Load modeling parameter identification method and related device

Abstract

The embodiment of the invention discloses a load modeling parameter identification method and a related device, and the method comprises the steps: obtaining a load data set of a power system, the load data set comprises load data samples of all time points, and the load data samples comprise current, voltage, active power and reactive power; a support vector machine is adopted to establish a load model, input data of the load model are voltage and current, and output data of the load model are active power and reactive power; and identifying model parameters of the load model based on a leave-one-out cross validation method and the load data set to obtain optimal load model parameters. Through the above mode, the method for optimizing the model parameters of the load model constructed by the support vector machine (SVM) based on the Leave-One-Out Cross Validation (LOCV) is adopted, so that the high-precision identification of the load model of the power system is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of power system load modeling, and in particular to a load modeling parameter identification method and related devices. Background Art

[0002] The safe and stable operation of the power system is inseparable from the accurate modeling of the system load. Due to the conservative nature of its construction, traditional load models are often unable to cope with the variable load characteristics of contemporary complex power systems. Existing load modeling methods include statistical synthesis and overall identification. Although the statistical synthesis method has the advantage of a clear physical model, it is difficult to obtain and determine parameters due to the randomness and complexity of the power load. The overall identification method does not rely on the internal structure of the load and is more suitable for the modeling of complex systems.

[0003] Support vector machine (SVM) is a machine learning algorithm widely used in nonlinear system modeling. It can handle complex nonlinear relationships, but its performance depends largely on the selection of model parameters. Traditional parameter identification methods, such as particle swarm optimization and genetic algorithm, may fall into local optimal solutions during the optimization process.

[0004] Therefore, there is still a lack of high-precision identification methods for power system load models, resulting in low efficiency in the prediction and control of power system operation. Summary of the invention

[0005] The main purpose of the present invention is to provide a load modeling parameter identification method and related devices, which can solve the problem in the prior art of lacking high-precision identification means of power system load models, resulting in low prediction and control efficiency of power system operation.

[0006] To achieve the above object, the present invention provides a load modeling parameter identification method in a first aspect, the method comprising:

[0007] Acquire a load data set of the power system, wherein the load data set includes load data samples at various time points, and the load data samples include current, voltage, active power, and reactive power;

[0008] A support vector machine is used to establish a load model, wherein the input data of the load model is voltage and current, and the output data of the load model is active power and reactive power;

[0009] Based on the leave-one-out cross-validation method and the load data set, the model parameters of the load model are identified to obtain the optimal load model parameters.

[0010] To achieve the above object, the second aspect of the present invention provides a load modeling parameter identification device based on a leave-one-out cross-validation method, the device comprising:

[0011] Data acquisition module: used to acquire a load data set of the power system, wherein the load data set includes load data samples at various time points, and the load data samples include current, voltage, active power and reactive power;

[0012] Model building module: used to establish a load model using a support vector machine, the input data of the load model is voltage and current, and the output data of the load model is active power and reactive power;

[0013] Parameter determination module: used to identify the model parameters of the load model based on the leave-one-out cross-validation method and the load data set to obtain the optimal load model parameters.

[0014] To achieve the above objective, the third aspect of the present invention provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps shown in the first aspect.

[0015] To achieve the above objectives, the fourth aspect of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps shown in the first aspect.

[0016] The embodiments of the present invention have the following beneficial effects:

[0017] The present invention provides a load modeling parameter identification method, the method comprising: obtaining a load data set of an electric power system, the load data set comprising load data samples at various time points, the load data samples comprising current, voltage, active power and reactive power; using a support vector machine to establish a load model, the input data of the load model being voltage and current, and the output data of the load model being active power and reactive power; identifying the model parameters of the load model based on a leave-one-out cross validation method and the load data set, and obtaining the optimal load model parameters. Through the above manner, a method for optimizing the model parameters of a load model constructed by a support vector machine (SVM) based on a leave-one-out cross validation method (LOOCV) is adopted to achieve high-precision identification of the load model of the electric power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] in:

[0020] Figure 1 A flow chart of a load modeling parameter identification method according to an embodiment of the present invention;

[0021] Figure 2 This is a schematic diagram of an SVM structure in an embodiment of the present invention;

[0022] Figure 3 A comparison diagram of an actual value and a simulated value of active power in an embodiment of the present invention;

