Power grid operation risk distribution monitoring method, system, equipment and medium

Abstract

The invention relates to the technical field of power system safety monitoring, and discloses a power grid operation risk distribution monitoring method, system and device and a medium, and the method comprises the steps: obtaining sample data, carrying out the fractal-dimension preprocessing, and providing reliable basic data for subsequent steps; a risk frequency sequence is established, so that the occurrence frequency of each risk dimension can be intuitively known; each risk dimension is monitored in combination with the first risk distribution monitoring strategy, and potential risks can be found in time; risk coupling operation is carried out on monitoring results of all risk dimensions, and integrated analysis of risks is realized; a risk abnormity criterion index is preset, so that a clear standard is provided for risk judgment; and finally, according to the real-time data of the target power grid, power grid operation risk distribution monitoring is carried out in combination with the risk coupling operation and the risk abnormity criterion index, and dynamic and real-time monitoring of the power grid operation risk is realized. According to the invention, an active prevention and control scheme with high precision and strong interpretability is provided for power grid operation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system safety monitoring, and in particular to a power grid operation risk distribution monitoring method, system, device and medium. BACKGROUND

[0002] In the operation and maintenance process of the power grid, operation risk monitoring is a key link to ensure the stable operation of the power grid and the safety of the operation personnel. The operation environment of the power grid is complex and changeable, and there are many uncertain factors, which may lead to the occurrence of operation risk. With the deepening of the construction of the smart grid, the scale and complexity of the power system operation continue to increase, and the safety hazards existing in the operation and maintenance of the power grid equipment may lead to equipment damage, power interruption and even personal injury accidents. The traditional risk assessment method mainly adopts a threshold alarm mechanism, but the actual operation data has significant non-Gaussian characteristics and multi-dimensional coupling characteristics, and a simple single variable control chart method cannot accurately capture the dynamic evolution law of the risk distribution. Therefore, it is of great practical significance to study a more scientific and effective power grid operation risk distribution monitoring method.

[0003] The traditional power grid operation risk monitoring method mainly relies on artificial experience judgment and simple statistical analysis, and these methods have many limitations when facing the complex and changeable operation environment of the power grid. First of all, artificial experience judgment has significant subjectivity, and different personnel may have different risk assessment results, which makes it difficult to ensure the accuracy and consistency of the assessment results. Secondly, simple statistical analysis methods (such as mean- range control chart) can only preliminarily analyze historical data, and cannot deeply mine the hidden risk characteristics and rules in the data, making it difficult to accurately predict future risks. Such methods usually assume that the risk indicators follow a normal distribution, but the actual power grid operation risk data often presents complex distribution patterns such as sharp peak, thick tail, asymmetry or multi-peak. For example, in the high-risk frequency dimension, risk events may concentrate in certain specific working conditions, while in other dimensions (such as medium-risk frequency, low-risk frequency and acceptable risk frequency), the data distribution may present long-tail characteristics or multi-mode characteristics. This distribution characteristic is seriously inconsistent with the assumption of the traditional method, resulting in a significant reduction in the sensitivity and specificity of risk identification. The traditional control chart assumes that the data follows a normal distribution, while the actual risk indicators often present complex distribution patterns such as asymmetry and multi-peak. This distribution difference makes it difficult for a single model to cover the global risk characteristics. Existing methods mostly process single variable risk dimensions independently, without considering the statistical correlation between multi-dimensional indicators. The power grid operation risk is essentially a multi-dimensional coupled systemic problem, and this correlation is ignored in the traditional method. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the application provides a power grid operation risk distribution monitoring method, system, device and medium, which can solve the limitations of traditional risk monitoring methods in the face of non-Gaussian characteristics and multi-dimensional coupling characteristics of power grid operation risk data, and realize more accurate and comprehensive risk monitoring.

[0006] To solve the above technical problems, the application provides the following technical solutions.

[0007] In a first aspect, the application provides a power grid operation risk distribution monitoring method, comprising:

[0008] Obtaining sample data of a plurality of risk dimensions of a target power grid, and performing dimension-by-dimension preprocessing on the sample data;

[0009] The plurality of risk dimensions include high-risk dimensions, medium-risk dimensions, low-risk dimensions and acceptable-risk dimensions of the power grid;

[0010] Establishing a risk frequency sequence according to the results of the dimension-by-dimension preprocessing;

[0011] Monitoring each risk dimension according to the risk frequency sequence and a first risk distribution monitoring strategy;

[0012] Performing risk coupling operation on the monitoring results of all risk dimensions;

[0013] Predefining risk anomaly criterion indicators;

[0014] According to the real-time data of the target power grid, combining the risk coupling operation and the risk anomaly criterion indicators to monitor the power grid operation risk distribution.

[0015] As a preferred scheme of the power grid operation risk distribution monitoring method, the method comprises:

[0016] Obtaining risk distribution data of each dimension after dimension-by-dimension preprocessing, wherein the risk distribution data of each dimension is a plurality of variable samples of different dimensions;

[0017] Standardizing the risk distribution data of each dimension by using the total planned operation quantity corresponding to each sample;

[0018] Mapping the risk values obtained after standardizing each dimension to a unified dimension to generate a standardized risk frequency sequence.

[0019] As a preferred scheme of the power grid operation risk distribution monitoring method, the first risk distribution monitoring strategy comprises:

[0020] Modeling the non-Gaussian characteristics of single-dimensional risk distribution by using a generalized Gaussian distribution model;

[0021] Solving the shape parameter and the mean and variance by a maximum likelihood estimation algorithm;

[0022] Obtaining the peak thick tail or other non-Gaussian distribution mode of the risk distribution data of each dimension;

[0023] The preset confidence level adopts a quantile function to solve the control boundary.

