Power distribution network power flow regression method and device based on SVR data driving

Abstract

The invention provides a power distribution network power flow regression method and device based on SVR data driving, and relates to the technical field of power distribution network analysis, and the method comprises the steps: obtaining target input information containing the state information of each node of a power distribution network; the nodes of the power distribution network comprise at least one of PV nodes and PQ nodes; the state information of the PV node comprises active power and voltage amplitude; the state information of the PQ node comprises active power and reactive power; inputting the target input information into the power flow regression model to obtain a target output result; the model is constructed based on a support vector regression (SVR) model; the objective function of the model is obtained based on a second mapping relation between the known quantity and the unknown quantity. According to the power distribution network power flow regression method and device based on SVR data driving provided by the invention, the high-accuracy power flow solution is calculated through the power distribution network power flow regression model based on SVR data driving, and then reasonable allocation of the power distribution network is realized.

Description

Technical Field

[0001] The present application relates to the technical field of distribution network analysis, and in particular to a distribution network power flow regression method and device driven by SVR data. Background Art

[0002] Power flow calculation is the most basic and important part of power system analysis. It is also the basis of system safety, economic operation analysis, and real-time control and dispatch. Therefore, power flow calculation is of great significance to power system planning and design.

[0003] In recent years, a high proportion of renewable energy has been connected to the grid on a large scale, and the penetration rate of equipment such as energy storage and electric vehicles in the distribution network has gradually increased, leading to the diversification of power generation forms; source and load fluctuations have strong randomness and uncertainty, and problems such as uneven distribution of distribution network power flow and voltage over-limit have frequently occurred.

[0004] Based on this, there is an urgent need for a power flow calculation method for active distribution networks to obtain a high-accuracy power flow solution and thus achieve reasonable deployment of the distribution network. Summary of the invention

[0005] The purpose of this application is to provide a distribution network power flow regression method and device based on SVR data drive, which is used to calculate a high-accuracy power flow solution through a distribution network power flow regression model driven by SVR data, thereby achieving reasonable deployment of the distribution network.

[0006] The present application provides a distribution network power flow regression method based on SVR data drive, including: Obtain target input information including status information of each node of the distribution network; the nodes of the distribution network include at least one of the following: PV nodes and PQ nodes; the status information of the PV nodes includes: active power and voltage amplitude; the status information of the PQ nodes includes: active power and reactive power; input the target input information into the power flow regression model to obtain a target output result; wherein the target output result includes at least one of the following: reactive power and voltage phase angle of the PV node, voltage amplitude and voltage phase angle of the PQ node; the power flow regression model is constructed based on the support vector regression SVR model; the objective function of the power flow regression model is: obtained based on the second mapping relationship between known quantities and unknown quantities; the second mapping relationship is determined based on the power flow equations corresponding to different bus types of the distribution network; the kernel function is used in the power flow regression model to replace the characteristic mapping inner product function of the SVR model; the kernel function is used to reduce the calculation dimension of the power flow regression model.

[0007] Optionally, the power flow regression model is constructed based on the following steps: establishing a first mapping relationship between power and voltage, and determining known quantities and unknown quantities corresponding to different bus types based on the bus type of the distribution network node; the power includes: active power and reactive power; the voltage includes: voltage amplitude and voltage phase angle; constructing a first mapping relationship equation based on the first mapping relationship, and converting the first mapping relationship equation into a second mapping relationship equation based on the known quantities and unknown quantities corresponding to different bus types; solving the regression parameter matrix and constant terms in the second mapping relationship equation, and performing power flow mapping based on the solved second mapping relationship equation; the power flow mapping is used to establish a second mapping relationship between known quantities and unknown quantities; establishing the objective function based on the second mapping relationship, and constructing the kernel function based on the regression parameter matrix, and constructing the power flow regression model based on the objective function and the kernel function.

[0008] Optionally, the first mapping relationship equation is expressed by the following formula:

[0009] in, H, N, M, L are all regression parameter matrices; and is a constant term; The second mapping relationship equation is expressed by the following formula:

[0010] Among them, the subscript L express PQ Busbar, subscript S express PV Busbar; is the regression parameter matrix; is a constant term; V is the voltage amplitude, is the voltage phase angle, P is the active power, Q is the reactive power.

[0011] Optionally, the second mapping relationship is expressed by the following formula:

[0012] in, and is a known quantity, and is the unknown quantity; K is calculated based on the regression parameter matrix; is a constant term.

[0013] Optionally, the objective function is expressed by the following formula:

[0014] in, W represents the regression parameter matrix, represents the dependent variable matrix, represents the independent variable matrix; represents the Frobenius norm; C is the penalty coefficient; and is the slack variable; is the tolerance deviation; the penalty coefficient is used to constrain the tolerance deviation.

[0015] Optionally, the kernel function is expressed by the following formula:

[0016] in, and is the Lagrange multiplier, represents the independent variable matrix, b is a constant term.

[0017] Optionally, the power flow regression model uses the following loss function to train the model:

[0018] in, is the dependent variable matrix, is the independent variable matrix, Tolerance for deviations; is the output of the model.

