Characterization method and system for net load uncertainty of power distribution network
By constructing a Dirichlet model and a Gaussian mixture model, combined with the Gibbs sampling method, the problem of characterizing net load uncertainty in power systems was solved, improving the accuracy and adaptability of net load forecasting for distribution networks, and enhancing the safety and economic benefits of power systems.
Patent Information
- Application Number
- CN202311802998.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-26
- Publication Date
- 2026-01-30
AI Technical Summary
With changes in the power system structure and the large-scale integration of renewable energy, the uncertainty of net load on the user side has increased, which has a significant impact on the operation and planning of the power system. Existing technologies are unable to accurately characterize the uncertainty of net load, affecting system security and economic efficiency.
A method for characterizing the uncertainty of net load in distribution networks is constructed by using the Dirichlet model and a data-driven Gaussian mixture model, combined with the Gibbs sampling method. The model is built by collecting power grid data, updating parameters, predicting the conditional probability distribution of error, and using five evaluation indicators to comprehensively evaluate the model performance.
It improves the adaptability and prediction accuracy of net load fluctuations in the distribution network, enhances the safety and economic efficiency of the power system, increases the utilization rate of renewable energy, and reduces the consumption of fossil fuels.
Smart Images

Figure CN121434824A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of user-side energy storage operation and control, in particular to a power distribution network net load uncertainty characterization method and system. BACKGROUND
[0002] With the development of renewable energy, the net load in the power system has attracted more and more attention. The net load of the power system refers to the power demand that needs to be met by traditional power sources (such as coal-fired, gas or nuclear power plants) after considering the output of renewable energy (such as wind and solar energy). The concept of net load has become increasingly important today as renewable energy plays an increasingly important role in the power system. Because the net load is influenced by both renewable energy output and total power demand, the uncertainty factor is large. If the uncertainty of renewable energy and user-side power demand is not fully considered in power system dispatching, it will seriously endanger the safe and stable operation of the new power system. In addition, with the increasing competition in the electricity market, power system operators are under pressure to improve the efficiency and economic benefits of the power system. Under this background, accurately characterizing the uncertainty of the net load of the power system can not only improve the safety and stability of the power system, but also improve the efficiency and economic benefits of the power system, thereby improving the competitiveness of the electricity market. By accurately characterizing the uncertainty of the net load, power system operators can more effectively dispatch renewable energy and traditional power sources, thereby improving the utilization rate of renewable energy and reducing the consumption of fossil energy, which helps to reduce carbon emissions of the power system.
[0003] Therefore, it is a challenging but urgent task to develop a data-driven based method to characterize the variation of the net load of the power system, thereby accurately characterizing the uncertainty of the net load. SUMMARY
[0004] In view of the above problems, the present application is proposed.
[0005] Therefore, the technical problem solved by the present application is that with the change of the structure of the power system and the large-scale access of renewable energy, the uncertainty of the user-side net load is becoming more and more obvious, which has an important influence on the operation and planning of the power system.
[0006] To solve the above technical problems, the present application provides the following technical solutions: a power distribution network net load uncertainty characterization method, comprising: collecting grid data and constructing a Dirichlet model.
[0007] Sampling from the conditional distribution and updating the model parameters.
[0008] Constructing a data-driven based Gaussian mixture model to predict the conditional probability distribution of the error.
[0009] The performance of the model is evaluated comprehensively by using five evaluation indexes.
[0010] As a preferred scheme of the power distribution network net load uncertainty characterization method, the Dirichlet model is expressed as:
[0011] G~DP (φ, Z (Ω))
[0012] w n ~Beta (w n |1, φ)
[0013]
[0014] The length of the selected stick each time is 1, and is expressed as:
[0015]
[0016] Wherein, φ represents the degree of discretization, Z represents the probability distribution function, Ω represents the random variable, λ n represents the weight coefficient, w n represents the length of each section, and is subject to Beta (w m |1, φ) distribution. w s represents the length of the broken stick taken s times, and n represents the number of times of folding the stick.
[0017] As a preferred scheme of the power distribution network net load uncertainty characterization method, the process of sampling and updating the model parameters from the conditional distribution includes: using the Gibbs sampling method to sample and update the model parameters from the conditional distribution, and the process is:
[0018] Data is D, and parameters are grouped as X1, X2,..., X N .
[0019] According to the Bayesian theorem, the conditional probability density function is derived after each group of parameters is verified:
[0020] p (X1|D, X2,..., X N ), p (X2|D, X1,..., X N ),..., p (X N |D, X1,..., X N )
[0021] Initialize parameters X1, X2,..., X N :
[0022]
[0023] Wherein, represents the initial value of the Nth parameter.