[0024] As a preferred scheme of the power grid operation risk distribution monitoring method, wherein: the risk coupling operation on the monitoring results of all risk dimensions comprises:

[0025] Establishing a Gaussian mixture model, wherein the Gaussian mixture model comprises setting initial mean values, covariance matrices and mixing coefficients of Gaussian components based on random sampling and uniform initialization strategies;

[0026] In the iterative optimization stage, a pseudo-inverse matrix is used to replace the traditional normal distribution function library.

[0027] This preferred scheme can effectively reduce the computational complexity, improve the model convergence speed, and at the same time ensure the accuracy of risk coupling. Through the Gaussian mixture model, the complex relationship between each risk dimension can be described more finely, providing strong support for subsequent risk assessment and early warning. The application of the pseudo-inverse matrix further enhances the stability and robustness of the model in processing nonlinear and high-dimensional data, making the entire monitoring method more adaptable to the complexity and variability of the actual power grid operation environment.

[0028] As a preferred scheme of the power grid operation risk distribution monitoring method, wherein: the preset risk anomaly criterion index comprises:

[0029] The risk anomaly criterion index is obtained by the squared Mahalanobis distance of different samples from the center of the Gaussian component and the posterior probability of the corresponding Gaussian component;

[0030] Calculating the control limit based on the risk anomaly criterion index;

[0031] According to the control limit, the current sample risk distribution state is judged.

[0032] As a preferred scheme of the power grid operation risk distribution monitoring method, wherein: the preset confidence level adopts a quantile function to solve the control boundary comprises:

[0033] The preset confidence level;

[0034] Calculating the upper quantile and the lower quantile of the risk distribution data distribution of each dimension can determine the confidence interval of the risk distribution of each dimension;

[0035] If the current risk frequency sequence data exceeds the upper limit of the control boundary or is lower than the upper limit of the control boundary, it is determined that a single-dimension risk abnormal event occurs.

[0036] As a preferred solution of the power grid operation risk distribution monitoring method, the control limit based on the risk abnormal criterion index is obtained by a kernel density estimation method.

[0037] In a second aspect, the present application provides a power grid operation risk distribution monitoring system, comprising:

[0038] A data acquisition and processing module is configured to acquire sample data of a plurality of risk dimensions of a target power grid and perform dimension-by-dimension preprocessing on the sample data.

[0039] The plurality of risk dimensions include a high-risk dimension of a power grid, a medium-risk dimension of a power grid, a low-risk dimension of a power grid, and an acceptable-risk dimension of a power grid.

[0040] A sequence establishing module is configured to establish a risk frequency sequence based on the results of the dimension-by-dimension preprocessing.

[0041] A single-dimension monitoring module is configured to monitor each risk dimension based on the risk frequency sequence and a first risk distribution monitoring strategy.

[0042] A coupling module is configured to perform a risk coupling operation on the monitoring results of all risk dimensions.

[0043] An index presetting module is configured to preset a risk abnormal criterion index.

[0044] A multi-dimension detecting module is configured to perform power grid operation risk distribution monitoring based on real-time data of a target power grid, a risk coupling operation, and a risk abnormal criterion index.

[0045] In a third aspect, the present application provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method when executing the computer program.

[0046] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the method.

[0047] Compared with the prior art, the present application has the beneficial effects that: the present application proposes a power grid operation risk distribution monitoring method, obtains sample data and performs dimension-by-dimension preprocessing, thereby providing reliable basic data for subsequent steps; a risk frequency sequence is established, which is helpful to intuitively understand the occurrence frequency of each risk dimension; each risk dimension is monitored by combining the first risk distribution monitoring strategy, so that potential risks can be found in time; the monitoring results of all risk dimensions are subjected to risk coupling operation, thereby realizing integrated analysis of risks; preset risk abnormality criterion indexes are provided, thereby providing clear standards for risk judgment; finally, according to target power grid real-time data, the power grid operation risk distribution monitoring is performed in combination with the risk coupling operation and the risk abnormality criterion indexes, thereby realizing dynamic and real-time monitoring of power grid operation risks. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0049] Figure 1 A method flowchart of a power grid operation risk distribution monitoring method provided by an embodiment of the present application.

[0050] Figure 2 A single-dimension risk distribution monitoring diagram of a power grid operation risk distribution monitoring method provided by an embodiment of the present application, (a) high risk, (b) medium risk, (c) low risk, (d) acceptable risk.

[0051] Figure 3 A multi-dimension risk distribution monitoring diagram of a power grid operation risk distribution monitoring method provided by an embodiment of the present application.

[0052] Figure 4 A multi-dimension sample visual monitoring diagram of a power grid operation risk distribution monitoring method provided by an embodiment of the present application.

[0053] Figure 5 An internal structure diagram of an electronic device of a power grid operation risk distribution monitoring method provided by an embodiment of the present application. DETAILED DESCRIPTION

[0054] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor should fall within the scope of protection of the present application.

[0055] Embodiment 1, reference Figures 1-4 For the first embodiment of the present application, the embodiment provides a power grid operation risk distribution monitoring method, comprising:

[0056] In the prior art, there are some problems, for example, the accuracy and real-time performance of risk monitoring are insufficient, and it is difficult to fully reflect the dynamic changes of power grid operation risk.