[0019] The present application also provides a distribution network power flow regression device driven by SVR data, comprising: A data acquisition module is used to acquire target input information including status information of each node of a distribution network; the nodes of the distribution network include at least one of the following: a PV node and a PQ node; the status information of the PV node includes: active power and voltage amplitude; the status information of the PQ node includes: active power and reactive power; a power flow calculation module is used to input the target input information into a power flow regression model to obtain a target output result; wherein the target output result includes at least one of the following: reactive power and voltage phase angle of the PV node, voltage amplitude and voltage phase angle of the PQ node; the power flow regression model is constructed based on a support vector regression SVR model; the objective function of the power flow regression model is: obtained based on a second mapping relationship between known quantities and unknown quantities; the second mapping relationship is determined based on power flow equations corresponding to different bus types of the distribution network; a kernel function is used in the power flow regression model to replace the characteristic mapping inner product function of the SVR model; the kernel function is used to reduce the calculation dimension of the power flow regression model.

[0020] Optionally, the device also includes: a model building module; the model building module is used to establish a first mapping relationship between power and voltage, and determine the known quantities and unknown quantities corresponding to different bus types based on the bus type of the distribution network node; the power includes: active power and reactive power; the voltage includes: voltage amplitude and voltage phase angle; the model building module is also used to construct a first mapping relationship equation based on the first mapping relationship, and convert the first mapping relationship equation into a second mapping relationship equation based on the known quantities and unknown quantities corresponding to different bus types; the model building module is also used to solve the regression parameter matrix and constant terms in the second mapping relationship equation, and perform flow mapping based on the solved second mapping relationship equation; the flow mapping is used to establish a second mapping relationship between known quantities and unknown quantities; the model building module is also used to establish the objective function based on the second mapping relationship, and construct the kernel function based on the regression parameter matrix, and construct the flow regression model based on the objective function and the kernel function.

[0021] Optionally, the first mapping relationship equation is expressed by the following formula:

[0022] in, H, N, M, L are all regression parameter matrices; and is a constant term; The second mapping relationship equation is expressed by the following formula:

[0023] Among them, the subscript L express PQ Busbar, subscript S express PV Busbar; is the regression parameter matrix; is a constant term; V is the voltage amplitude, is the voltage phase angle, P is the active power, Q is the reactive power.

[0024] Optionally, the second mapping relationship is expressed by the following formula:

[0025] in, and is a known quantity, and is the unknown quantity; K is calculated based on the regression parameter matrix; is a constant term.

[0026] Optionally, the objective function is expressed by the following formula:

[0027] in, W represents the regression parameter matrix, represents the dependent variable matrix, represents the independent variable matrix; represents the Frobenius norm; C is the penalty coefficient; and is the slack variable; is the tolerance deviation; the penalty coefficient is used to constrain the tolerance deviation.

[0028] Optionally, the kernel function is expressed by the following formula:

[0029] in, and is the Lagrange multiplier, represents the independent variable matrix, b is a constant term.

[0030] Optionally, the power flow regression model uses the following loss function to train the model:

[0031] in, is the dependent variable matrix, is the independent variable matrix, Tolerance for deviations; is the output of the model.

[0032] The present application also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of any of the above-mentioned distribution network power flow regression methods driven by SVR data.

[0033] The present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any of the above-described distribution network power flow regression methods driven by SVR data are implemented.

[0034] The present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of any of the above-mentioned distribution network power flow regression methods driven by SVR data are implemented.

[0035] The distribution network power flow regression method and device based on SVR data driving provided in the present application first obtain target input information including status information of each node of the distribution network; the nodes of the distribution network include at least one of the following: PV nodes and PQ nodes; the status information of the PV nodes includes: active power and voltage amplitude; the status information of the PQ nodes includes: active power and reactive power; then, the target input information is input into the power flow regression model to obtain the target output result; wherein, the target output result includes at least one of the following: reactive power and voltage phase angle of the PV node, voltage amplitude and voltage phase angle of the PQ node; the power flow regression model is constructed based on the support vector regression SVR model; the objective function of the power flow regression model is: obtained based on the second mapping relationship between known quantities and unknown quantities; the second mapping relationship is determined based on the power flow equation corresponding to different bus types of the distribution network; the characteristic mapping inner product function of the SVR model is replaced by a kernel function in the power flow regression model; the kernel function is used to reduce the calculation dimension of the power flow regression model. In this way, a high-accuracy power flow solution can be calculated through the distribution network power flow regression model driven by SVR data, thereby achieving reasonable deployment of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the present application or 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 some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0037] Figure 1 It is one of the flow diagrams of the distribution network power flow regression method based on SVR data drive provided in this application; Figure 2 This is the second flow chart of the distribution network power flow regression method based on SVR data drive provided by this application; Figure 3 It is a schematic diagram of the improved IEEE33 node calculation example provided by this application; Figure 4 This is a comparison chart of the accuracy of different SVR kernel functions provided by this application; Figure 5 It is the PQ node voltage regression accuracy diagram of the four methods provided in this application; Figure 6 It is the PQ node phase angle regression accuracy diagram of the four methods provided in this application; Figure 7 It is the PQ node voltage regression accuracy diagram of the four methods after bus type transformation provided by this application; Figure 8 It is the PQ node phase angle regression accuracy diagram of the four methods after bus type transformation provided by this application; Fig. 9 It is a structural schematic diagram of a distribution network power flow regression device driven by SVR data provided in this application; Fig.10 It is a structural schematic diagram of the electronic device provided by this application. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0039] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0040] The power flow equation is established based on the knowledge of system topology and line parameters. Since the electrical parameters of medium and low voltage distribution networks are often not collected and the measurement system is not as complete as the transmission network, the topological changes caused by the switching action are difficult to be reflected in the monitoring system in real time. Therefore, it is difficult to apply the power flow method of the transmission network based on the admittance matrix to the medium and low voltage distribution network in practice.

