098] However, power system failures are often not caused by a single factor, but the result of the interaction of multiple factors. It is necessary to create composite features (such as equipment utilization rate, load fluctuation rate, etc.) through feature engineering. For example, equipment utilization rate (operating time / total time) may be more indicative of equipment failure risk than simple operating time:

[0099] ,

[0100] Among them, φ k is the feature conversion function, h k is an aggregate function; In X is the original feature vector that needs to be transformed by feature engineering.

[0101] The random forest model and feature engineering results are weighted and integrated to combine the prediction capability based on the tree structure and the deep insights based on the composite features. This allows the model to consider both the predictive power of the original features and the complex relationships revealed by the composite features, thus providing a more comprehensive and accurate fault prediction:

[0102] ,

[0103] Specifically expanded as follows:

[0104] MDFM ( X ) = a ⋅ 1 B ∑ b = 1 B [ β ⋅ ∑ i = 1 N w i ⋅ f b ( x i ) + ( 1 + β ) ⋅ exp( − γ ⋅ IG b ) ∑ j = 1 M exp( − γ ⋅ IG j ) ] + ( 1 − a ) ⋅ g k ( X ) ,

[0105] Among them, MDFM(X) is the output of the multi-dimensional fault mode model, X is the input feature vector, α is the model fusion parameter, B is the number of decision trees in the random forest, β is the internal balance parameter of the decision tree, N is the number of features, and w i is the weight of the i-th feature, f b (x i ) is the prediction result of the bth decision tree for the i-th feature, γ is the sensitivity parameter of information gain, IG b is the information gain of the bth decision tree, M is the total number of all decision trees, K is the number of composite features after feature engineering, and g k (X) is the function of the kth composite feature; It is the contribution measure of the jth feature to data classification.

[0106] The structure of the decision tree is determined by information gain, that is, at each node, the feature that can maximize the information gain is selected for splitting. Features with high information gain are more likely to appear at the top of the decision tree. The individual predictions of the decision tree are then integrated by random forest. Combined with feature engineering, the model can not only predict the probability of failure, but also explain the prediction results by analyzing the contributions of different components, providing valuable insights for the maintenance and optimization of power systems.

[0107] Furthermore, a basic simulation model is established, including all key components such as power generation, transmission, distribution and load. This model should be able to accurately reflect the system topology and the characteristics of each component. Next, real-time data and failure mode models from smart sensor networks are integrated into this basic model. The failure mode model includes various possible failure types, probability of occurrence and impact range.

[0108] The dynamic sampling algorithm is further developed to generate a large number of possible failure scenarios based on the current system status and predicted load changes. In order to improve the simulation efficiency, parallel computing technology is used to distribute a large number of scenarios to multiple computing nodes for simultaneous simulation. Each simulation scenario will calculate the reliability index of the system in that case. Finally, all simulation results are statistically analyzed to obtain the probability distribution and risk assessment of the overall reliability of the system.

[0109] Preferably, starting from the basic idea of ​​the Monte Carlo method, a weighted average framework is constructed:

[0110] ,

[0111] For each scenario, consider the reliability of all key components:

[0112] ,

[0113] Furthermore, different components have different impacts on system reliability. For example, a trunk line failure is more serious than a branch line. It is also necessary to consider that the failure rate of power system components usually varies over time, so as to analyze the scenario:

[0114] ,

[0115] Among them, R sys is the system reliability index, N is the number of dynamically generated fault scenarios, M is the number of components in the system, ω i is the weight of the i-th scenario, P f,j is the failure probability of the jth component, obtained from the failure mode model, I c,j is the importance coefficient of component j, λ j is the failure rate of component j, t i is the simulation time of the i-th scenario; is the failure rate, is the failure rate of the jth component, is the simulation time of the ith scenario, e is the base of natural logarithms.

[0116] Specifically, according to the above parameters of the actual system comfort, including the number of failure scenarios and the number of system components, the component importance coefficient is determined, and then the failure probability and failure rate of each component are obtained through the failure mode model.

[0117] For each scenario and component, calculate its reliability contribution, add the weight of the current scenario, and then divide it by the total weight to get the system reliability index. Then, based on the system reliability and failure scenarios, predict the change in system load.