[0023] Figure 4 A comparison diagram of the actual value and the simulated value of reactive power in an embodiment of the present invention;

[0024] Figure 5 This is a local comparison diagram of active power fitting in an embodiment of the present invention;

[0025] Figure 6 A reactive power fitting local comparison diagram in an embodiment of the present invention;

[0026] Figure 7 It is a structural block diagram of a load modeling parameter identification device in an embodiment of the present invention;

[0027] Figure 8 4 is a structural block diagram of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] Specifically, the present invention uses the Gaussian radial basis kernel function as the kernel function of the SVM model, uses the actual synchronized phasor measurement unit (PMU) data in the power system as input, and uses active power and reactive power as output. In the parameter identification process, the kernel parameters and regularization parameters of the model are optimized by the LOOCV method to ensure that the load model can achieve higher fitting accuracy while ensuring generalization capability. In the model identification process, the PMU measured current and voltage data are used as input variables, and active power and reactive power are used as output variables. The load modeling method based on measurement uses a load model parameter identification strategy optimized by the leave-one-out cross-validation method. The model uses SVM to establish a load model, avoiding the local minimum problem that plagues neural networks; the load model parameters optimized by the leave-one-out cross-validation method obtain a relatively considerable model fit. Through the LOOCV method optimization, the optimal comprehensive load model parameters are obtained, so that the model can better fit the power curve. Through the analysis of the PMU measured modeling parameter identification example, it is shown that the SVM load model based on the LOOCV method to optimize the parameters proposed in the present invention has better performance than the SVM load model with particle swarm algorithm optimization parameters and the SVM load model with bootstrap method optimization parameters. For details, please refer to the following content.

[0030] See also Figure 1 , Figure 1 Flow chart of a load modeling parameter identification method in an embodiment of the present invention. The method can be applied to both a terminal and a server. The terminal can be a desktop terminal or a mobile terminal. The mobile terminal can be at least one of a mobile phone, a tablet computer, a laptop computer, etc. The server can be implemented by an independent server or a server cluster composed of multiple servers. This embodiment is described by applying to a terminal as an example. Figure 1 The methods shown include:

[0031] 101. Obtain a load data set of a power system, wherein the load data set includes load data samples at various time points, and the load data samples include current, voltage, active power, and reactive power;

[0032] It should be noted that the measured wave data of the power system is recorded by a collection device such as a phasor measurement unit (PMU), and the measured data such as the voltage, current, active power, and reactive power of the load nodes at each moment are obtained as input features of the support vector machine (SVM) model, that is, step 101 may include: obtaining the measured wave data of each load node recorded by the phasor measurement unit of the power system, the measured wave data including the voltage, current, active power, and reactive power of each load node; based on the measured wave data, a load data set is obtained. Among them, the collected raw data can also be preprocessed to improve the data quality, including but not limited to normalization and denoising, etc.

[0033] 102. Use a support vector machine to establish a load model, wherein the input data of the load model is voltage and current, and the output data of the load model is active power and reactive power;

[0034] It should be noted that the load model is a prediction model that uses current and voltage to predict active power and reactive power, and a regression prediction model that has learned the fitting relationship between historical current and voltage and active power and reactive power. It is a regression prediction model that can use historical data to predict the data at the next moment.

[0035] In one possible implementation, see Figure 2 , Figure 2 FIG. 1 is a schematic diagram of an SVM structure in an embodiment of the present invention. The present invention adopts a support vector machine to establish a load model and selects a Gaussian radial basis kernel function to solve nonlinear and high-dimensional problems. For power system load modeling, we construct an optimal decision function, such as formula (1):

[0036]

[0037] in, represents the characteristics of the input data, which can be current and voltage in the present invention, w and b represent the weight vector and the deviation value respectively, and w and b are found by using the structural risk minimization principle to construct the support vector machine optimization risk function, as shown in formula (2):

[0038]

[0039] ξ represents the introduced slack variable and is penalized by the regularization parameter γ in the objective function. The Lagrange multipliers α and β are introduced to construct the Lagrange function, as shown in formula (3):

[0040]

[0041] In order to minimize the objective function, find the saddle point of the Lagrange function in the above formula and set the partial derivatives of each variable in L to zero, as shown in formula (4):

[0042]

[0043] In order to facilitate calculation, the kernel function is defined Substituting the above extreme value conditions into the equation, we get the following dual problem, as shown in equation (5):