[0041] The power industry has a good accumulation of data measurement, transmission and storage. The data-driven method can monitor the operating parameters and various operating modes of the distribution network, which lays a good foundation for data-driven analysis and enhances the accuracy, real-time performance and operational stability of distribution network flow data acquisition.

[0042] In power flow calculations, changes in bus types are not uncommon. When a distributed power source (DG) is connected, most bus nodes can be regarded as PV nodes. For example, distributed photovoltaics can be equivalent to PV nodes when using voltage inverters; fuel cells are usually equivalent to PV nodes; micro gas turbines can be equivalent to PV nodes when the control mode is voltage control; some PV buses must be converted into PQ buses due to the limitation of reactive output. However, the conversion of the bus will change the fixed known and unknown variables, and the previous power flow model is no longer applicable. In the existing data-driven power flow calculations, there is no good solution to this problem. Based on this, the embodiment of the present application provides a data-driven power flow regression method that can cope with bus type changes.

[0043] The distribution network power flow regression method based on SVR data driving provided in the embodiment of the present application includes the following steps: First, the data-driven power flow equations under different bus types are established, including: establishing data-driven power flow equations based on traditional physical power flow equations; establishing power flow equations under different bus types. After that, the update method of the bus type transformation of the model is derived. Finally, a power flow regression model based on SVR is constructed, including: establishing the objective function, solving the parameters, and embedding the kernel function.

[0044] The power flow regression model based on support vector regression (SVR) constructed in the embodiment of the present application better fits the nonlinearity of the power flow by embedding a Gaussian kernel function. The model is used to perform simulation on an improved IEEE33-node system. In the scenarios where the distribution network is unobservable and has outliers, it is verified that the proposed method can effectively obtain a high-accuracy power flow solution before and after the bus type is transformed.

[0045] The following, in conjunction with the accompanying drawings, describes in detail the SVR data-driven distribution network power flow regression method provided in the embodiment of the present application through specific embodiments and their application scenarios.

[0046] like Figure 1 As shown, an embodiment of the present application provides a distribution network power flow regression method based on SVR data driving, and the method may include the following steps 101 and 102: Step 101: Obtain target input information including status information of each node in the distribution network.

[0047] The nodes of the distribution network include at least one of the following: a PV node and a PQ node; the status information of the PV node includes: active power and voltage amplitude; the status information of the PQ node includes: active power and reactive power.

[0048] For example, in order to perform power flow regression calculation on the distribution network, it is first necessary to obtain the status information of each node of the distribution network. When the distribution network only includes PV nodes, only the status information of the PV nodes can be obtained, including: active power and voltage amplitude; when the distribution network only includes PQ nodes, only the status information of the PQ nodes can be obtained, including: active power and reactive power; when the distribution network includes both PV nodes and PQ nodes, the status information of both nodes can be obtained at the same time.

[0049] It should be noted that different types of nodes have different bus types. The bus type of a PV node is a PV bus, and the bus type of a PQ node is a PQ bus.

[0050] Step 102: input the target input information into the power flow regression model to obtain the target output result.

[0051] Among them, the target output result includes at least one of the following: the reactive power and voltage phase angle of the PV node, the voltage amplitude and voltage phase angle of the PQ node; the power flow regression model is constructed based on the support vector regression SVR model; the objective function of the power flow regression model is: obtained based on the second mapping relationship between known quantities and unknown quantities; the second mapping relationship is determined based on the power flow equation corresponding to different bus types of the distribution network; the kernel function is used in the power flow regression model to replace the characteristic mapping inner product function of the SVR model; the kernel function is used to reduce the calculation dimension of the power flow regression model.

[0052] Exemplarily, after obtaining the above target input information, the target input information is input into the power flow regression model to obtain the target output result. The target output result includes the output result corresponding to each node in the target input information. The output result of the PV node includes: the reactive power and voltage phase angle of the PV node; the output result of the PQ node includes: the voltage amplitude and voltage phase angle of the PQ node.

[0053] Optionally, in the embodiment of the present application, firstly, the mapping relationship between known and unknown quantities of different types of buses is established; secondly, the updating method of bus type transformation of the model is further derived; finally, a power flow regression model based on SVR is constructed to better fit the nonlinearity of power flow by embedding Gaussian kernel function. The proposed method is simulated in the improved IEEE33 node system, and it is verified that the proposed method can effectively obtain high-accuracy power flow solutions before and after bus type transformation in the scenario of unobservable distribution network and outliers.

[0054] For example, in the embodiment of the present application, the steps for constructing the above power flow regression model may include the following steps 201 to 204: Step 201: establish a first mapping relationship between power and voltage, and determine known quantities and unknown quantities corresponding to different bus types based on the bus types of the distribution network nodes.

[0055] The power includes active power and reactive power; the voltage includes voltage amplitude and voltage phase angle.

[0056] For example, before building a power flow regression model, it is necessary to first establish data-driven power flow equations under different bus types. For traditional physical model power flow calculation, the power equation of the node is written through Kirchhoff's theorem, where the physical model power in polar coordinate form is shown in the following formula 1:

[0057] in, i =1,2,3,…,n, is the active power injection on bus i, For bus i Reactive power injection on For bus i The voltage value, For bus k Voltage value; as well as is the admittance matrix Y = G + j · B In ( i , k )element; j is an imaginary unit, For bus i and busbar k The phase angle difference between them.

