Section transmission power probability prediction method, apparatus and device, and storage medium

Abstract

The invention discloses a section transmission power probability prediction method and device, equipment and a storage medium, and relates to the technical field of power prediction. The method comprises the following steps: inputting a power grid topological structure, and converting the power grid topological structure into a power system diagram by taking a bus in the power grid topological structure as a node and a line as an edge; inputting the power system diagram into a diagram neural network to obtain a node embedding vector; generating an initial value based on the power system diagram, and performing power system load flow calculation on the initial value to construct a training data set; wherein the initial value comprises load demand and unit output data, and the training data set comprises the initial value and corresponding section transmission power; inputting the node embedding vector into a spatio-temporal feature extraction model to obtain spatio-temporal feature information; and inputting the spatial-temporal characteristic information and the training data set into a mixed density network to obtain section transmission power probability distribution. According to the invention, the accuracy and efficiency of power probability prediction can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power prediction, and particularly to a method, device, equipment and storage medium for predicting the probability of section transmission power. Background Art

[0002] In a power system, in order to ensure the transmission and distribution of electric energy, it is necessary to monitor and control the transmission power of different sections. The section transmission power refers to the power of the electric energy passing through a certain set of specific devices (such as transmission lines, transformers, etc.) in the power system. Among them, the section is usually a certain node or line in the power grid, and the power passing through this section can reflect the load and energy transmission situation of this part of the power grid.

[0003] The accurate prediction of the section transmission power is crucial for the safe and stable operation of the power system. By real-time monitoring and analyzing the transmission power of different sections, the grid operator can timely discover and solve potential problems, avoid overload and faults, and ensure the reliable operation of the grid.

[0004] However, traditional methods model the power system based on matrices or vectors, resulting in the inability to efficiently process unstructured data and also unable to adapt to the dynamically changing grid topology, reducing the accuracy of the power prediction results. Summary of the Invention

[0005] The purpose of the embodiments of the present invention is to provide a method, device, equipment and storage medium for predicting the probability of section transmission power, which can improve the accuracy and efficiency of power probability prediction to assist the safety analysis and power dispatching of the power system.

[0006] To achieve the above purpose, the embodiments of the present invention provide a method for predicting the probability of section transmission power, including:

[0007] Input the grid topology structure, take the buses in the grid topology structure as nodes and the lines as edges, and convert the grid topology structure into a power system graph;

[0008] Input the power system graph into a graph neural network to obtain node embedding vectors;

[0009] Generate an initial value based on the power system graph, perform power system load flow calculation on the initial value to construct a training data set; wherein, the initial value includes load demand and unit output data, and the training data set includes the initial value and the corresponding section transmission power;

[0010] Input the node embedding vectors into a spatio-temporal feature extraction model to obtain spatio-temporal feature information;

[0011] Input the spatio-temporal feature information and the training data set into the mixture density network to obtain the probability distribution of the section transmission power.

[0012] As an improvement to the above solution, the spatio-temporal feature extraction model includes a convolutional neural network and a long short-term memory network, and the spatio-temporal feature information is obtained in the following way:

[0013] Input the node embedding vector into the convolutional neural network to obtain intermediate features;

[0014] Input the intermediate features and the training data set into the long short-term memory network to obtain spatio-temporal feature information.

[0015] As an improvement to the above solution, the probability distribution of the section transmission power is shown as follows:

[0016]

[0017] where, P ′ (y|x) represents the probability distribution of the section transmission power; y represents the target vector; x represents the spatio-temporal feature information; N(y|μ i (x), σ i (x)) represents the conditional density of the target vector of the i-th kernel; g represents the total number of kernels of the mixture density network; μ i (x) represents the center of the i-th kernel; σ i (x) represents the common variance.

[0018] As an improvement to the above solution, the expression of N(y|μ i (x), σ i (x)) is shown as follows:

[0019]

[0020] where, y represents the target vector; x represents the spatio-temporal feature information; π i (x) represents the mixing coefficient; N(y|μ i (x), σ i (x)) represents the conditional density of the target vector of the i-th kernel; μ i (x) represents the center of the i-th kernel; σ i (x) represents the common variance; c represents the dimension of the target vector.