[0044]

[0045] Select α j >0, calculate b, as shown in formula (6):

[0046]

[0047] From formula (4), we can get Substituting into formula (1), the final expression of the load model is as follows:

[0048]

[0049] Among them, the kernel function (KF) is the key factor in determining the feature space. Different kernel functions in SVM form different algorithms and constitute different SVM models. When using the SVM model, there is no need to pay attention to the complex feature space, but the most critical kernel function selection can determine the accuracy and fitness of load modeling. Therefore, how to choose a suitable kernel function has always been a key step in SVM modeling. Currently, the commonly used kernel functions mainly include linear kernel function, polynomial kernel function, Sigmoid function, Gaussian radial basis function, exponential radial basis kernel function, Fourier kernel function, etc. By comparing a large number of literature and experimental materials, the Gaussian radial basis kernel function has better performance.

[0050] The kernel function in the present invention selects Gaussian radial basis function:

[0051]

[0052] The Gaussian radial basis function is extended to the load model, where the methods of modeling the power system load are mainly divided into statistical synthesis method and overall identification method. Because the support vector machine modeling method regards the load as a whole "black box" and uses input-output to determine the load model, this idea is similar to the overall identification method. Therefore, the support vector machine load modeling can be regarded as a kind of overall identification method. The active power and reactive power of the SVM load model structure can be expressed as formula (8):

[0053]

[0054] The load model includes the SVM active power model P and the SVM reactive power model Q, where N P 、N Q are the number of support vectors for active power and reactive power models respectively; σ P , σ Q are the kernel widths of the SVM active power and reactive power models respectively; x Pi =[U i ,I i ] T 、x Qi =[U i ,I i ] T , i = 1, 2, ..., l, are the input data of the SVM active power model and the SVM reactive power model, i.e., the training samples when the active power and reactive power models are trained. The training samples include current and voltage. The present invention adopts leave-one-out cross validation. During each training (regression fitting), the training samples are obtained by leaving one sample in the load data set as a validation set and the remaining samples as training samples. x is the input value of the function, i.e., x P =[U,I] T 、x Q =[U,I] T ; α pi is the active power network weight, α Qi is the reactive power network weight; b P , b Q is the deviation value of the SVM active power and reactive power model. The load model expression of the support vector machine has nothing to do with the load composition, but identifies the model according to the actual input and output training data to reduce the model inaccuracy caused by human qualitative analysis. The load model established by the present invention only needs to identify α i With the kernel parameter σ, given that 0≤α in formula (5) i ≤γ, identify α i can be converted into the regularization parameter γ, and b can be calculated by equation (6).

[0055] 103. Based on a leave-one-out cross-validation method and a load data set, model parameters of the load model are identified to obtain optimal load model parameters.

[0056] Furthermore, the leave-one-out cross-validation method is used to optimize the kernel width and regularization parameter of SVM. The regularization parameter is used to balance the empirical risk and model complexity, and control the degree of penalty for samples exceeding the precision (insensitive loss parameter), so that the required f(x) has better generalization ability. The regularization parameter and the insensitive loss function are pre-set parameters. In the present invention, an initial value is given, and then the leave-one-out cross-validation method is used to find its optimal value. Among them, the optimal load model parameter refers to the optimal value of the regularization parameter γ and the kernel width σ, and its deviation term does not need to be identified, and can be calculated by the following formula: By removing one sample and using the remaining samples for training, the optimal parameter combination is found to maximize the model's fit and predictive ability.

[0057] It should be noted that one of the most useful methods in selection problems is the cross-validation (CV) method. In the past decade, CV methods have been quite extensively developed in the literature, especially in the field of nonparametric curve estimation.