[0058] Exemplarily, with reference to the independent variable and the dependent variable of the above formula 1, firstly, a mapping relationship between active power and reactive power (i.e., the above first mapping relationship) and a mapping relationship between voltage amplitude and voltage phase angle (i.e., the above second mapping relationship) are established. The first mapping relationship equation constructed based on the above two mapping relationships can be specifically expressed by the following formula 2: (Formula 2)

[0059] in, H, N, M, L are all regression parameter matrices; and is a constant term.

[0060] Step 202: construct a first mapping relationship equation based on the first mapping relationship, and convert the first mapping relationship equation into a second mapping relationship equation based on known quantities and unknown quantities corresponding to different bus types.

[0061] Exemplarily, the known quantities and unknown quantities of the distribution network are divided according to different bus types; the bus types in the distribution network and their corresponding known and unknown variables are shown in the following Table 1:

[0062] Table 1 Busbar types of distribution network

[0063] For example, when considering different bus types, according to the power flow equation (i.e., the above formula 1), the active and reactive power of different bus types are first used as dependent variables, and the different types of buses and phase angles are used as independent variables. Considering that there is generally only one bus in the distribution network, The busbar is a (balanced busbar), whose voltage is given, the phase angle is generally set to 0, and the active power can be almost determined by the active power injection of other buses; regardless of Busbar or The active power of the bus is an independent variable, and The active power of the bus is the dependent variable. If it is added, it will lead to regression collinearity. In summary, The mother line is removed from the regression, so Formula 2 is converted to Formula 3 below, that is, the above second mapping relationship equation is expressed by Formula 3 below: (Formula 3)

[0064] Among them, the subscript L express PQ Busbar, subscript S express PV Busbar; is the regression parameter matrix; is a constant term; V is the voltage amplitude, is the voltage phase angle, P is the active power, Q is the reactive power. The task of power flow analysis is to calculate the unknown variables based on the known variables. and to After all the parameters are set, power flow mapping can be performed.

[0065] Step 203: Solve the regression parameter matrix and the constant term in the second mapping relationship equation, and perform power flow mapping based on the solved second mapping relationship equation.

[0066] Exemplarily, the specific steps of solving the regression parameter matrix and the constant term in the second mapping relationship equation are as follows: First, let , ;in, and is a known quantity, and is an unknown quantity; therefore, the above formula 3 can be equivalent to the following formula 4: (Formula 4)

[0067] in, is the simplified regression parameter matrix.

[0068] Then, the unknown variables in the above formula 4 are calculated by the following formula 5:

[0069] For example, the unknown quantity can be calculated by the above formula 5: and .

[0070] For example, after calculating the above unknown quantities, a second mapping relationship between the known quantities and the unknown quantities can be established through power flow mapping. The specific steps are as follows: First, replace the above formula 5 Substitute into In the formula 6, we get: (Formula 6)

[0071] in, ; ; .

[0072] Afterwards, After conversion, we get the following formula 7: (Formula 7)

[0073] In the formula, ; ; .

[0074] Finally, a mapping relationship between known quantities and unknown quantities is established, that is, the second mapping relationship can be expressed by the following formulas 8 and 9: (Formula 8)

[0075] (Formula 9)

[0076] in, and is a known quantity, and is the unknown quantity; K is calculated based on the regression parameter matrix; is a constant term.

[0077] For example, according to the above formula 9, without changing the bus type, after the regression model training is completed, only one matrix calculation is required to obtain the unknown quantity and There is no need for complex calculation process or calculation of inverse matrix, which ensures the calculation efficiency.

[0078] Step 204: establishing the objective function based on the second mapping relationship, constructing the kernel function based on the regression parameter matrix, and constructing the power flow regression model based on the objective function and the kernel function.

[0079] Exemplarily, the specific steps of establishing the objective function are as follows: First, the above formula 9 is for a single operation snapshot. In order to consider the regression problem of time asynchrony, the above formula 9 needs to be converted into the following formula 10:

[0080] Where W represents the regression parameter matrix, represents the dependent variable matrix, represents the independent variable matrix, and b is the constant term. It corresponds to the regression parameter matrix in the above formula 9 , dependent variable, independent variable and constant term In the data-driven trend of returning, and It is a matrix rather than a vector. The number of columns represents different data sets of the distribution network at different times, and the number of rows represents different variables.

[0081] Afterwards, before establishing the above-mentioned power flow regression model, a candidate model must first be determined, which approximates the target function of the input and output relationship. In the embodiment of the present application, an effective approximate function related to the above-mentioned formula 9 is selected. .

[0082] Exemplarily, the objective function is expressed by the following formula 11:

[0083] in, W represents the regression parameter matrix, represents the dependent variable matrix, represents the independent variable matrix; represents the Frobenius norm; C is the penalty coefficient; and is the slack variable; is the tolerance deviation; the penalty coefficient is used to constrain the tolerance deviation.