[0024] For n = 1,...,T, sampling is performed:
[0025]
[0026]
[0027] Wherein, T represents the Tth moment, Xn represents the N-1 samples drawn from the X1th group in the joint distribution at the nth moment.
[0028] Removing initial unstable state sample points, X1,X2,...,X N Sampling points, conforming to the probability density function.
[0029] The removing initial unstable state sample points comprises:
[0030] The initial sample data is cleaned to remove outliers and missing values.
[0031] Remove redundant sample points and keep representative sample points.
[0032] Re-fitting the distribution.
[0033] As a preferred scheme of the power distribution network net load uncertainty characterization method, wherein: the construction of the data-driven Gaussian mixture model comprises: using the actual value and the predicted value of the net load data to realize the modeling of the probability distribution, and then characterizing the net load uncertainty.
[0034] Let x be a 2-dimensional feature vector, and the Gaussian mixture model be composed of K Gaussian components, denoted as:
[0035]
[0036] Wherein, ω i The weight coefficient of the ith Gaussian component is represented as V, which satisfies:
[0037]
[0038] P i (x) represents the probability density function of the ith Gaussian component in the GMM, and the formula is as follows:
[0039]
[0040] Wherein,, μ i The mean of the ith Gaussian component is represented as S i The covariance of the ith Gaussian component is represented as Vi, and N represents the dimension of the feature vector.
[0041] As a preferred scheme of the power distribution network net load uncertainty characterization method, the performance of the comprehensive evaluation model is evaluated by using five evaluation indexes, including the performance of the design scheme under different levels of prediction error of the data set, and a radar chart is drawn to characterize the indexes, and the five evaluation indexes are the average percentage error of deviation, the determination coefficient, the mean square error, the error harmonic mean, and the average absolute error.
[0042] As a preferred scheme of the power distribution network net load uncertainty characterization method, the average percentage error of deviation is used to measure the deviation degree of the prediction model in the fitting scatter plot, and the average deviation of the predicted value relative to the true value is measured, and if the value of the MBPE is small, it indicates that the fitting degree of the prediction model for the scatter plot is good, and the error is small. If the value of the MBPE is large, it indicates that the fitting degree of the prediction model for the scatter plot is poor. The MBPE is represented as:
[0043]
[0044] wherein y i represents the true value, and represents the actual value of the i-th observation data point. represents the predicted value, represents the predicted value of the i-th observation data point, and n represents the total number of observation data points.
[0045] The determination coefficient is used to measure the variability of the model scatter plot. If the R-squared is closer to 1, the fitting degree of the model for the scatter plot is better, and is represented as:
[0046]
[0047] wherein SS res represents the sum of squares of residuals, and is the sum of squares of the difference between the predicted value of the model and the actual value. SS tot represents the total sum of squares, and is the sum of squares of the difference between the actual value and the average value.
[0048] As a preferred scheme of the power distribution network net load uncertainty characterization method, the mean square error is used to measure the error size of the prediction model in the fitting scatter plot, and the average value of the square of the difference between the predicted value and the actual value is calculated. If the value of the MSE is small, it indicates that the fitting degree of the prediction model for the scatter plot is good, and the error is small, and is represented as:
[0049]
[0050] wherein y i represents the true value, and is the actual value of the i-th observation data point. represents the predicted value, and is the predicted value of the i-th observation data point. N represents the total number of observation data points.
[0051] The error harmonic mean is used to measure the prediction model error, which is calculated based on the arithmetic mean of the inverse of the prediction error. If the HME value is small, it indicates that the model prediction error is small, and the fitting degree is good, which is represented as:
[0052]
[0053]
[0054] where e i represents the prediction error of the ith observation data point, which is the difference between the true value and the predicted value. n represents the total number of observation data points.
[0055] The mean absolute error is used to measure the prediction model error in fitting the scatter plot, which calculates the average value of the absolute value of the difference between the predicted value and the actual value. If the MAE value is small, it indicates that the prediction model has a good fitting degree for the scatter plot and the error is small, which is represented as:
[0056]
[0057] where y i represents the true value, which is the actual value of the ith observation data point. represents the predicted value, which is the predicted value of the ith observation data point. n represents the total number of observation data points.
[0058] A power distribution network net load uncertainty characterization system, characterized in that it comprises,
[0059] A model construction module acquires power grid data and constructs a Dirichlet model.