[0057] The present application provides a method that can effectively solve the above-mentioned problems. Next, how to realize the power grid operation risk distribution monitoring method will be described in detail in conjunction with multiple embodiments;

[0058] Figure 1 A method flowchart of a power grid operation risk distribution monitoring method is shown, comprising:

[0059] S101, obtaining sample data of several risk dimensions of a target power grid, and performing dimension-by-dimension preprocessing on the sample data;

[0060] It should be noted that in the field of traditional power grid operation risk monitoring, the monitoring method often relies on manual experience judgment and simple statistical analysis. However, these methods have many limitations and shortcomings when dealing with complex and variable power grid operation environments. Due to the complexity and variability of the power grid operation environment, it is often difficult to comprehensively and accurately identify and evaluate potential risk factors by relying solely on manual experience judgment and simple statistical analysis. In addition, these methods also have certain hysteresis in risk warning and response, and cannot respond to sudden situations in a timely and effective manner. Therefore, in order to improve the accuracy and efficiency of power grid operation risk monitoring, it is necessary to introduce more advanced technical means and methods to realize comprehensive and real-time monitoring and early warning of power grid operation risk.

[0061] It should be noted that according to the purpose of the present application, the present application obtains some sample data related to power grid operation at the beginning, and these data include data of different dimensions. In actual application, different dimensions may produce different dimensional risks. Therefore, how to simultaneously monitor different dimensional operation risks of the entire power grid operation scenario is the biggest problem to be solved at this stage.

[0062] In embodiments of the present application, the several risk dimensions include a high grid risk dimension, a medium grid risk dimension, a low grid risk dimension, and an acceptable grid risk dimension.

[0063] In some specific embodiments, the differentiation of the several risk dimensions can be achieved through different monitoring modules, each of which monitors a specific risk dimension. For example, for the high grid risk dimension, a dedicated high-risk monitoring module can be set up, which can monitor the data changes of the high-risk operation area in real time and trigger the early warning mechanism as soon as abnormal data is detected, notifying relevant personnel to take timely measures. Similarly, for the medium grid risk dimension, the low grid risk dimension, and the acceptable grid risk dimension, corresponding monitoring modules can also be set up respectively to achieve comprehensive coverage and real-time monitoring of the entire grid operation scenario.

[0064] In some specific embodiments, the differentiation of the several risk dimensions can also be achieved by setting threshold ranges. Each risk dimension corresponds to a specific threshold range, and the risk dimension to which the sample data belongs is determined according to the range to which the risk value of the sample data belongs. For example, the threshold range of the high risk dimension can be set to a risk value greater than or equal to a certain specific value, while the medium risk dimension, the low risk dimension, and the acceptable risk dimension correspond to different threshold ranges respectively. This method can more intuitively reflect the risk level of different risk dimensions, helping relevant personnel to more accurately judge the risk situation and take corresponding measures.

[0065] In embodiments of the present application, the dimension-specific preprocessing can include, but is not limited to, cleaning, denoising, and normalizing sample data, etc. to improve data quality and the accuracy of subsequent analysis. The cleaning step aims to eliminate invalid or abnormal data to ensure the representativeness of the sample data; the denoising operation helps to reduce random fluctuations in the data and highlight the risk characteristics; and the normalization process can convert data of different dimensions to the same dimension, facilitating subsequent risk assessment and comparison. Through these preprocessing means, the present application can more effectively mine and utilize the information in the sample data, providing a solid foundation for subsequent risk monitoring.

[0066] It should be noted that obtaining sample data of several risk dimensions of the target power grid and performing dimension-by-dimension preprocessing on the sample data can ensure the accuracy and consistency of the data and provide a reliable basis for subsequent risk monitoring. At the same time, dimension-by-dimension preprocessing also helps to reduce noise and redundant information in the data, improving the efficiency and accuracy of data analysis. In specific implementation, appropriate preprocessing methods and parameters can be selected according to the characteristics and needs of different risk dimensions to achieve the best data processing effect. In addition, the data acquisition and processing module, as an important part of the power grid operation risk distribution monitoring system of the present application, undertakes the key task of data collection and preprocessing, and the stability and accuracy of its performance are directly related to the effect of the whole monitoring system. Therefore, in actual application, attention should be paid to the design and optimization of the data acquisition and processing module to ensure that it can meet the actual monitoring needs.

[0067] S102, establishing a risk frequency sequence according to the results of the dimension-by-dimension preprocessing;

[0068] It should be noted that the processed data cannot be directly used, so the processed data needs to be further operated.

[0069] In some specific embodiments, establishing a risk frequency sequence according to the results of the dimension-by-dimension preprocessing can further reveal the distribution law and trend of each risk dimension over time. The establishment of the risk frequency sequence helps to intuitively show the occurrence frequency of each risk dimension in different time periods, thereby providing important decision basis for risk managers. Through analysis of the risk frequency sequence, risk high-incidence periods and low-incidence periods can be identified, and corresponding risk prevention and control measures can be taken to optimize resource allocation and improve risk response efficiency.

[0070] In some specific embodiments, statistical methods or data mining techniques can be used to establish the risk frequency sequence. For example, a time series analysis model can be used to fit and predict historical data, thereby obtaining the risk frequency of each risk dimension in future time periods. In addition, machine learning algorithms can also be combined to perform pattern recognition and trend prediction on the risk frequency sequence, further improving the accuracy and foresight of risk management. By establishing the risk frequency sequence, not only can a more intuitive and comprehensive risk view be provided for risk managers, but also strong support can be provided for formulating scientific and reasonable risk prevention and control strategies.