[0084] For example, the Frobenius norm means the square root of the sum of the squares of each number in the matrix. To minimize the norm, the matrix Minimum is equivalent to maximizing the margin, that is, (the "distance" to the farthest sample point on the hyperplane is the smallest). SVR creates a "gap" on both sides of the function with a spacing of (Also called tolerance deviation, which is an empirical value set manually), which indicates the tolerance and There are at most When the time difference between the two is less than When the deviation between the two is not less than When , the loss is not 0, which makes the regression less sensitive to abnormal data, which is very friendly to today's distribution network. This loss can be expressed by formula 12.

[0085] Exemplarily, the above power flow regression model uses the following loss function to train the model: (Formula 12)

[0086] in, is the dependent variable matrix, is the independent variable matrix, Tolerance for deviations; and These are the output results of the model.

[0087] It should be noted that in real tasks, it is often difficult to directly determine the appropriate , ensuring that most of the data are within the interval band. If ignoring it will lead to learning errors. However, this paper hopes that all training data are within the interval band, so a slack variable is added and , which is to strike a balance between tolerating bad data and identifying the correct power flow mapping, while also preventing overfitting. C is the penalty coefficient, which serves as a constraint role.

[0088] For example, for the above tolerance deviation and penalty coefficient C The solution and function selection can be achieved through the following steps: First, use k-fold cross validation to select tuning parameters C and . When all data is used to train the model, overfitting may occur. By using k-fold cross validation, the model can be "tested" on k different data sets to prevent overfitting. First, set k=10 and split the training data set into k parts; second, in the k-fold samples, (k-1) parts are used for training and 1 part is used for validation, and the performance of each model is recorded; finally, the performance of all k models is compared to select the optimal parameters.

[0089] Afterwards, we can know that the above convex optimization problem can first be solved by the Lagrange multiplier method, and then equivalent to the dual problem through the KKT conditions (Karush–Kuhn–Tucker conditions, KKT conditions): (Formula 13)

[0090] (Formula 14)

[0091] in, and are the corresponding Lagrange multipliers in constraint formula 13 and formula 14, is the inner product form. After solving the dual problem, use Formula 13 from and Recover the regression matrix from the solution of . Hence, they are the weights given to the features to determine which data samples become the support vectors so that the located hyperplane is optimal.

[0092] For example, kernel functions generally have two benefits: solving inherent nonlinear problems and reducing the amount of calculation. When using formula 14 for regression, it is nonlinear because formula 14 contains and The inner product of , so the feature map can be selected in the RHKS (Hilbert kernel space) Hilbert kernel space Project the variables into a new high-dimensional feature space, thereby realizing the mapping of the feature space from low dimension to high dimension. Instead of the inner product, the nonlinearity of the power flow can be better captured, which can be specifically expressed by the following formula 15 and formula 16: (Formula 15)

[0093] (Formula 16)

[0094] However, since the amount of calculation of the high-dimensional function after mapping is unacceptably large and difficult to solve, in order to solve this problem, the kernel function can be used instead of the feature mapping inner product function, as shown in the following formula 17: (Formula 17)

[0095] Exemplarily, the kernel function is an extension of the vector inner product space, which not only makes the nonlinear regression problem better capture the nonlinearity of the trend after the conversion of the kernel function, but also can be equivalently implemented through low-dimensional and low-cost calculations, avoiding high-dimensional calculations. The kernel function in the embodiment of the present application can be expressed by the following formula 18:

[0096] in, and is the Lagrange multiplier, represents the independent variable matrix, and b is the constant term.

[0097] For example, Figure 3 As shown, the embodiment is to perform simulation analysis on the improved IEEE33 node example, and add PV type distributed photovoltaic to 6, 12, and 24 buses on the IEEE33 node system. Different kernel functions have different effects on regression. In order to select a suitable kernel function, the present invention selects four most commonly used kernel functions: Gaussian kernel function, polynomial second-order kernel function, polynomial third-order kernel function and hyperbolic tangent kernel function for comparison on the IEEE33 node system. Figure 4 The average RMSE of the regression accuracy of the node voltage and phase of all 29 PQ loads in the node system is given. Figure 4 As shown in the figure, the Gaussian kernel function has the smallest RMSE and the highest accuracy. The Gaussian kernel function is nonlinear and suitable for the nonlinear characteristics of the power flow equation. Therefore, the Gaussian kernel function is selected as the kernel function of SVR. The simulation platform is MATLAB. The data used is the historical load and photovoltaic power generation data of a company, which includes smart meter and PMU readings. The data is sampled every fifteen minutes to generate 2688 sets of power flow sections as data sets. The prepared data is divided into 70% for training the model, 10% for verifying the model's hyperparameters (such as C in SVR), and 20% for testing.

[0098] Due to the unobservability problem in the distribution network, the present invention selects 10% of the buses to be unmeasurable (no voltage and power measurement) during the learning process. In addition, in order to consider the measurement error, random noise (with a mean of zero and a relative standard deviation of 1%) that obeys a Gaussian distribution is added to the data. 1% of the data is randomly changed to anomalies, increasing their values ​​by 5-10 times.

[0099] For comparison, representative learning models are selected to compare with the method proposed in this invention: NN, BYS, and LS. Specifically, NN uses a 3-layer multilayer perceptron with ReLu activation function; BYS regards parameters as random variables and calculates their posteriori through the prior of model parameters (weight coefficients); LS model is an unconstrained programming model with the minimum residual sum of squares as the goal and introduces the Huber loss function. For evaluation, an intuitive method is to calculate the actual output Output of the learning model In this paper, the mean absolute error (MAE) is used to evaluate the power flow regression results of a single bus node, and the root mean square error (RMSE) is used to evaluate the power flow regression results of the entire node system. The lower the MAE and RMSE, the higher the accuracy.