[0058] After load modeling is performed using SVM, the present invention uses the “leave one fit goodness of fit” method to evaluate the out-of-sample fitting ability of the regression model. Let Y = (y1,…,y n ) T is the response vector, i.e. the output data of the load model, X = (x ij ) is the data matrix of the input data of the load model, i=1, ..., n, j=1, ..., K, K refers to the number of input features of the model, that is, the dimension or feature number used by the model to predict the output. In the present invention, K=2, that is, voltage and current, is the load model design matrix defined as formula (9):

[0059] Y=Xβ+ε (9)

[0060] In the formula, ε=(ε1,…,ε n ) T is a random disturbance vector; assuming there are k0 covariates in β, This corresponds to the case where the K covariates are pre-ordered according to their importance, so that only the number of covariates needs to be determined. is the fitted value, y i is the response variable; residual When β has a constant term in the established load model, the sum of squared deviations can be expressed as follows:

[0061]

[0062] in, is the average value of the response data. When the model has a constant term, the explanatory variables of the model contain a column that is all constants, indicating that the model has an intercept term. For the regression model, this constant term can help the model better fit the average trend of the data. It can be seen that the sum of squares of the deviations of the explained variables can be decomposed into the part explained by the model and the residual part that cannot be explained by the model. Therefore, the goodness of fit R is defined 2 The expression (11) is:

[0063]

[0064] The LOOR explored by the present invention 2 Based on the leave-one-out regression, the leave-one-out regression estimates the regression model parameters by removing one sample from the whole sample. The introduction of the leave-one-out cross-validation method, which estimates the regression model parameters by removing one sample from the whole sample, is to solve this problem in the load parameter identification process. Let X (-i) is the data matrix with row i removed from matrix X, y (-i) To eliminate the explained variable vector of the i-th sample, the variable coefficient estimate obtained by the leave-one-out cross-validation method is formula (12):

[0065]

[0066] Where, X (-i) From the matrix X = (x ij ) in the matrix with the i-th row of data removed, y (-i) is the response vector that eliminates the i-th sample, X′ (-i) Yes X (-i) The transposed matrix of (-i) X (-i) is the remaining data matrix X (-i) The covariance matrix of (-i) X (-i) ) -1 X′ (-i) It is the pseudo-inverse formula for calculating parameter estimates; X(-i) means deleting the data in the i-th row. The matrix of X(-i) will become n-1 rows and K columns. X′ (-i) It is the data matrix with row i removed from matrix X and then transposed. (-i) X (-i) is the remaining data matrix X (-i) The covariance matrix of (-i) X (-i) is a symmetric matrix that can reflect the correlation between variables. (-i) X (-i) ) -1 X′ (-i)It is the pseudo-inverse formula used to calculate the parameter estimate. (12) gives the variable coefficient estimate in the leave-one-out cross-validation method. This is the standard formula for the least squares method. After deleting the i-th row, the remaining data is used to fit the model. The goal is to minimize the residual sum of squares. The least squares method obtains parameter estimates through the pseudo-inverse calculation of the matrix.

[0067] When the leave-one-out cross-validation method is used to estimate After that, the i-th sample in the original sample will be deleted, and the updated fitting value can be used express, The expression of is as follows:

[0068]

[0069] For a given sample value, the sample value here is the sample value obtained from the pre-collected load data set. The above operation will be repeated, and different samples will be removed each time to repeat the above operation, and it will be repeated n times, that is, each sample is used as a validation set to validate the load model, so as to obtain After the above operation, the range of goodness of fit LOOR2 can be further obtained, as shown in formula (14):

[0070]

[0071] in, for y i and The correlation coefficient of It is the fit of the regression model outside the sample, so it represents the true prediction value of the model. The leave-one-out goodness of fit established by the true prediction value can better reflect the true fitting ability of the regression model.

[0072] It should be noted that in leave-one-out cross-validation, formula (12) participates in the selection of the optimal model parameters through the following process: each time a sample i is deleted, the remaining n-1 samples are used to fit the linear regression model, and the model parameters after the sample is removed are calculated The predicted value of the i-th sample is predicted using the fitted model. Repeat the above steps n times to cover all samples. Finally, the leave-one-out cross-validation goodness of fit LOOR is calculated using formula (11): 2 , evaluate the fitting ability of the model, where the leave-one-out goodness of fit refers to the goodness of fit of the predicted value of the i-th sample removed each time, which can be obtained using the true value and predicted value of the i-th sample, where the true value is the measured active and reactive power, and the predicted value is the active and reactive power predicted by the load model. 2 The present invention uses an evaluation index, and the LOOR explored by the present invention is 2Based on the leave-one-out regression, the leave-one-out regression estimates the regression model parameters by removing one sample from the entire sample. The closer the value is to 1, the better the effect.