[0071] In the embodiment of the present application, establishing a risk frequency sequence according to the results of the dimension-by-dimension preprocessing includes:

[0072] Obtaining risk distribution data of each dimension after dimension-by-dimension preprocessing, the risk distribution data of each dimension being a plurality of variable samples of different dimensions;

[0073] The risk distribution data of each dimension is standardized by using the total planning operation quantity corresponding to each sample;

[0074] The risk value obtained after standardizing each dimension is mapped to a unified dimension to generate a standardized risk frequency sequence. Figure 2

[0075] Suppose the risk distribution data X m×n There are m samples and n-dimensional variables, and the original risk distribution data of each dimension is standardized by using the total planning operation quantity T(i), i = 1, 2,..., m corresponding to each sample. The risk value of each dimension is mapped to a unified dimension to generate a standardized risk frequency sequence, and the formula is as follows:

[0076]

[0077]

[0078] Where j = 1, 2,..., 4 represents the jth risk variable, t j (i) represents the ith sample value of the jth variable, x j (i) represents its risk frequency value, E(x j ) and D(x j ) represent the mean and variance of the jth variable, respectively.

[0079] It should be noted that the risk frequency sequence is established according to the results of the dimension-by-dimension preprocessing, which can standardize the risk data of different dimensions, so that the risk data of each dimension can be compared and analyzed under the same dimension. This processing method helps to eliminate the dimensional differences between different dimensions, so that risk managers can more clearly understand the distribution and trend of each risk dimension. At the same time, by establishing the risk frequency sequence, more accurate and reliable data basis can be provided for subsequent risk monitoring and early warning.

[0080] S103, according to the risk frequency sequence combined with the first risk distribution monitoring strategy to monitor each risk dimension;

[0081] It should be noted that after obtaining the previous sequence, the monitoring of each risk dimension can be started. During the monitoring process, the present application adopts the first risk distribution monitoring strategy, which is an effective monitoring method for different risk dimensions based on in-depth analysis and mining of historical data. Through real-time monitoring and analysis of the risk frequency sequence, abnormal conditions of each risk dimension can be found in time, providing timely and accurate information support for subsequent risk judgment and response measures.

[0082] ​In specific implementation, the first risk distribution monitoring strategy can be customized and optimized according to the characteristics and needs of different risk dimensions to ensure that it can maximize the monitoring effect. For example, for high-risk dimensions, more stringent and frequent monitoring strategies can be adopted to ensure that potential high-risk events can be discovered and addressed in a timely manner.

[0083] In some specific embodiments, the first risk distribution monitoring strategy can be implemented through expert systems or intelligent algorithms that can automatically identify abnormal changes in each risk dimension based on risk frequency sequences and historical data, and trigger corresponding early warning mechanisms. The early warning mechanism can include sending early warning information to relevant personnel, starting emergency plans, and other measures to ensure that potential risks can be addressed in a timely and effective manner. In addition, the first risk distribution monitoring strategy can also combine real-time data and power grid operation status to dynamically evaluate and predict risks, providing a more comprehensive and accurate risk view for risk managers, which helps them make more scientific and reasonable decisions.

[0084] In the embodiments of the present application, the first risk distribution monitoring strategy includes:

[0085] The single-dimensional risk distribution is modeled using a non-Gaussian characteristic of a generalized Gaussian distribution model;

[0086] The shape parameter and the mean and variance are solved by a maximum likelihood estimation algorithm;

[0087] Obtain the peak thick tail or other non-Gaussian distribution morphology of the risk distribution data of each dimension;

[0088] The preset confidence level uses a quantile function to calculate the control boundary.

[0089] In the embodiments of the present application, the preset confidence level uses a quantile function to calculate the control boundary, which includes:

[0090] The preset confidence level;

[0091] The upper and lower quantiles of the risk distribution data distribution of each dimension can determine the confidence interval of the risk distribution of each dimension;

[0092] If the current risk frequency sequence data exceeds or is lower than the upper limit of the control boundary, it is determined that there is a single-dimensional risk abnormal event.

[0093] It is noted that the non-Gaussian modeling of single-dimensional risk distribution using the generalized Gaussian distribution model is to use the mathematical model of generalized Gaussian distribution (GGD) to describe the data distribution of a single risk dimension in power grid operations. Traditional methods usually assume that the data follows a normal distribution, but the actual power grid operation risk data often presents complex distribution patterns such as sharp peak, thick tail, asymmetry or multi-peak. GGD can more accurately capture these non-Gaussian characteristics, thereby providing more accurate risk assessment.

[0094] It is noted that the maximum likelihood estimation (MLE) is a commonly used parameter estimation method for inferring the parameter value of the probability distribution that best describes the data set from sample data. In this context, MLE is used to determine the optimal shape parameter, mean and variance of the GGD model to best fit the actual observed risk distribution data.

[0095] It is noted that obtaining the sharp peak, thick tail or other non-Gaussian distribution patterns of risk distribution data in each dimension emphasizes the identification and quantification of special distribution characteristics in power grid operation risk data, such as sharp peak and thick tail phenomenon (i.e. the data is concentrated around the mean, but there are also many extreme values away from the mean). These characteristics do not conform to the traditional normal distribution assumption, so special treatment is needed to improve the accuracy of risk assessment.

[0096] It is noted that the pre-set confidence level uses quantile function to calculate the control boundary. The pre-set confidence level refers to a pre-set probability threshold used to define the extent or range that the application considers a result to be "normal". For example, a 95% confidence level means that the application accepts 95% of the observed values within the range defined by the application.