[0100] It should be noted that, under the same data set, the proposed method is compared with three representative methods in MATLAB, and the voltage and phase angle regression accuracies of the PQ load node are as follows: Figure 3 , 4 shown.

[0101] The improved IEEE33-bus system has 29 PQ load nodes, 3 PV load nodes, and 1 balance node. Two of the PV load nodes are converted into PQ load nodes. The model is updated after the variables are reordered. The RMSE of the PQ load node voltage and phase angle regression accuracy before and after the bus type transformation of the four methods is shown in Table 2. The MAE of the PQ load node voltage and phase angle regression accuracy after the bus type transformation of the four methods is shown in Table 2. Figure 5 Figure 6 shown.

[0102] From Table 2, we can see that when some bus types are changed, although the calculation accuracy is slightly reduced, the accuracy remains at the same level, and the voltage and phase of the PQ node can be accurately regressed. Figure 5 Figure 6 It can be seen that the proposed SVR power flow regression method has the highest accuracy, the smallest error, and is relatively stable compared to the other three methods. Figure 7 and Figure 8 It shows that the SVR power flow regression can still maintain good accuracy after the bus type transformation. The above shows the effectiveness of the proposed bus type transformation method, and it still has good accuracy in the face of frequently changing distribution network power flow regression.

[0103]

[0104] Table 2 RMSE of voltage phase angle accuracy of different power flow regression methods after bus transformation

[0105] The above embodiments are only preferred technical solutions of the embodiments of the present application and should not be regarded as limitations on the embodiments of the present application. The protection scope of the embodiments of the present application should be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present application.

[0106] The distribution network power flow regression method based on SVR data driving provided in the embodiment of the present application first obtains target input information including status information of each node of the distribution network; the nodes of the distribution network include at least one of the following: PV nodes and PQ nodes; the status information of the PV nodes includes: active power and voltage amplitude; the status information of the PQ nodes includes: active power and reactive power; then, the target input information is input into the power flow regression model to obtain the target output result; wherein, the target output result includes at least one of the following: reactive power and voltage phase angle of the PV node, voltage amplitude and voltage phase angle of the PQ node; the power flow regression model is constructed based on the support vector regression SVR model; the objective function of the power flow regression model is: obtained based on the second mapping relationship between known quantities and unknown quantities; the second mapping relationship is determined based on the power flow equation corresponding to different bus types of the distribution network; the kernel function is used in the power flow regression model to replace the characteristic mapping inner product function of the SVR model; the kernel function is used to reduce the calculation dimension of the power flow regression model. In this way, a high-accuracy power flow solution can be calculated through the distribution network power flow regression model driven by SVR data, thereby achieving reasonable deployment of the distribution network.

[0107] It should be noted that the distribution network power flow regression method based on SVR data drive provided in the embodiment of the present application can be executed by a distribution network power flow regression device based on SVR data drive, or a control module in the distribution network power flow regression device based on SVR data drive for executing the distribution network power flow regression method based on SVR data drive. In the embodiment of the present application, the distribution network power flow regression device based on SVR data drive is taken as an example to illustrate the distribution network power flow regression device based on SVR data drive provided in the embodiment of the present application.

[0108] It should be noted that, in the embodiments of the present application, the distribution network flow regression method based on SVR data driven is illustrated in the drawings of the above-mentioned methods. All of them are illustrated by way of example in combination with one of the drawings in the embodiments of the present application. In specific implementation, the distribution network flow regression method based on SVR data driven shown in the drawings of the above-mentioned methods can also be implemented in combination with any other drawings that can be combined as shown in the above-mentioned embodiments, which will not be described in detail here.

[0109] The following is a description of the distribution network power flow regression device based on SVR data driving provided in the present application. The distribution network power flow regression method based on SVR data driving described below and described above can refer to each other.

[0110] Fig. 9 A schematic diagram of the structure of a distribution network power flow regression device driven by SVR data provided in an embodiment of the present application is shown in FIG. Fig. 9As shown, specifically including: A data acquisition module 901 is used to acquire target input information including status information of each node of a distribution network; the nodes of the distribution network include at least one of the following: a PV node and a PQ node; the status information of the PV node includes: active power and voltage amplitude; the status information of the PQ node includes: active power and reactive power; a power flow calculation module 902 is used to input the target input information into a power flow regression model to obtain a target output result; wherein the target output result includes at least one of the following: reactive power and voltage phase angle of the PV node, voltage amplitude and voltage phase angle of the PQ node; the power flow regression model is constructed based on a support vector regression SVR model; the objective function of the power flow regression model is: obtained based on a second mapping relationship between known quantities and unknown quantities; the second mapping relationship is determined based on power flow equations corresponding to different bus types of the distribution network; a kernel function is used in the power flow regression model to replace the characteristic mapping inner product function of the SVR model; the kernel function is used to reduce the calculation dimension of the power flow regression model.