[0021] As an improvement to the above solution, the training data set is constructed in the following way:

[0022] Initialize the load demand power of each edge and the unit output data of each bus node in the power system diagram; among them, the unit output data includes the active power and reactive power injected by the generator into the bus node;

[0023] Based on the unit output data and the load demand power, construct a power balance equation;

[0024] Iteratively solve the power balance equation to obtain the section transmission power under steady state;

[0025] Take the unit output data, the load demand power, and the corresponding section transmission power as a set of data and add them to the training dataset.

[0026] As an improvement to the above solution, in each iteration, calculate the voltage magnitude correction amount and the voltage phase angle correction amount for each bus node, and based on the voltage magnitude correction amount and the voltage phase angle correction amount, correct the voltage magnitude and voltage phase angle of each bus node, and then enter a new round of iteration;

[0027] When both the voltage magnitude correction amount and the voltage phase angle correction amount are less than the preset error upper limit, calculate the section transmission power between the bus nodes based on the current voltage magnitude of each bus node and the current voltage phase angle difference between the bus nodes.

[0028] As an improvement to the above solution, the inputting the power system diagram into the graph neural network to obtain node embedding vectors includes:

[0029] Input the initial feature matrix and the adjacency matrix of the power system diagram;

[0030] Based on the adjacency matrix, aggregate the nodes in the power system diagram to obtain the hidden layer features;

[0031] Perform linear regression prediction on the hidden layer features to obtain node embedding vectors.

[0032] To achieve the above object, an embodiment of the present invention further provides a section transmission power probability prediction device, including:

[0033] A graph construction module, configured to input a power grid topology structure, use the buses in the power grid topology structure as nodes, use the lines as edges, and convert the power grid topology structure into a power system diagram;

[0034] A graph feature extraction module, configured to input the power system diagram into the graph neural network to obtain node embedding vectors;

[0035] A training data generation module, configured to generate an initial value based on the power system diagram, perform power system power flow calculation on the initial value to construct a training dataset; wherein, the initial value includes load demand and unit output data, and the training dataset includes the initial value and the corresponding section transmission power;

[0036] A spatio-temporal feature information extraction module, configured to input the node embedding vector into a spatio-temporal feature extraction model to obtain spatio-temporal feature information;

[0037] A power probability prediction module, configured to input the spatio-temporal feature information and the training data set into a mixture density network to obtain a cross-section transmission power probability distribution.

[0038] To achieve the above object, an embodiment of the present invention further provides a cross-section transmission power probability prediction device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the cross-section transmission power probability prediction method described in any of the above embodiments is implemented.

[0039] To achieve the above object, an embodiment of the present invention further provides a computer-readable storage medium. The computer-readable storage medium includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the cross-section transmission power probability prediction method described in any of the above embodiments.

[0040] Compared with the prior art, the cross-section transmission power probability prediction method, device, equipment, and storage medium provided by the embodiments of the present invention input a power grid topology structure, use the buses in the power grid topology structure as nodes, use the lines as edges, and convert the power grid topology structure into a power system graph; input the power system graph into a graph neural network to obtain node embedding vectors; generate initial values based on the power system graph, and perform power system power flow calculations on the initial values to construct a training data set. The initial values include load demand and unit output data, and the training data set includes the initial values and the corresponding cross-section transmission power; input the node embedding vectors into a spatio-temporal feature extraction model to obtain spatio-temporal feature information; input the spatio-temporal feature information and the training data set into a mixture density network to obtain a cross-section transmission power probability distribution. Compared with the prior art, the embodiments of the present invention can improve the accuracy and efficiency of power probability prediction. Description of the Drawings

[0041] Figure 1 is a flowchart of a cross-section transmission power probability prediction method provided by an embodiment of the present invention;

[0042] Figure 2 is a schematic diagram of a graph convolutional neural network and a multi-task regression module provided by an embodiment of the present invention;

[0043] Figure 3 is a schematic diagram of a self-weighted loss training module provided by an embodiment of the present invention;

[0044] Figure 4It is a schematic structural diagram of a cross-section transmission power probability prediction device provided by an embodiment of the present invention;

[0045] Figure 5 It is a schematic structural diagram of a cross-section transmission power probability prediction device provided by an embodiment of the present invention. Specific embodiments

[0046] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

[0047] In the description of the present application, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present application, unless otherwise specified, the meaning of "plurality" is two or more.