[0118] Furthermore, in the power system, different components and factors have different degrees of influence on system reliability. The decision tree uses information gain as the splitting criterion, which in turn affects the prediction accuracy of the multidimensional fault model. Therefore, optimizing the information gain sensitivity parameter can directly improve the prediction accuracy of the multidimensional fault model. The calculation of the system reliability index is based on the multidimensional fault model, which further optimizes the final decision support, as follows:

[0119] First, a reasonable search range for the information gain sensitivity parameter is defined and discretized with a fixed step size. For each sensitivity parameter, the multidimensional fault model is updated and the failure probability of each component is calculated. Then, a dynamic Monte Carlo simulation is run to generate multiple fault scenarios and calculate the average system reliability. By comparing the system reliability index under different information gain sensitivity parameter values, the sensitivity parameter value that makes the system reliability the highest is selected. On this basis, the step size is further reduced and the above steps are repeated to obtain the optimal sensitivity parameter. Finally, independent test data is used to verify the performance of the optimized model, and multiple Monte Carlo simulations are performed to evaluate the stability of the results.

[0120] It should be noted that the optimal sensitivity parameter is the information gain sensitivity value that can maximize the system reliability index under the current power system state.

[0121] Based on the optimized multi-dimensional failure mode model and dynamic Monte Carlo simulation, the system reliability index under the current real-time data is determined.

[0122] Furthermore, this embodiment also provides a power system reliability analysis system, including:

[0123] The data acquisition module is used to collect and pre-process the power system data and extract features from the pre-processed data;

[0124] A first model building module, used to build a first model based on the preprocessed data, and determine a first result according to the first model;

[0125] A second model building module, used to build a second model according to the real-time data and the first result to obtain a second result;

[0126] The optimization module is used to optimize the parameters in the first model using the second result, and determine the final system reliability index according to the optimized first model and the second model.

[0127] This embodiment also provides a computer device, which is suitable for the case of the power system reliability analysis method, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to implement the power system reliability analysis method proposed in the above embodiment.

[0128] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0129] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for analyzing the reliability of a power system as proposed in the above embodiment is implemented.

[0130] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0131] Example 2

[0132] This is the second embodiment of the present invention. This embodiment provides a power system reliability analysis method and system. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0133] To verify the effectiveness of the proposed power system reliability analysis method, a six-month experiment was conducted in a distribution network of a provincial power grid company. The distribution network covers urban, suburban and rural areas and contains various types of users and substations of different capacities.

[0134] First, a comprehensive set of IoT devices and smart sensor networks were deployed at key nodes of the distribution network. These devices include 250 smart meters, 100 phasor measurement units (PMUs), 500 environmental parameter monitors (temperature, humidity, vibration, etc.), and 300 traditional voltage and current sensors. All devices are configured to collect data once a second and transmit it to the central data processing center in real time through a secure communication protocol.

[0135] At the data center, the received data is first quality checked. Using anomaly detection algorithms based on statistics and machine learning, about 2.3% of outliers and 0.5% of missing data were successfully identified and marked. For detected anomalies, intelligent interpolation is performed using a time series prediction model based on a long short-term memory network (LSTM), which takes into account historical patterns and data from adjacent sensors, and the interpolation accuracy reaches 97.8%.

[0136] Next, an advanced feature extraction technique combining autoencoders and mutual information is applied to extract key features from the original signal. The structure of the autoencoder is a 3-layer encoder and a 3-layer decoder, with the bottleneck layer in the middle containing 50 neurons. Through repeated experiments, the optimal learning rate was determined to be 0.001, the batch size was 64, and the number of training rounds was 100. This structure can effectively capture the potential relationship between voltage fluctuations, load changes, and device status.

[0137] In order to further improve the effect of feature extraction, a dynamic weight adjustment mechanism based on mutual information is introduced. First, the entropy of each original feature and the entropy of the target variable (system reliability index) are calculated, and then their joint entropy is calculated. Through these results, the mutual information between each feature and the system reliability is calculated and used to initialize the feature attenuation factor.

[0138] In the experiment, nearly five years of data were collected from the historical operation records of the power system, including more than 10,000 equipment failure records, more than 1 million hours of operating parameters and environmental conditions data. After preprocessing, these data were used to train a multidimensional fault pattern recognition model. The model uses a random forest algorithm and contains 500 decision trees, with the maximum depth of each tree set to 10.

[0139] In order to evaluate the contribution of different features to fault prediction, the information gain of each feature was calculated. The experiment found that transformer oil temperature, line load rate and ambient humidity are the three most significant factors affecting system reliability. Based on these findings, new composite features such as "equipment utilization rate" (operating time / total time) and "load fluctuation rate" (standard deviation of load within 24 hours / average load) were created through feature engineering.