[0073] Through the above steps, the load model parameter identification and optimization are completed. Figures 3 to 6 , Figure 3 A comparison diagram of an actual value and a simulated value of active power in an embodiment of the present invention; Figure 4 A comparison diagram of the actual value and the simulated value of reactive power in an embodiment of the present invention; Figure 5 This is a local comparison diagram of active power fitting in an embodiment of the present invention; Figure 6 This is a local comparison diagram of reactive power fitting in the embodiment of the present invention, and then compares the SVM load model (PSO_SVM) with particle swarm optimization parameters and the SVM load model (B_SVM) with bootstrap optimization parameters, showing that the method proposed in this scheme can improve the speed of identification and the accuracy of identification. Among them, the simulation value is the estimated value predicted by the load model, and the actual value is the actual collected recorded data.

[0074] The present invention is applicable to load modeling and parameter identification of power systems, and is particularly applicable to scenarios where complex nonlinear loads need to be processed. The method can effectively improve the accuracy of load modeling of power systems, thereby improving the safety and economy of power systems.

[0075] The present invention provides a load modeling parameter identification method, the method comprising: obtaining a load data set of an electric power system, the load data set comprising load data samples at various time points, the load data samples comprising current, voltage, active power and reactive power; using a support vector machine to establish a load model, the input data of the load model being voltage and current, and the output data of the load model being active power and reactive power; identifying the model parameters of the load model based on a leave-one-out cross validation method and the load data set, and obtaining the optimal load model parameters. Through the above manner, a method for optimizing the model parameters of a load model constructed by a support vector machine (SVM) based on a leave-one-out cross validation method (LOOCV) is adopted to achieve high-precision identification of the load model of the electric power system.

[0076] See also Figure 7 , Figure 7 FIG. 1 is a structural block diagram of a load modeling parameter identification device based on a leave-one-out cross-validation method in an embodiment of the present invention. Figure 7 The device shown comprises:

[0077] Data acquisition module 701: used to acquire a load data set of a power system, wherein the load data set includes load data samples at various time points, and the load data samples include current, voltage, active power and reactive power;

[0078] Model building module 702: used to establish a load model using a support vector machine, the input data of the load model is voltage and current, and the output data of the load model is active power and reactive power;

[0079] Parameter determination module 703: used to identify the model parameters of the load model based on the leave-one-out cross-validation method and the load data set to obtain the optimal load model parameters.

[0080] It should be noted that Figure 7 The functions of each module in the device shown are Figure 1 The contents of each step in the method shown are similar, so they are not described here to avoid repetition. For details, please refer to Figure 1 The content of each step in the method shown.

[0081] The present invention provides a load modeling parameter identification device, the device includes: a data acquisition module: used to obtain a load data set of an electric power system, the load data set includes load data samples at various time points, and the load data samples include current, voltage, active power and reactive power; a model construction module: used to establish a load model using a support vector machine, the input data of the load model is voltage and current, and the output data of the load model is active power and reactive power; a parameter determination module: used to identify the model parameters of the load model based on a leave-one-out cross validation method and a load data set to obtain optimal load model parameters. Through the above device, a method for optimizing the model parameters of a load model constructed by a support vector machine (SVM) based on a leave-one-out cross validation method (LOOCV) is adopted to achieve high-precision identification of the load model of the electric power system.

[0082] Figure 8 FIG. 1 shows an internal structure diagram of a computer device in an embodiment. The computer device may be a terminal or a server. Figure 8 As shown, the computer device includes a processor, a memory and a network interface connected via a system bus. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor can implement the above method. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor can implement the above method. Those skilled in the art can understand that Figure 8The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0083] In one embodiment, a computer device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the following steps: Figure 1 Steps shown.

[0084] In one embodiment, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the processor performs the following steps: Figure 1 Steps shown.

[0085] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0086] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0087] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for identifying load modeling parameters, characterized in that: The method comprises: Acquire a load data set of the power system, wherein the load data set includes load data samples at various time points, and the load data samples include current, voltage, active power, and reactive power; A support vector machine is used to establish a load model, wherein the input data of the load model are voltage and current, and the output data of the load model are active power and reactive power; Based on the leave-one-out cross-validation method and the load data set, the model parameters of the load model are identified to obtain the optimal load model parameters.