[0097] The quantile function is a function used to determine the position of data points at a given probability. In this context, it is used to calculate the upper and lower quantiles of the risk distribution data in each dimension to define the confidence interval of the risk distribution in each dimension. If the risk frequency sequence data exceeds this interval, it is considered to be an abnormal event.

[0098] It is noted that if the current risk frequency sequence data exceeds or is below the upper limit of the control boundary, it is determined that the single-dimensional risk abnormal event. This condition specifies how to make risk judgments based on the control boundary established in the previous steps. Specifically, when the risk frequency data in a certain dimension exceeds the control boundary (including the upper and lower limits) calculated according to the confidence level, it is considered that the dimension has an abnormal situation and needs to be further investigated or measures are taken.

[0099] In the embodiments of the present application, the single-dimensional risk frequency data is modeled by using a generalized Gaussian distribution (GGD), and the probability density function of the GGD is:

[0100]

[0101] wherein μ represents the mean of the GGD, σ 2 and γ represent the variance and shape parameter, and the constants a and b corresponding to the four dimensions are calculated as shown below:

[0102]

[0103] wherein, is the known gamma function.

[0104] Further, for the jth dimension variable, the method of moment estimation is used, and a monotonic function of the shape parameter γ j of the jth dimension is constructed based on the absolute moment to inversely solve the value of γ j , and then the GGD function of the risk data of the dimension is solved:

[0105]

[0106] wherein σ j is the standard deviation of the risk data of the jth dimension, and E 2 [|x j |] represents the square of the mean of the absolute value of the data, and the corresponding calculation formula is as follows:

[0107]

[0108] wherein μ j is the mean of the risk data of the jth dimension.

[0109] Finally, the obtained parameter values are substituted into formula (3) to obtain the probability density function f(·) of the GGD of the jth dimension risk variable, and the parameter estimation of the GGD is completed.

[0110] Further, the risk safety threshold boundary is solved: based on the fitted GGD distribution, a preset confidence level α is set, and the upper quantile and the lower quantile of the data distribution are calculated to determine the confidence interval of the risk distribution of each dimension, i.e., the safety threshold boundary, and the solving process is as shown below:

[0111]

[0112] Finally, if the current risk frequency data exceeds the upper limit of the control boundary or is lower than the lower limit of the control boundary, it is determined as a single-dimensional risk abnormal event.

[0113] It should be noted that monitoring each risk dimension according to the risk frequency sequence combined with the first risk distribution monitoring strategy can capture the dynamic changes of each risk dimension in power grid operation in real time and discover potential risk hazards in time. Through detailed monitoring and analysis of the risk frequency sequence, the system can automatically compare the current risk data with the preset safety threshold boundary, and once the data exceeds or is lower than the control boundary, the early warning mechanism is triggered immediately to remind the risk manager to take corresponding measures. This way not only improves the efficiency and accuracy of risk monitoring, but also greatly shortens the risk response time, providing a strong guarantee for the safe and stable operation of power grid operation.

[0114] S104, risk coupling operation is performed on the monitoring results of all risk dimensions;

[0115] It should be noted that after the single-dimension risk assessment, multi-dimension comprehensive risk acquisition needs to be performed. It should be noted that all single-dimension risk monitoring results cannot be directly superimposed because different dimensions have certain correlations with each other, which may cause mutual influence and superposition effect between risks. Therefore, the present application proposes a risk coupling operation method, which can comprehensively consider the interaction between risk dimensions to obtain more accurate and comprehensive comprehensive risk assessment results.

[0116] In some specific embodiments, the risk coupling operation can be realized using risk matrix method, Bayesian network, fuzzy comprehensive evaluation method or neural network model, etc. These methods can construct complex risk relationship models based on the monitoring results of each risk dimension and their correlations. Through calculation and analysis, the coupling degree between each risk dimension and the comprehensive risk level can be obtained.

[0117] When the risk matrix method is used to realize, the specific steps can be as follows:

[0118] First, determine the severity and possibility levels of each risk dimension to construct a risk matrix;

[0119] Then, according to the monitoring results of each risk dimension, locate the corresponding risk level in the risk matrix;

[0120] Next, comprehensively consider the mutual influence between each risk dimension to adjust the risk level;

[0121] Finally, according to the adjusted risk level, the comprehensive risk assessment result is obtained. This process fully considers the correlations between each risk dimension, making the assessment result more accurate and comprehensive.

[0122] In the embodiment of the present application, the risk coupling operation on the monitoring results of all risk dimensions includes:

[0123] establish a Gaussian mixture model, the Gaussian mixture model includes setting initial mean, covariance matrix and mixing coefficient of Gaussian component based on random sampling and uniform initialization strategy;

[0124] In the iterative optimization phase, the pseudo-inverse matrix is used to replace the traditional normal distribution function library.

[0125] Specifically, first, randomly select K samples from the data set as the initial Gaussian component mean μ k (0) , to ensure that the initial center point covers the data distribution characteristics, and second, uniformly initialize the mixing coefficient weight π k (0) = 1 / K, and satisfy the following probability normalization constraint:

[0126]

[0127] Second, use the covariance matrix ∑ k of the k = 1, 2, … K Gaussian component to solve the posterior probability λ ik of the corresponding Gaussian component:

[0128]

[0129] Next, according to the solved posterior probability, further calculate the effective sample number N k of each component, and update the mixing coefficient:

[0130]

[0131] Second, use the posterior probability to calculate the new mean value, realize the adaptive adjustment of the data distribution center, and update the covariance iteratively:

[0132]

[0133] Finally, superimpose the weighted probability density of each Gaussian component to obtain the global probability distribution of the i-th risk sample:

[0134]

[0135] It should be noted that the risk coupling operation of the monitoring results of all risk dimensions through the probability distribution can realize the comprehensive assessment of the power grid operation risk. Since power grid operation involves multiple risk dimensions, each dimension may have different probability distribution characteristics. By coupling the monitoring results of each dimension, the mutual influence between each risk dimension can be considered comprehensively, so as to obtain a more accurate and comprehensive risk distribution. This method not only improves the accuracy of risk monitoring, but also provides strong support for risk management and decision-making of power grid operation.