[0060] An update parameter module samples and updates the model parameters from the conditional distribution.
[0061] A prediction conditional probability distribution module constructs a data-driven Gaussian mixture model to predict the conditional probability distribution of the error.
[0062] An evaluation module uses five evaluation indexes to comprehensively evaluate the performance of the model.
[0063] A computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the steps of the method as described above when executing the computer program.
[0064] A computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the steps of the method as described above.
[0065] The beneficial effects of this invention are as follows: By constructing a Dirichlet model and a data-driven Gaussian mixture model, the uncertainty of the net load of the distribution network can be characterized more accurately. The introduction of the Dirichlet model enhances the model's ability to grasp the probability distribution of data, while the application of the Gaussian mixture model provides flexibility to adapt to different data characteristics and changes. The use of the Gibbs sampling method to update model parameters allows the model to dynamically adapt to data changes, improving its adaptability to fluctuations in the net load of the distribution network. The application of five evaluation indicators provides a multi-dimensional perspective for evaluating and improving the model. This ensures that the model's performance in different aspects is fully considered, thereby providing comprehensive and reliable prediction results. Attached Figure Description
[0066] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:
[0067] Figure 1 The following is an overall flowchart of a method for characterizing the uncertainty of net load in a distribution network, provided in the first embodiment of the present invention.
[0068] Figure 2 This is a data graph showing the predicted value and prediction error normalized for each data point in a method for characterizing the uncertainty of net load in a distribution network, provided in the second embodiment of the present invention.
[0069] Figure 3 The image shows a three-dimensional plot of the normalized weighted summation average probability density function of a method for characterizing the uncertainty of net load in a distribution network, provided in the second embodiment of the present invention.
[0070] Figure 4 The second embodiment of the present invention provides a prediction error expectation diagram under different prediction value conditions for a method for characterizing the uncertainty of net load in a distribution network.
[0071] Figure 5 The second embodiment of the present invention provides a method for characterizing the uncertainty of net load in a distribution network, and compares the performance of the present invention with that of the KNN model under five indicators. Detailed Implementation
[0072] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.
[0073] Example 1
[0074] Reference Figure 1 As an embodiment of the present invention, a method for characterizing the uncertainty of net load in a distribution network is provided, comprising:
[0075] S1: Construct the Dirichlet model.
[0076] It should be noted that the Dirichlet mixture model is a non-parametric Bayesian clustering method that adaptively determines the number of clusters. In traditional mixture models, the number of clusters needs to be pre-defined, but in practical applications, this number is usually unknown. The Dirichlet mixture model uses the Dirichlet process as prior knowledge, allowing the data to automatically determine the number of clusters.
[0077] A Dirichlet mixture model is a distribution of distributions, not a distribution of parameters. In other words, each drawing from a Dirichlet mixture model is itself a distribution, not a vector of parameters.
[0078] The Dirichlet model is represented as:
[0079] G~DP(φ,Z(Ω))
[0080] w n ~Beta(w n |1,φ)
[0081]
[0082] The sum of the lengths of the sticks selected each time is 1, which is represented as:
[0083]
[0084] Where φ represents the degree of discretization, Z represents the probability distribution function, Ω represents the random variable, and λ n w represents the weighting coefficient. n This represents the length of each segment, following the formula Beta(w). m |1,φ) distribution. w s Let represent the length of the broken rod taken in the s-th attempt, and n represent the number of times the rod is bent.
[0085] S2: Sample from the conditional distribution and update the model parameters.
[0086] It should be noted that in Dirichlet process mixture models (DPMMs), Gibbs sampling can be used to learn the parameters of clusters and the cluster assignments of data points. DPMM is a Bayesian nonparametric method used for clustering and density estimation. Its core idea is to assign data to an infinite number of latent clusters, whose weights are generated by a Dirichlet process. Through folded-rod construction, an approximate representation of the Dirichlet process can be generated, and the data can be assigned to these clusters. Gibbs sampling can then be used to learn the parameters of the clusters and the cluster assignments of data points.
[0087] The Gibbs sampling method is used to sample from the conditional distribution and update the model parameters. The process is as follows:
[0088] S2.1: Given data D, group the parameters into X1, X2, ..., X... N .
[0089] S2.2: After deriving the parameters of each group based on Bayes' theorem, test the conditional probability density function p(X1|D,X2,...,X). N ),p(X2|D,X1,...,X N ),...,p(X N |D,X1,...,X N ).