[0111] Optionally, the device also includes: a model building module; the model building module is used to establish a first mapping relationship between power and voltage, and determine the known quantities and unknown quantities corresponding to different bus types based on the bus type of the distribution network node; the power includes: active power and reactive power; the voltage includes: voltage amplitude and voltage phase angle; the model building module is also used to construct a first mapping relationship equation based on the first mapping relationship, and convert the first mapping relationship equation into a second mapping relationship equation based on the known quantities and unknown quantities corresponding to different bus types; the model building module is also used to solve the regression parameter matrix and constant terms in the second mapping relationship equation, and perform flow mapping based on the solved second mapping relationship equation; the flow mapping is used to establish a second mapping relationship between known quantities and unknown quantities; the model building module is also used to establish the objective function based on the second mapping relationship, and construct the kernel function based on the regression parameter matrix, and construct the flow regression model based on the objective function and the kernel function.

[0112] Optionally, the first mapping relationship equation is expressed by the following formula:

[0113] in, H, N, M, L are all regression parameter matrices; and is a constant term; The second mapping relationship equation is expressed by the following formula:

[0114] Among them, the subscript L express PQ Busbar, subscript S express PV Busbar; is the regression parameter matrix; is a constant term; V is the voltage amplitude, is the voltage phase angle, P is the active power, Q is the reactive power.

[0115] Optionally, the second mapping relationship is expressed by the following formula:

[0116] in, and is a known quantity, and is the unknown quantity; K is calculated based on the regression parameter matrix; is a constant term.

[0117] Optionally, the objective function is expressed by the following formula:

[0118] in, W represents the regression parameter matrix, represents the dependent variable matrix, represents the independent variable matrix; represents the Frobenius norm; C is the penalty coefficient; and is the slack variable; is the tolerance deviation; the penalty coefficient is used to constrain the tolerance deviation.

[0119] Optionally, the kernel function is expressed by the following formula:

[0120] in, and is the Lagrange multiplier, represents the independent variable matrix, b is a constant term.

[0121] Optionally, the power flow regression model uses the following loss function to train the model:

[0122] in, is the dependent variable matrix, is the independent variable matrix, Tolerance for deviations; is the output of the model.

[0123] The distribution network power flow regression device based on SVR data drive provided in the present application first obtains target input information including status information of each node of the distribution network; the nodes of the distribution network include at least one of the following: PV nodes and PQ nodes; the status information of the PV nodes includes: active power and voltage amplitude; the status information of the PQ nodes includes: active power and reactive power; then, the target input information is input into the power flow regression model to obtain the target output result; wherein, the target output result includes at least one of the following: reactive power and voltage phase angle of the PV node, voltage amplitude and voltage phase angle of the PQ node; the power flow regression model is constructed based on the support vector regression SVR model; the objective function of the power flow regression model is: obtained based on the second mapping relationship between known quantities and unknown quantities; the second mapping relationship is determined based on the power flow equation corresponding to different bus types of the distribution network; the characteristic mapping inner product function of the SVR model is replaced by a kernel function in the power flow regression model; the kernel function is used to reduce the calculation dimension of the power flow regression model. In this way, a high-accuracy power flow solution can be calculated through the distribution network power flow regression model driven by SVR data, thereby achieving reasonable deployment of the distribution network.

[0124] Fig.10 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.10As shown, the electronic device may include: a processor (processor) 1010 , a communication interface (Communications Interface) 1020 , a memory (memory) 1030 and a communication bus 1040 , wherein the processor 1010 , the communication interface 1020 , and the memory 1030 communicate with each other via the communication bus 1040 . The processor 1010 can call the logic instructions in the memory 1030 to execute the distribution network power flow regression method driven by SVR data, which includes: obtaining target input information including status information of each node of the distribution network; the nodes of the distribution network include at least one of the following: PV nodes and PQ nodes; the status information of the PV nodes includes: active power and voltage amplitude; the status information of the PQ nodes includes: active power and reactive power; inputting the target input information into the power flow regression model to obtain a target output result; wherein the target output result includes at least one of the following: reactive power and voltage phase angle of the PV node, voltage amplitude and voltage phase angle of the PQ node; the power flow regression model is constructed based on the support vector regression SVR model; the objective function of the power flow regression model is: obtained based on the second mapping relationship between known quantities and unknown quantities; the second mapping relationship is determined based on the power flow equation corresponding to different bus types of the distribution network; in the power flow regression model, a kernel function is used to replace the characteristic mapping inner product function of the SVR model; the kernel function is used to reduce the calculation dimension of the power flow regression model.

[0125] In addition, the logic instructions in the above-mentioned memory 1030 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0126] On the other hand, the present application also provides a computer program product, the computer program product includes a computer program stored on a computer-readable storage medium, the computer program includes program instructions, when the program instructions are executed by the computer, the computer can execute the distribution network power flow regression method based on SVR data driven provided by the above methods, the method includes: obtaining target input information including status information of each node of the distribution network; the nodes of the distribution network include at least one of the following: PV nodes and PQ nodes; the status information of the PV nodes includes: active power and voltage amplitude; the status information of the PQ nodes includes: active power and reactive power; the target The target input information is input into the power flow regression model to obtain the target output result; wherein, the target output result includes at least one of the following: the reactive power and voltage phase angle of the PV node, the voltage amplitude and voltage phase angle of the PQ node; the power flow regression model is constructed based on the support vector regression SVR model; the objective function of the power flow regression model is: obtained based on the second mapping relationship between known quantities and unknown quantities; the second mapping relationship is determined based on the power flow equation corresponding to different bus types of the distribution network; the kernel function is used in the power flow regression model to replace the characteristic mapping inner product function of the SVR model; the kernel function is used to reduce the calculation dimension of the power flow regression model.