[0136] S105, preset risk anomaly criterion index;

[0137] It should be noted that after obtaining the coupling result, it is necessary to judge whether there is a risk anomaly, and therefore, the preset risk anomaly criterion index is used to quantitatively evaluate the comprehensive risk level after coupling, so as to identify and handle potential risk anomalies in a timely manner. These indexes can be set based on historical data, expert experience or industry standards to ensure that the actual situation of the risk can be reflected and clear decision basis can be provided for risk managers. For example, when a certain comprehensive risk value exceeds a certain threshold, it is determined that there is a risk anomaly, and immediate measures need to be taken for intervention. In this way, the present application can realize dynamic monitoring and early warning of power grid operation risk, and provide strong guarantee for safe and stable operation of the power grid.

[0138] In the embodiment of the present application, the preset risk anomaly criterion index includes:

[0139] The risk anomaly criterion index is obtained by the squared Mahalanobis distance of different sample distances from the center of the Gaussian component and the posterior probability of the corresponding Gaussian component;

[0140] The control limit based on the risk anomaly criterion index is calculated;

[0141] The current sample risk distribution state is judged according to the control limit.

[0142] In the embodiment of the present application, the control limit based on the risk anomaly criterion index is calculated by a kernel density estimation method. As shown in Figure 3 .

[0143] Specifically, based on the obtained joint probability density and posterior probability, the global BID monitoring index of the i-th sample is calculated using the Mahalanobis distance:

[0144]

[0145] Wherein, D ik represents the squared Mahalanobis distance of the i-th sample x i to the k-th Gaussian component C k center, and the calculation formula is:

[0146] D ik =(x i -μ k ) T ∑ k -1 (x i -μ k ) (14)

[0147] The control limit of the BID monitoring index is denoted as BIDlmt, which is obtained by presetting a significance level ξ and then using a kernel density estimation method. When the BID index is below the control limit, it is considered that the current sample risk distribution is normal; otherwise, it is considered that the risk distribution is abnormal, an alarm is sent out, and then a multi-dimensional risk distribution monitoring signal is output.

[0148] It should be noted that the preset risk anomaly criterion index can provide a quantitative standard for accurately evaluating the comprehensive risk level of the coupled grid operation risk. The setting of this index enables the risk manager to quickly determine whether there is risk anomaly in the current grid operation according to the specific data results, so as to timely take corresponding risk management measures. This not only enhances the scientificity and accuracy of risk management, but also effectively improves the safety and stability of grid operation. By presetting the risk anomaly criterion index, the present application realizes fine management and intelligent monitoring of grid operation risk, and provides a solid technical support for safe operation of the grid.

[0149] S106, according to the target grid real-time data, combining the risk coupling operation and the risk anomaly criterion index, the grid operation risk distribution monitoring is performed.

[0150] In summary, the present application proposes a grid operation risk distribution monitoring method, which obtains sample data and performs dimension-by-dimension preprocessing, providing reliable basic data for subsequent steps; establishes a risk frequency sequence, which helps to intuitively understand the occurrence frequency of each risk dimension; combines the first risk distribution monitoring strategy to monitor each risk dimension, which can timely discover potential risks; performs risk coupling operation on the monitoring results of all risk dimensions, realizing integrated analysis of risks; presets the risk anomaly criterion index, which provides a clear standard for risk determination; finally, according to the target grid real-time data, combining the risk coupling operation and the risk anomaly criterion index, the grid operation risk distribution monitoring is performed, realizing dynamic and real-time monitoring of grid operation risk. The present application provides an active prevention and control scheme with high precision and strong explanation for grid operation.

[0151] Embodiment 2, in a preferred embodiment, the following embodiments are only used to more clearly illustrate the technical solutions of the present application, and cannot be used to limit the protection scope of the present application, the application object is certain grid company's risk distribution data in a certain year, there are 4-dimensional risk distribution of high risk, medium risk, low risk and acceptable risk and 368 samples.

[0152] Step 1, independent single-dimensional risk distribution monitoring: based on the generalized Gaussian distribution (GGD), the single-dimensional risk dynamic monitoring is realized for each dimensional risk distribution. Through the preprocessing of multi-dimensional data such as high risk, medium risk, low risk and acceptable risk of power grid, the quantifiable risk frequency sequence is generated. The GGD model is used for non-Gaussian modeling of single-dimensional risk distribution. The shape parameter, mean and variance are solved by maximum likelihood estimation algorithm. The risk data of sharp peak and thick tail or other non-Gaussian distribution form is accurately described. Finally, the control boundary is calculated by using the quantile function with pre-set significance level, and the monitoring of independent single-dimensional risk distribution is realized.

[0153] (1) Risk data preprocessing: risk distribution data X 368×4 There are 368 samples and 4-dimensional variables. The total planned operation quantity T(i), i = 1, 2,..., 368 corresponding to each sample is used to standardize the original risk distribution data of each dimension, and the risk value of each dimension is mapped to a unified dimension to generate a standardized risk frequency sequence. The formula is as follows:

[0154]

[0155] Where j = 1, 2,..., 4 represents the jth-dimensional risk variable, t j (i) represents the ith sample value of the jth variable, x j (i) represents its risk frequency value, E(x j ) and D(x j ) represent the mean and variance of the jth variable, respectively.