[0048] In the description of the present application, it should be noted that, unless otherwise clearly defined and limited, the terms "installation", "connection", and "connection" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected, or indirectly connected through an intermediate medium, and it may be the internal communication of two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific situations.

[0049] See Figure 1 , which is a flowchart of a cross-section transmission power probability prediction method provided by an embodiment of the present invention, including steps S1 to S5:

[0050] S1. Input the power grid topology structure, use the buses in the power grid topology structure as nodes, use the lines as edges, and convert the power grid topology structure into a power system diagram;

[0051] S2. Input the power system diagram into a graph neural network to obtain node embedding vectors;

[0052] S3. Generate initial values based on the power system diagram, perform power system power flow calculations on the initial values to construct a training data set; wherein, the initial values include load demand and unit output data, and the training data set includes the initial values and the corresponding cross-section transmission power;

[0053] S4. Input the node embedding vectors into the spatio-temporal feature extraction model to obtain spatio-temporal feature information;

[0054] S5. Input the spatio-temporal feature information and the training data set into the mixture density network to obtain the probability distribution of the section transmission power.

[0055] In step S1, it can be understood that the line can be regarded as a transmission line. In the power grid topology, the lines connect the buses. Therefore, in the embodiments of the present invention, the buses are used as the nodes of the power system diagram, and the lines are used as the edges of the power system diagram. It should be noted that in the following text, the "node" is also referred to as the "bus node".

[0056] Furthermore, in step S2, by inputting the power system diagram into the graph neural network, not only can the efficiency of feature extraction be improved, but also the features of the power grid topology can be extracted globally. Compared with the prior art that only extracts local features, the embodiments of the present invention can extract the global features of the power grid topology, and the features extracted by the embodiments of the present invention include not only the power grid connection relationship, but also the attribute information of each element in the power grid topology, thereby improving the accuracy and reliability of subsequent power probability prediction.

[0057] Furthermore, in step S3, the training data set includes multiple groups of data. Each group of data includes generator output data, load demand, and the section transmission power between bus nodes. Among them, the generator output data and the load demand need to be randomly initialized, and the section transmission power between bus nodes is calculated based on the generator output data and the load demand. Further, the generator output data includes the active power and reactive power injected by the generator into the bus node, and the load demand is the power required by the load.

[0058] Furthermore, in step S4, the spatio-temporal feature extraction model may include a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM). In the embodiments of the present invention, the temporal information of the graph features is extracted by the CNN and the LSTM, and spatio-temporal information fusion is performed.

[0059] Furthermore, in step S5, the future section power level is predicted based on the Mixture Density Networks (MDN). The spatio-temporal feature information extracted by the LSTM is used as the input, and the probability distribution of the section transmission power is output, thereby more comprehensively reflecting the uncertainty and risk of the power on the transmission line and providing more accurate information for the operation of the power system.

[0060] Further, after obtaining the cross-section transmission power probability distribution model, in some embodiments, the power grid can be analyzed for security based on the cross-section transmission power probability distribution model, or power dispatching can be performed based on the cross-section transmission power probability distribution model, which is not limited herein.

[0061] Compared with the prior art, the embodiments of the present invention can more accurately capture the characteristics of the power grid topology by introducing graph neural networks, and combined with the spatio-temporal feature extraction model and the mixture density network, can improve the accuracy and efficiency of power probability prediction.