[0090] S2.3: Initialize parameters
[0091] S2.4: For n = 1, ..., T, sampling T represents the T-th time. For the nth time step, N-1 samples are drawn from the X1th group of the joint distribution.
[0092] S2.5: Remove the initial unstable state sample points to obtain X1, X2, ..., X N The sampling points conform to their joint posterior probability density function.
[0093] It should be noted that the specific process for removing initial unstable state sample points is as follows:
[0094] The initial sample data is cleaned to remove unstable sample points such as outliers and missing values.
[0095] Remove redundant sample points and retain representative sample points.
[0096] Refit the distribution.
[0097] Furthermore, in the Gibbs sampling of DPMM, data points are first randomly assigned to clusters. Then, for each data point, a new cluster assignment is sampled from the conditional probability distribution of its respective cluster. This conditional probability distribution is related to the cluster weights and the probability density of the data point within the cluster. Specifically, the conditional probability distribution is proportional to both the cluster size and the probability density of the data point within the cluster. Through multiple iterations of this process, the cluster assignments can be made to converge to the posterior distribution.
[0098] S3: Construct a data-driven Gaussian mixture model to predict the conditional probability distribution of the error.
[0099] By using the actual and predicted values of net load data, a probability distribution model is achieved, thereby characterizing the uncertainty of net load.
[0100] Let x be a 2-dimensional eigenvector. A Gaussian mixture model is composed of K Gaussian components, expressed as:
[0101]
[0102] Where, ω i Let V represent the weighting coefficient of the i-th Gaussian component, and let V represent the component, which conforms to:
[0103]
[0104] P i (x) represents the probability density function of the i-th Gaussian component in the GMM, as shown in the following formula:
[0105]
[0106] Where, μ i S represents the mean of the i-th Gaussian component. i Let Vi represent the covariance of the i-th Gaussian component, Vi represent the i-th component, and N represent the dimension of the eigenvector.
[0107] It should be noted that by using a Gaussian mixture model, the probability distribution of net load can be fitted based on historical data, thereby providing probability density estimates of net load under different conditions. These probability density estimates can serve as important input information for decision support and optimization algorithms, thereby improving the operating efficiency and reliability of energy systems. This invention utilizes two characteristics of net load data, actual values and predicted values, to model the probability distribution and thus characterize net load uncertainty.
[0108] S4: Five evaluation indicators are used to comprehensively evaluate the performance of the model.
[0109] To evaluate the performance of the design schemes under different levels of prediction error, radar charts were plotted to characterize the indicators. The five evaluation indicators are the mean deviation percentage error, coefficient of determination, mean square error, harmonic mean of error, and mean absolute error.
[0110] The Mean Percentage Error (MBPE) measures the degree of deviation of the prediction model from the actual scatter plot. It measures the average deviation of the predicted values from the true values. A small MBPE value indicates a good fit and low error in the prediction model. A large MBPE value indicates a poor fit. MBPE is expressed as:
[0111]
[0112] Among them, y i represents the true value, and represents the actual value of the i-th observation data point. Let represent the predicted value, and let represent the predicted value for the i-th observation data point. Let n represent the total number of observation data points.
[0113] The coefficient of determination (R-squared) measures the degree of variability in the model's scatter plot. The closer R-squared is to 1, the better the model fits the scatter plot, expressed as:
[0114]
[0115] Among them, SS res This represents the sum of squared residuals, which is the sum of the squared differences between the model's predicted and actual values. SS tot The total sum of squares represents the sum of the squares of the differences between the actual values and the average value.
[0116] Mean Squared Error (MSE) measures the magnitude of the error of a prediction model in fitting a scatter plot. It is calculated as the average of the squares of the differences between the predicted and actual values. A small MSE value indicates a good fit between the prediction model and the scatter plot, with a small error, expressed as:
[0117]
[0118] Among them, y i This represents the true value, which is the actual value of the i-th observation data point. Let represent the predicted value, which is the predicted value for the i-th observation data point. n represents the total number of observation data points.
[0119] The harmonic mean error (HME) measures the error of a prediction model and is calculated as the arithmetic mean of the reciprocals of the prediction errors. A small HME value indicates a small prediction error and a good fit, expressed as:
[0120]
[0121]
[0122] Among them, e i Let represent the prediction error for the i-th observation data point, which is the difference between the actual value and the predicted value. Let n represent the total number of observation data points.