[0156] (2) GGD fitting and parameter estimation: the single-dimensional risk frequency data is modeled by using the generalized Gaussian distribution (GGD), and its probability density function is:

[0157]

[0158] Where μ represents the mean of GGD, σ 2 and γ represent the variance and shape parameter, and the calculation values of the constants a, b under 4 dimensions are as follows:

[0159]

[0160] Where, is the known gamma function.

[0161] For the jth variable, the moment estimation method is used. Based on the absolute moment, a monotonic function of the shape parameter γ j of this dimension is constructed to inversely solve γ j value. The shape parameters of each dimension are calculated as follows:

[0162] [γ1, γ2, γ3, γ4] = [0.8550, 0.4330, 1.1860, 0.8400] (19)

[0163] (3) Solve the risk safety threshold boundary: based on the fitted GGD distribution, set the preset confidence level a as 0.95, and calculate the upper and lower quantiles of the data distribution to determine the confidence interval of each dimension risk distribution, that is, the safety threshold boundary, the solving process is as follows:

[0164]

[0165] The lower and upper limits of the control boundary are calculated as:

[0166] L CL = [-2.1820, -2.2219, -2.2581, -2.3491] (21)

[0167] U CL = [2.1820, 2.2219, 2.2581, 2.3491] (22)

[0168] Finally, if the current risk frequency data exceeds the upper limit of the control boundary or is lower than the upper limit of the control boundary, it is determined as a single-dimensional risk abnormal event.

[0169] Step 2, solve the probability density value of the overall data based on the GMM model: first, based on random sampling and uniform initialization strategy, set the initial mean, covariance matrix and mixing coefficient of Gaussian component, ensure that the model covers the multimodal distribution characteristics; in the iterative optimization stage, use the pseudo-inverse matrix instead of the traditional normal distribution function library, calculate the posterior probability after E step, update the mixing coefficient and regularized covariance to finally output the sample joint probability density function.

[0170] First, randomly select 3 samples from the data set as the initial Gaussian component mean μ k (0) , ensure that the initial center point covers the data distribution characteristics, secondly, uniformly initialize the mixing coefficient weight π k (0) = 1 / 3, and satisfy the following probability normalization constraint:

[0171]

[0172] Secondly, use the covariance matrix ∑ k of the k = 1, 2, 3 Gaussian components to solve the posterior probability λ ik of the corresponding Gaussian component:

[0173]

[0174] Next, according to the desired posterior probability, the effective number of samples N of each component is calculated. k , and update the mixing coefficient:

[0175]

[0176] Secondly, the new mean is calculated by weighting the posterior probability to achieve adaptive adjustment of the data distribution center.

[0177]

[0178] Finally, the weighted probability density of each Gaussian component is superimposed to obtain the global probability density value corresponding to each sample. After normalization, the data is as follows:

[0179]

[0180] Step 3: Output the multidimensional risk distribution monitoring signal based on the BID indicator: Based on the joint probability density of the overall data samples solved in the second part, calculate the BID indicator value of each sample, and use the kernel density estimation method to obtain the corresponding control limit, thereby completing the monitoring of the multidimensional risk distribution.

[0181] Based on the joint probability density and posterior probability, the global BID monitoring index of the i-th sample is calculated using the Mahalanobis distance. The results are as follows:

[0182]

[0183] The significance level is set to ξ = 0.05, and the control limit BIDlmt is calculated by the kernel density estimation method to be 6.7837. Figure 4 shown.

[0184] Example 3, reference Figure 5 This embodiment also provides a power grid operation risk distribution monitoring system, including:

[0185] The data acquisition and processing module is used to obtain sample data of several risk dimensions of the target power grid and perform dimensional preprocessing on the sample data;

[0186] Several risk dimensions include a high-risk dimension for the power grid, a medium-risk dimension for the power grid, a low-risk dimension for the power grid, and an acceptable risk dimension for the power grid;

[0187] A sequence building module is used to build a risk frequency sequence based on the results of dimension preprocessing;

[0188] A single-dimension detection module is used to monitor each risk dimension according to the risk frequency sequence in combination with the first risk distribution monitoring strategy;

[0189] A coupling module is configured to perform a risk coupling operation on the monitoring results of all risk dimensions.

[0190] An index presetting module is configured to preset a risk anomaly criterion index.

[0191] A multi-dimensional detection module is configured to perform power grid operation risk distribution monitoring according to target power grid real-time data, in combination with the risk coupling operation and the risk anomaly criterion index.

[0192] The above modules can be embedded in or independent of a processor in an electronic device in hardware form, or stored in a memory in the electronic device in software form, so as to be called and executed by the processor to perform the operations corresponding to the above modules.

[0193] The embodiment also provides an electronic device, which can be a terminal, and an internal structure diagram of the electronic device can be as shown in Figure 5 The electronic device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic 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 operating system and the computer program in the non-volatile storage medium to run. The communication interface of the electronic device is configured to perform wired or wireless communication with an external terminal. The wireless communication can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The computer program is executed by the processor to implement a power grid operation risk distribution monitoring method. The display screen of the electronic device can be a liquid crystal display screen or an electronic ink display screen. The input device of the electronic device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the electronic device, or an external keyboard, touchpad or mouse, etc.