[0062] As an optional implementation manner, in step S2, inputting the power system graph into the graph neural network to obtain node embedding vectors includes:

[0063] Inputting the initial feature matrix and the adjacency matrix of the power system graph;

[0064] Aggregating the nodes in the power system graph based on the adjacency matrix to obtain hidden layer features;

[0065] Performing linear regression prediction on the hidden layer features to obtain node embedding vectors.

[0066] It can be understood that after obtaining the power system graph, graph feature extraction is required. In some embodiments, the graph neural network mainly includes three modules: a graph convolutional neural network (Graph Convolutional Networks, GCN) module, a multi-task regression module, and a self-weighted loss training module.

[0067] See Figure 2 , which is a schematic diagram of the graph convolutional neural network and the multi-task regression module (i.e., Figure 2 's "multi-branch fully connected layer") provided by an embodiment of the present invention. As can be seen from Figure 2 , the graph convolutional neural network module includes an input module and an intermediate layer, and the function of the graph convolutional neural network module is to learn the implicit information between nodes through a preset adjacency matrix A and generate hidden layer features using the aggregation rule of spectral domain convolution, where the hidden layer features contain the structural information between nodes, and linear regression prediction is performed through the connected multi-branch fully connected layer to obtain the node embedding vectors of different nodes. It can be understood that the graph convolutional neural network module has good scalability and can adapt to more complex power system scenarios; and through the graph convolutional neural network module and the multi-task regression module, the information between different nodes is fully transmitted and fused, improving the performance of the model. See Figure 3 , which is a schematic diagram of the self-weighted loss training module provided by an embodiment of the present invention. In Figure 3Among them, except for the solid arrows marked indicating backpropagation, the other solid arrows represent forward propagation. The self-weighted loss training module is introduced to further improve the model robustness and parameter identification accuracy. By reforming the original loss function that averages each node, the feature distribution of the intermediate layer of the graph convolutional neural network will not become overly sharp due to the differences between different nodes, realizing outlier peak shaving. Compared with the prior art, the embodiment of the present invention can improve the model robustness by introducing the self-weighted loss training module, can avoid overfitting and noise interference, and improves the generalization ability of the model.

[0068] Further, it can be seen from Figure 2 the left box that, different from image data, the power grid branch parameter data source is in the form of a two-dimensional matrix. Each row represents a time series data record, and the number of columns represents the number of operating parameters measured by the SCADA (Supervisory Control And Data Acquisition) system. Moreover, based on the requirements of the graph neural network for the characteristics of input data, it is necessary to structure the input layer data into graph data.

[0069] The multi-task regression module uses an MLP (Multilayer Perceptron) as the regressor. Specifically, according to the dimension of the node embedding vector, the Sigmoid function is selected as the activation function, which can prevent a large number of outlier data from appearing. This model can predict multiple node embedding vectors simultaneously, thereby improving the efficiency and accuracy of the model.

[0070] The input is divided into two parts: the input feature matrix X with an input dimension of (N, F0), where N is the number of nodes in the network and F0 is the number of input features of each node; the structural description of the graph: the adjacency matrix A ∈ R N×N . Therefore, the feature H of the hidden layer after the input passes through the graph neural network can be expressed as:

[0071] H (l+1) = f(H (l) , A) (1)

[0072] where H (0) = X, the dimension of the hidden layer feature matrix H (l) is N×F(1), and f(·) is the information aggregation rule of the graph neural network. In each layer, the graph neural network uses the propagation rule f(·) to aggregate this information to form the features of the lower layer. Under this framework, various variants of the graph neural network only differ in the propagation rule f(·).

[0073] In machine learning, computational complexity implies stronger fitting ability and robustness. Our model needs to consider the topological structure of the branches and cannot ignore the influence of the original branch characteristics on parameter identification. GCN uses the operation of the convolution kernel to convolve the data with the connection relationship of the power grid branches. Define the Laplacian matrix of the graph L = D - A, normalize the Laplacian matrix to Δ, and decompose the normalized matrix Δ.

[0074]

[0075] Among them, △ represents the normalized matrix; I represents the identity matrix; D represents the degree matrix; A represents the adjacency matrix.