[0140] Finally, a comprehensive simulation model including power generation, transmission, distribution and load was established. The model accurately reflects the topology and component characteristics of the entire provincial power grid. Through parallel computing technology, dynamic Monte Carlo simulations were run simultaneously on 100 high-performance computing nodes, and 10,000 possible fault scenarios were generated each time. The following are some of the key data collected in the experiment:

[0141] Table 1 Experimental data table

[0142] ,

[0143] By comparing the data in the table, it can be clearly seen that the comprehensive method proposed in the present invention has achieved significant improvements in multiple key indicators.

[0144] First, in terms of feature extraction time, the proposed comprehensive method only takes 45.2 seconds, which is 82.36% less than the 256.3 seconds of the traditional method. This significant efficiency improvement is mainly due to the advanced feature extraction technology that combines autoencoders and mutual information, which can quickly capture complex nonlinear relationships and spatiotemporal correlations, greatly shortening data processing time.

[0145] The reduction in the number of features is also an important improvement. The traditional method uses 85 features, while the proposed method only uses 37, a reduction of 56.47%. This not only reduces the computational complexity, but also improves the generalization ability of the model. Through the dynamic weight adjustment mechanism of mutual information, the method can automatically identify and retain the most relevant features, effectively reducing the noise and redundancy of the data.

[0146] In terms of model training time, the new method only takes 72.4 minutes, which is 61.39% less than the 187.5 minutes of the traditional method. This efficiency improvement mainly comes from the reduction in the number of features and the optimized use of the random forest algorithm. Through the precise calculation of information gain and the dynamic adjustment of feature importance, the model can converge to the optimal solution more quickly.

[0147] The most critical thing is the improvement in fault prediction accuracy. The proposed method achieved an accuracy of 95.3%, which is 15.24% higher than the 82.7% of the traditional method. This significant improvement is attributed to the introduction of a multidimensional fault pattern recognition model and the innovative application of feature engineering. By considering composite features such as equipment utilization and load fluctuation, the model can capture potential fault modes more comprehensively.

[0148] The system reliability assessment error was reduced from 8.6% in the traditional method to 2.8%, an improvement of 67.44%. This improvement in accuracy is mainly due to the optimization of dynamic Monte Carlo simulation and information gain sensitivity parameters. By generating a large number of possible failure scenarios and performing parallel simulations, the method can more accurately assess the overall reliability of the system.

[0149] Finally, in terms of computing resource usage, the new method only requires 32 CPU cores, which is 50% less than the traditional method. This improvement in resource utilization efficiency mainly comes from the reduction in the number of features and the optimization of the algorithm, making the entire analysis process lighter and more efficient.

[0150] It is worth noting that the p-values ​​for all improvement indicators are less than 0.001, indicating that these improvements are extremely significant in statistics and are unlikely to be caused by random factors.

[0151] In summary, the power system reliability analysis method proposed in this invention has made significant progress in efficiency, accuracy, resource utilization and other aspects. It can not only process large-scale data more quickly, but also predict potential faults and evaluate system reliability more accurately. This method provides a powerful tool for power system operation management and decision support, which is expected to significantly improve the safety and stability of the power grid, reduce operating costs, and ultimately provide users with a more reliable power supply.