[0194] The embodiment also provides a computer readable storage medium having a computer program stored thereon. The computer program is executed by the processor to implement the following steps:

[0195] Obtain sample data of a plurality of risk dimensions of a target power grid, and perform dimension-by-dimension preprocessing on the sample data;

[0196] The plurality of risk dimensions include a high-risk dimension of a power grid, a medium-risk dimension of a power grid, a low-risk dimension of a power grid and an acceptable-risk dimension of a power grid.

[0197] Establish a risk frequency sequence according to the results of the dimension-by-dimension preprocessing;

[0198] Monitor each risk dimension according to the risk frequency sequence and a first risk distribution monitoring strategy;

[0199] The monitoring results of all risk dimensions are subjected to a risk coupling operation;

[0200] A preset risk anomaly criterion index;

[0201] According to target power grid real-time data, the power grid operation risk distribution is monitored in combination with the risk coupling operation and the risk anomaly criterion index.

[0202] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application 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 application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and all modifications and replacements should be covered in the scope of the claims of the present application.

[0203] Although the preferred embodiments of the present application have been described, those skilled in the art can make further changes and modifications to the embodiments once they understand the basic inventive concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0204] Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and changes of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and changes.

Claims

1. A method for monitoring the risk distribution of power grid operations, characterized in that: include: Obtaining sample data of several risk dimensions of the target power grid and performing dimensional preprocessing on the sample data; The risk dimensions include a high-risk dimension for a power grid, a medium-risk dimension for a power grid, a low-risk dimension for a power grid, and an acceptable risk dimension for a power grid; Establish risk frequency sequence based on the results of dimension preprocessing; Monitor each risk dimension according to the risk frequency sequence in combination with the first risk distribution monitoring strategy; Conduct risk coupling operations on the monitoring results of all risk dimensions; Preset risk abnormality judgment indicators; Based on the real-time data of the target power grid, the risk distribution of power grid operations is monitored in combination with risk coupling operations and risk anomaly judgment indicators.

2. A method for monitoring the distribution of power grid operation risks according to claim 1, characterized in that: The step of establishing a risk frequency sequence based on the result of the dimension preprocessing includes: Obtain risk distribution data for each dimension after dimensional preprocessing, wherein the risk distribution data for each dimension is variable samples of several different dimensions; The risk distribution data of each dimension is standardized using the total number of planned operations corresponding to each sample; The risk values ​​obtained after standardization of each dimension are mapped to a unified dimension to generate a standardized risk frequency series.

3. A method for monitoring the distribution of power grid operation risks according to claim 2, characterized in that: The first risk distribution monitoring strategy includes: The generalized Gaussian distribution model is used to model the non-Gaussian characteristics of the unidimensional risk distribution; The shape parameters, mean and variance are solved by the maximum likelihood estimation algorithm; Obtain peaked, fat-tailed, or other non-Gaussian distribution forms of risk distribution data for each dimension; The control boundaries are solved using the quantile function at the preset confidence level.

4. A method for monitoring the distribution of power grid operation risks according to claim 3, characterized in that: The risk coupling operation for the monitoring results of all risk dimensions includes: Establishing a Gaussian mixture model, wherein the Gaussian mixture model includes setting the initial mean, covariance matrix and mixing coefficient of the Gaussian components based on random sampling and uniform initialization strategy; In the iterative optimization stage, the pseudo-inverse matrix is ​​used to replace the traditional normal distribution function library.

5. A method for monitoring the distribution of power grid operation risks according to claim 4, characterized in that: The preset risk abnormality judgment indicators include: The risk anomaly criterion index is obtained by the square of the Mahalanobis distance of different samples from the center of the Gaussian component and the posterior probability of the corresponding Gaussian component; Calculating control limits based on the risk anomaly criterion indicator; The current sample risk distribution status is determined based on the control limits.

6. A method for monitoring the distribution of power grid operation risks according to claim 5, characterized in that: The preset confidence level uses the quantile function to solve the control boundary, including: Preset confidence level; Calculating the upper and lower quantiles of the risk distribution data for each dimension can determine the confidence interval of the risk distribution for each dimension; If the current risk frequency series data exceeds the upper limit of the control boundary or is lower than the upper limit of the control boundary, it is determined to be a single-dimensional risk abnormal event.

7. A method for monitoring the distribution of power grid operation risks according to claim 6, characterized in that: The calculation is based on the control limit of the risk anomaly criterion indicator obtained by a kernel density estimation method.

8. A power grid operation risk distribution monitoring system, applying the method according to any one of claims 1 to 7, characterized in that: include: A data acquisition and processing module is used to acquire sample data of several risk dimensions of the target power grid and perform dimensional preprocessing on the sample data; The risk dimensions include a high-risk dimension for a power grid, a medium-risk dimension for a power grid, a low-risk dimension for a power grid, and an acceptable risk dimension for a power grid; Sequence building module, used to build risk frequency sequence based on the result of dimension preprocessing; A single-dimensional monitoring module, configured to monitor each risk dimension according to the risk frequency sequence in combination with the first risk distribution monitoring strategy; The coupling module is used to perform risk coupling operations on the monitoring results of all risk dimensions; Indicator preset module, used to preset risk anomaly judgment indicators; The multi-dimensional detection module is used to monitor the risk distribution of power grid operations based on real-time data of the target power grid, combined with risk coupling operations and risk anomaly judgment indicators.

9. An electronic 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 a method for monitoring the distribution of risk of power grid operations according to any one of claims 1 to 7 are implemented.

10. 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 a method for monitoring the distribution of risk of power grid operations according to any one of claims 1 to 7 are implemented.

Images