[0127] On the other hand, the present application also provides a computer-readable storage medium having a computer program stored thereon, which is implemented when the processor executes the above-mentioned distribution network power flow regression method based on SVR data drive, the method comprising: obtaining target input information including status information of each node of the distribution network; the nodes of the distribution network include at least one of the following: PV nodes and PQ nodes; the status information of the PV nodes includes: active power and voltage amplitude; the status information of the PQ nodes includes: active power and reactive power; inputting the target input information into the power flow regression model to obtain a target output result; wherein the target output result includes at least one of the following: reactive power and voltage phase angle of the PV node, voltage amplitude and voltage phase angle of the PQ node; the power flow regression model is constructed based on the support vector regression SVR model; the objective function of the power flow regression model is: obtained based on a second mapping relationship between known quantities and unknown quantities; the second mapping relationship is determined based on the power flow equation corresponding to different bus types of the distribution network; in the power flow regression model, a kernel function is used to replace the characteristic mapping inner product function of the SVR model; the kernel function is used to reduce the calculation dimension of the power flow regression model.

[0128] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0129] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A distribution network power flow regression method based on SVR data drive, characterized in that: include: Obtain target input information including status information of each node in the distribution network; The nodes of the distribution network include at least one of the following: a PV node and a PQ node; The state information of the PV node includes: active power and voltage amplitude; the state information of the PQ node includes: active power and reactive power; Inputting the target input information into the power flow regression model to obtain the target output result; Among them, the target output result includes at least one of the following: the reactive power and voltage phase angle of the PV node, the voltage amplitude and voltage phase angle of the PQ node; the power flow regression model is constructed based on the support vector regression SVR model; the objective function of the power flow regression model is: obtained based on the second mapping relationship between known quantities and unknown quantities; the second mapping relationship is determined based on the power flow equation corresponding to different bus types of the distribution network; the kernel function is used in the power flow regression model to replace the characteristic mapping inner product function of the SVR model; the kernel function is used to reduce the calculation dimension of the power flow regression model.

2. The method according to claim 1, characterized in that The power flow regression model is constructed based on the following steps: Establishing a first mapping relationship between power and voltage, and determining known quantities and unknown quantities corresponding to different bus types based on the bus type of the distribution network node; the power includes: active power and reactive power; the voltage includes: voltage amplitude and voltage phase angle; Constructing a first mapping relationship equation based on the first mapping relationship, and converting the first mapping relationship equation into a second mapping relationship equation based on known quantities and unknown quantities corresponding to different bus types; Solving the regression parameter matrix and the constant term in the second mapping relationship equation, and performing power flow mapping based on the solved second mapping relationship equation; the power flow mapping is used to establish a second mapping relationship between known quantities and unknown quantities; The objective function is established based on the second mapping relationship, the kernel function is constructed based on the regression parameter matrix, and the power flow regression model is constructed based on the objective function and the kernel function.

3. The method according to claim 2, characterized in that The first mapping relationship equation is expressed by the following formula: in, H, N, M, L are all regression parameter matrices; and is a constant term; The second mapping relationship equation is expressed by the following formula: Among them, the subscript L express PQ Busbar, subscript S express PV Busbar; is the regression parameter matrix; is a constant term; V is the voltage amplitude, is the voltage phase angle, P is the active power, Q is the reactive power.

4. The method according to claim 2, characterized in that: The second mapping relationship is expressed by the following formula: in, and is a known quantity, and is the unknown quantity; K is calculated based on the regression parameter matrix; is a constant term.

5. The method according to claim 2, characterized in that: The objective function is expressed by the following formula: in, W represents the regression parameter matrix, represents the dependent variable matrix, represents the independent variable matrix; represents the Frobenius norm; C is the penalty coefficient; and is the slack variable; is the tolerance deviation; the penalty coefficient is used to constrain the tolerance deviation.

6. The method according to claim 2, characterized in that The kernel function is expressed by the following formula: in, and is the Lagrange multiplier, represents the independent variable matrix, b is a constant term.

7. The method according to claim 2, characterized in that The flow regression model uses the following loss function to train the model: in, is the dependent variable matrix, is the independent variable matrix, Tolerance for deviations; is the output of the model.

8. A distribution network power flow regression device driven by SVR data, characterized in that: The device comprises: A data acquisition module, used to acquire target input information including status information of each node of the distribution network; the nodes of the distribution network include at least one of the following: a PV node and a PQ node; the status information of the PV node includes: active power and voltage amplitude; the status information of the PQ node includes: active power and reactive power; A power flow calculation module, used for inputting the target input information into the power flow regression model to obtain a target output result; Among them, the target output result includes at least one of the following: the reactive power and voltage phase angle of the PV node, the voltage amplitude and voltage phase angle of the PQ node; the power flow regression model is constructed based on the support vector regression SVR model; the objective function of the power flow regression model is: obtained based on the second mapping relationship between known quantities and unknown quantities; the second mapping relationship is determined based on the power flow equation corresponding to different bus types of the distribution network; the kernel function is used in the power flow regression model to replace the characteristic mapping inner product function of the SVR model; the kernel function is used to reduce the calculation dimension of the power flow regression model.

9. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the distribution network power flow regression method based on SVR data driving as claimed in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the steps of the distribution network power flow regression method based on SVR data driving as claimed in any one of claims 1 to 7 are implemented.

Images