[0152] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for analyzing reliability of a power system, characterized in that: include, Collect power system data and preprocess it, and extract features from the preprocessed data; Building a first model based on the feature-extracted data, and obtaining a first result according to the first model; Constructing a second model according to the real-time data and the first result to obtain a second result, and optimizing the information gain sensitivity parameter in the first model using the second result; Determine the final system reliability index according to the optimized first model and the second model; The constructing of the first model comprises: Construct a decision tree based on the information gain criterion and evaluate feature importance; Ensemble the decision trees and ensure that each decision tree uses a random subset of features; Create composite features through feature engineering and weighted fusion random forest model; The construction of the second model includes: Establishing a basic simulation model and integrating the real-time data and the first result; Generate fault scenarios based on current system status and load changes using dynamic sampling algorithms; Simulating the fault scenarios simultaneously and calculating the reliability index under each fault scenario; The power system data includes voltage, current, temperature, humidity and vibration; The decision tree uses information gain as the splitting criterion and selects the most discriminative feature to divide the data. The information gain is calculated as follows: ， Among them, H(T) is the entropy of node T, a is the feature, T v is the subset when feature a takes value v, values(a) is all the values ​​of feature a, and IG() is the information gain of the decision tree; is the entropy value of the subset when the feature value is v; The decision tree ensemble is represented by the following formula: ， Among them, f b (x) is a single decision tree, B is the number of decision trees, L b is the set of leaf nodes of decision tree b, c l is the predicted value of leaf node l, I() is the indicator function, R l is the feature space region corresponding to leaf node l; Is a single decision tree for the input feature x The prediction output function of The feature engineering is shown in the following formula: ， Among them, φ k is the feature conversion function, h k is an aggregate function; In X is the original feature vector that needs to be transformed by feature engineering; The decision tree and the feature engineering are integrated to form a first model, as shown in the following formula: ， Among them, MDFM(X) is the output of the multi-dimensional fault mode model, X is the input feature vector, α is the model fusion parameter, B is the number of decision trees in the random forest, β is the internal balance parameter of the decision tree, N is the number of features, and w i is the weight of the i-th feature, f b (x i ) is the prediction result of the bth decision tree for the i-th feature, γ is the information gain sensitivity parameter, IG b is the information gain of the bth decision tree, M is the total number of all decision trees, K is the number of composite features after feature engineering, and g k (X) is the function of the kth composite feature; is the contribution measure of the jth feature to data classification; The dynamic sampling algorithm includes: Construct a weighted average framework as shown below: ； For each scenario, determine the reliability of all critical components , as shown below: ； Analyze system reliability over time for each scenario , as shown below: ； in, R sys is the system reliability index, N is the number of dynamically generated fault scenarios, M is the number of components in the system, ω i is the weight of the i-th scenario, P f,j is the failure probability of the jth component, obtained from the failure mode model, I c,j is the importance coefficient of component j, λ j is the failure rate of component j, t i is the simulation time of the i-th scenario; is the failure rate, is the failure rate of the jth component, is the simulation time of the ith scenario, e is the base of natural logarithms.

2. The power system reliability analysis method according to claim 1, characterized in that: The feature extraction extracts data points that have an impact on system reliability from the original signal according to importance through the self-editor and mutual information, as shown in the following formula: ， Among them, F advanced is the high-level extracted feature, X is the original data matrix, W1 and W2 are the weight matrices of the autoencoder, b1 and b2 are the bias vectors, σ is the sigmoid activation function, ReLU is the rectified linear unit activation function, m is the number of original features, and λ j is the attenuation factor of the jth feature, H(X j ) is the entropy of the jth feature, H(Y) is the entropy of the target variable Y, H(X j , Y) is X j The joint entropy of and Y.

3. The power system reliability analysis method according to claim 2, characterized in that: The step of optimizing the parameter of the first model using the second result as an information gain sensitivity parameter comprises the following steps: Define the search range of the information gain sensitivity parameter and discretize it with a fixed step size; For each information gain sensitivity parameter, updating the first model and calculating the result; generating failure scenarios through a second model and calculating average system reliability; Compare different information gain sensitivity parameter values ​​and select the information gain sensitivity parameter with the highest average system reliability; Based on the information gain sensitivity parameter with the highest system reliability, the step size is reduced and the optimization steps are repeated to determine the optimal information gain sensitivity parameter; The optimal information gain sensitivity parameter is the information gain sensitivity value that maximizes the system reliability index under the current system state.

4. A power system reliability analysis system, based on the power system reliability analysis method according to any one of claims 1 to 3, characterized in that: include, The data acquisition module is used to collect and pre-process the power system data and extract features from the pre-processed data; A first model building module, used to build a first model based on the preprocessed data, and determine a first result according to the first model; A second model building module, used to build a second model according to the real-time data and the first result to obtain a second result; An optimization module, configured to optimize the parameters in the first model using the second result, and determine a final system reliability index according to the optimized first model and the second model; The constructing of the first model comprises: Construct a decision tree based on the information gain criterion and evaluate feature importance; Ensemble the decision trees and ensure that each decision tree uses a random subset of features; Create composite features through feature engineering and weighted fusion random forest model; The construction of the second model includes: Establishing a basic simulation model and integrating the real-time data and the first result; Generate fault scenarios based on current system status and load changes using dynamic sampling algorithms; The fault scenarios are simulated simultaneously, and the reliability index under each fault scenario is calculated.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the power system reliability analysis method described in any one of claims 1 to 3 are implemented.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the power system reliability analysis method described in any one of claims 1 to 3 are implemented.

Images