[0123] Mean Absolute Error (MAE) measures the magnitude of the error of a prediction model in fitting a scatter plot. It is calculated as the average of the absolute values of the differences between the predicted and actual values. A smaller MAE value indicates a better fit of the prediction model to the scatter plot and a smaller error. This is expressed as:
[0124]
[0125] Among them, y i This represents the true value, which is the actual value of the i-th observation data point. Let represent the predicted value, which is the predicted value for the i-th observation data point. n represents the total number of observation data points.
[0126] The above embodiments also include a distribution network net load uncertainty characterization system, specifically:
[0127] The model module is constructed by collecting power grid data and building a Dirichlet model.
[0128] The parameter update module samples and updates model parameters from the conditional distribution.
[0129] The module for predicting conditional probability distribution constructs a data-driven Gaussian mixture model to predict the conditional probability distribution of the error.
[0130] The evaluation module uses five evaluation indicators to comprehensively evaluate the model's performance.
[0131] The computer device can be a server. This computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores data cluster data for a power monitoring system. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a method.
[0132] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0133] Example 2
[0134] Reference Figures 2 to 5 As an embodiment of the present invention, a method and system for characterizing the uncertainty of net load in a distribution network are provided. To verify the beneficial effects of the present invention, a comparative experiment is conducted for scientific demonstration.
[0135] In this example, a simulation experiment was conducted using the net load data of a certain park system in Zibo in 2021.
[0136] Figure 2This is a data plot after normalizing the predicted value and prediction error for each data point. The predicted value for each data point is normalized to obtain the x-axis of each data point in the net load scatter plot. Then, the predicted value for each data point is subtracted from the actual value to obtain the prediction error. The prediction error characteristics are then normalized to obtain the y-axis of each data point in the net load scatter plot.
[0137] Figure 3 This is a 3D plot of the normalized weighted average probability density function. To provide a more comprehensive and intuitive understanding of net load uncertainty, a 3D plot was generated based on the normalized weighted average probability density function of the prediction error under different predicted values to visualize the conditional probability density function. Each point in the plot represents the probability density of the prediction error being z when the predicted value is y. Through this plot, one can intuitively observe how the prediction error changes under different predicted values, thereby understanding the distribution characteristics of the prediction error.
[0138] Figure 4 The expected prediction error under different predicted values is used to further visualize the uncertainty of net load and to intuitively represent the prediction error. In addition, to facilitate the comparison of the fitting effect of different schemes on net load data, this paper solves the expected value of the normalized weighted summation average conditional probability density function of the prediction error under different predicted values, and draws two-dimensional curves.
[0139] Figure 5 This is a comparison chart of the performance indicators of the present invention and the KNN model under five indicators. Based on the prediction error curve of the KNN model and the expected prediction error curve under different prediction values, the scatter plot of net load data was compared in terms of fit. Then, five evaluation indicators, including HME and R... 2 The radar chart compares the performance of two design schemes using MAE, MBPE, and MSE. The performance metrics are normalized to ensure they are on the same order of magnitude for easy comparison. Each metric on the radar chart represents the performance level of its corresponding component. In other words, a higher value indicates better performance. As shown, although the two schemes perform similarly in MAE, this design scheme leads to varying degrees in the other four metrics. Of particular note is R... 2 In terms of the three metrics of HME and MBPE, this design not only leads but also shows a significant improvement. This result demonstrates that, considering the performance of all five evaluation metrics, this invention outperforms the KNN-based design overall. Especially in R... 2 In terms of the three evaluation metrics of HME and MBPE, the performance advantage of this design is very obvious.
[0140] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A method for representing uncertainty of a net load of a power distribution network, characterized by, The method comprises the following steps: Collecting power grid data, constructing a Dirichlet model; Sampling from the conditional distribution and updating the model parameters; Constructing a data-driven Gaussian mixture model to predict the conditional probability distribution of the error; Using five evaluation indexes to comprehensively evaluate the performance of the model.
2. The method of claim 1, wherein: The Dirichlet model is represented as: G ~ DP (φ, Z (Ω)) w n ~ Beta(w n |1,φ) The total length of the selected stick is 1, which is represented as: where φ denotes the degree of discretization, Z denotes a probability distribution function, Ω denotes a random variable, λ n denotes a weight coefficient, w n denotes the length of each segment, and is subject to a Beta(w m |1,φ) distribution; w s denotes the length of the broken rod taken s times, and n denotes the number of times of folding the rod.