2. The method according to claim 1, characterized in that: The use of a support vector machine to establish a load model includes: Model the power system load and construct the optimal decision function: in, Represents the characteristics of the input data, w and b represent the weight vector and the deviation value respectively. The structural risk minimization principle is used to find w and b, and the support vector machine optimization risk function is constructed: Among them, ξ represents the introduced slack variable, and is penalized by the regularization parameter γ in the objective function; Lagrange multipliers α and β are introduced to construct the Lagrange function: In order to minimize the objective function, find the saddle point of the Lagrange function in the above formula and set the partial derivatives of each variable in L to zero: In order to facilitate calculation, the kernel function is defined Substituting the above extreme value conditions into the equation, we get the following dual problem: Select α j >0, calculate b: Will Substituting the formula of the optimal decision function into the load model, we can get:

3. The method according to claim 2, characterized in that: The kernel function of the support vector machine selects the Gaussian radial basis function, and the expression of the Gaussian radial basis function is as follows: Here, σ represents the kernel width.

4. The method according to claim 3, characterized in that: The load model includes the expressions of the SVM active power model P and the SVM reactive power model Q as follows: Where N P 、N Q are the number of support vectors for the SVM active power and reactive power models respectively; σ P , σ Q are the kernel widths of the SVM active power and reactive power models respectively; x Pi =[U i ,I i ] T 、x Qi =[U i ,I i ] T , i = 1, 2, ..., l, are the input data of the SVM active power model and the SVM reactive power model respectively, x is the input value, including x P =[U,I] T 、x Q =[U,I] T ; α pi is the active power network weight, α Qi is the reactive power network weight; b P , b Q It is the deviation value of SVM active power and reactive power model.

5. The method according to claim 4, characterized in that: The method of identifying the model parameters of the load model based on the leave-one-out cross-validation method and the load data set to obtain the optimal load model parameters includes: Let Y = (y1,…,y n ) T is the response vector, i.e., the output data of the load model, X = (x ij ) is the data matrix of the input data of the load model, i=1,…,n, j=1,…,K; The design matrix for the load model is defined as: Y = Xβ + ε; Where ε=(ε1,…,ε n ) T is a random disturbance vector; β is the load model parameter. Assuming there are k0 covariates in β, The leave-one-out cross-validation method is introduced to estimate the regression model parameters using the remaining samples after removing one sample from the load data set. The estimated value of the load model parameters obtained by the leave-one-out cross-validation method is The expression is as follows: Where, X (-i) From the matrix X = (x ij ) in the matrix with the i-th row of data removed, y (-i) is the response vector to remove the i-th sample, X ( ' -i) Yes X (-i) The transposed matrix of ( ' -i) X (-i) is the remaining data matrix X (-i) The covariance matrix of ( ' -i) X (-i) ) -1 X ( ' -i) is the pseudo-inverse formula for calculating parameter estimates; When the leave-one-out cross-validation method is used to estimate After that, the updated fitted value of the deleted i-th sample is replaced by express, The expression is as follows: For a given load data set, the above operation is repeated by removing different samples each time, thereby obtaining And the final estimated model parameters The optimal load model parameters include the model parameters finally estimated 6. The method according to claim 5, characterized in that: The method further comprises: When the load model has a constant term in β, the response vector y i The expression of the sum of squared deviations is as follows: in, is the response vector y i The average value of is the fitted value, is the residual; The goodness of fit of the load model is obtained by using the sum of squared deviations. 2 The expression is as follows: In the formula, R 2 is the goodness of fit, is the residual, the fitted value The response vector y i .

7. The method according to claim 1, characterized in that: The step of obtaining a load data set of the power system includes: Acquire the measured wave data of each load node recorded by the phase measurement unit of the power system, wherein the measured wave data includes the voltage, current, active power and reactive power of each load node; Based on the measured recorded data, a load data set is obtained.

8. A load modeling parameter identification device, characterized in that: The device comprises: Data acquisition module: used to acquire a load data set of the power system, wherein the load data set includes load data samples at various time points, and the load data samples include current, voltage, active power and reactive power; Model building module: used to establish a load model using a support vector machine, the input data of the load model is voltage and current, and the output data of the load model is active power and reactive power; Parameter determination module: used to identify the model parameters of the load model based on the leave-one-out cross-validation method and the load data set to obtain the optimal load model parameters.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 7.

10. A computer device comprising a memory and a processor, characterized in that: The memory stores a computer program, and when the computer program is executed by the processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 7.

Images