3. The method of claim 2, wherein: The sampling from the conditional distribution and updating the model parameters comprises using the Gibbs sampling method to sample from the conditional distribution and update the model parameters, and the process is as follows: Data is D, parameters are grouped X1, X2,..., X N ; According to the Bayesian theorem, the conditional probability density function is derived after each group of parameters is verified: p(X1|D,X2,...,X N ),p(X2|D,X1,...,X N ),...,p(X N |D,X1,...,X N ) Initialization parameters X1, X2,..., X N : wherein, represents the initial value of the Nth parameter; For n = 1,...,T, sampling is performed: where T denotes the Tth time instant, denotes the N-1 samples drawn from the X1th group of the joint distribution at the nth time instant. Removing the initial unstable state sample points, the obtained X1, X2,..., X N Sampling points, conforming to the probability density function; The removal of initial unstable state sample points comprises the following steps: Cleaning the initial sample data to remove outliers and missing values; Removing redundant sample points and retaining representative sample points; Re-fitting the distribution.
4. The method of claim 3, wherein: The construction of the data-driven Gaussian mixture model comprises using the actual value and the predicted value of the net load data to realize the modeling of the probability distribution, and then representing the net load uncertainty; Let x be a 2-dimensional feature vector, and the Gaussian mixture model be composed of K Gaussian components, represented as: where ω i denotes the weighting factor of the i-th Gaussian component, V denotes the component, and complies with: P i (x) denotes the probability density function of the i-th Gaussian component in the GMM, and is given by where, μ i represents the mean of the i-th Gaussian component, S i represents the covariance of the i-th Gaussian component, V; represents the i-th component, and N represents the dimension of the eigenvector.
5. The method of claim 4, wherein: The five evaluation indexes are used to comprehensively evaluate the performance of the model, which comprises evaluating the performance of the design scheme under different levels of prediction error, drawing a radar chart to represent the indexes, and the five evaluation indexes are the average percentage error of deviation, the determination coefficient, the mean square error, the error harmonic mean and the average absolute error.
6. The method of power distribution grid net load uncertainty characterization of claim 5, wherein: The average percentage error of deviation is used to measure the deviation degree of the prediction model in the fitting scatter plot, measures the average deviation of the predicted value relative to the true value, and if the value of MBPE is small, it indicates that the fitting degree of the prediction model for the scatter plot is good and the error is small; if the value of MBPE is large, it indicates that the fitting degree of the prediction model for the scatter plot is poor, and MBPE is represented as: where y i represents the true value, represents the actual value of the i-th observation data point. represents the predicted value, represents the predicted value of the i-th observation data point. n represents the total number of observation data points; The determination coefficient is used to measure the variability of the model scatter plot; the closer R-squared is to 1, the better the fitting degree of the model for the scatter plot, and is represented as: where SS res represents the sum of squares of residuals, and is the sum of squares of the difference between the predicted value of the model and the actual value; SS tot represents the total sum of squares, and is the sum of squares of the difference between the actual value and the average value.
7. The method of claim 6, wherein: The mean square error is used to measure the error size of the prediction model in the fitting scatter plot, and the average value of the square of the difference between the predicted value and the actual value is calculated; if the value of MSE is small, it indicates that the fitting degree of the prediction model for the scatter plot is good and the error is small, and is represented as: where y i represents a true value, is the actual value of the i-th observation data point; represents a predicted value, is the predicted value of the i-th observation data point; n represents the total number of observation data points; The error harmonic mean is used to measure the prediction error of the model, which is calculated based on the arithmetic mean of the inverse of the prediction error; if the value of HME is small, it indicates that the prediction error of the model is small and the fitting degree is good, and is represented as: where e i represents the prediction error of the ith observation data point, which is the difference between the true value and the predicted value; n represents the total number of observation data points; The average absolute error is used to measure the error size of the prediction model in the fitting scatter plot, and the average value of the absolute value of the difference between the predicted value and the actual value is calculated; if the value of MAE is small, it indicates that the fitting degree of the prediction model for the scatter plot is good and the error is small, and is represented as: where y i represents a true value, is the actual value of the i-th observation data point; represents a predicted value, is the predicted value of the i-th observation data point; n represents the total number of observation data points.
8. A power distribution network net load uncertainty representation system using the method of any one of claims 1-7, characterized in that: The model construction module collects power grid data and constructs a Dirichlet model; The parameter updating module samples from the conditional distribution and updates the model parameters; A prediction conditional probability distribution module, which constructs a data-driven Gaussian mixture model to predict the conditional probability distribution of the error; An evaluation module, which uses five evaluation indexes to comprehensively evaluate the performance of the model. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor implements the steps of the method of any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method of any one of claims 1 to 7.