[0076] Furthermore, introduce Chebyshev polynomials to replace the convolution kernel to reduce the time complexity of the model based on Equation (2). According to the definition of Chebyshev polynomials, the convolution of the graph neural network can be approximately defined as shown in Equation (5):

[0077]

[0078] Among them, X represents the input feature; represents the convolution kernel or filter in the graph convolution operation, which is used to extract the local features of the graph; represents the parameter of the convolution kernel; represents the degree matrix of; is the alternative matrix of the adjacency matrix A, and I ∈ R n×n is the identity matrix.

[0079] It should be noted that using to replace A as the adjacency matrix is because when the input features increase exponentially, only calculating the weighted sum of the features of all neighbors of a node will ignore the features of the node itself. Therefore, an identity matrix I is added to A to replace the adjacency matrix A.

[0080] Thus, the convolution aggregation rule on the graph is obtained, and the final hidden layer features can be expressed as shown in Equation (4):

[0081]

[0082] Among them, represents the convolution aggregation rule of the graph neural network; W (l) represents the training weight matrix of the l-th layer; σ represents the non-linear activation function.

[0083] As an optional implementation manner, the training data set is constructed in the following way:

[0084] Initialize the load demand power of each edge and the generator output data of each bus node in the power system diagram; wherein, the generator output data includes the active power and reactive power injected by the generator into the bus node.

[0085] Based on the generator output data and the load demand power, construct a power balance equation.

[0086] Iteratively solve the power balance equation to obtain the section transmission power under steady state.

[0087] Add the generator output data, the load demand power, and the corresponding section transmission power as a set of data to the training dataset.

[0088] As an optional implementation manner, in each iteration, calculate the voltage magnitude correction amount and voltage phase angle correction amount of each bus node, and based on the voltage magnitude correction amount and the voltage phase angle correction amount, correct the voltage magnitude and voltage phase angle of each bus node, and then enter a new round of iteration.

[0089] When both the voltage magnitude correction amount and the voltage phase angle correction amount are less than the preset error upper limit, calculate the section transmission power between the bus nodes based on the current voltage magnitude of each bus node and the current voltage phase angle difference between the bus nodes.

[0090] Exemplarily, use the np.random.uniform function in the NumPy library to randomly generate generator output data and load demand power, and then perform power flow calculation of the power system. The power flow calculation aims to calculate the active power, reactive power, voltage magnitude, and voltage phase angle of each node in the power grid under steady state through the given partial known power grid parameters, that is, the power flow initial values.

[0091] In the actual power flow calculation of the power system, the buses are often divided into three types, namely constant power buses, voltage-regulated buses, and balanced buses. First, select a bus with a generator as the balanced bus to balance the system power. The magnitude V i and phase angle θ i of the voltage at this bus are given values, and the active power injection P i and reactive power injection Q i are calculated, so it is also called a Vθ node. The remaining buses with generators are used as voltage-regulated buses. The voltage-regulated buses need to maintain their set voltage magnitude through adjustable reactive power capacity. The active power injection P i and voltage magnitude V i at this bus are given values, and the reactive power Q i and voltage phase angle θ i are to be solved, so it is also called a PV node. Finally, set the remaining nodes as constant power buses, and their active power Pi With reactive power Q i being a given value, the voltage amplitude V i and the phase angle θ i are to be determined and are thus also called PQ nodes.

[0092] Power flow calculation first gives a set of non - linear power balance equations for the target power grid through Kirchhoff’s Law, and then solves for the steady - state solution of the power flow problem of the system. In polar coordinate form, the power balance equation is shown in Equation (5):

[0093]

[0094] where N is the total number of buses in the power grid; is the active power injected at bus i, and are the active powers of the generator and load at bus i respectively; is the reactive power injected at bus i, and are the reactive powers of the generator and load at this node respectively; V i is the voltage amplitude at bus i; θ ij = θ i - θ j is the difference in voltage phase angle between bus i and bus j; G ij and B ij are the conductance and susceptance values of the wire between bus i and bus j respectively; for a power system containing sources, loads, grids, and energy storage, the power injected at bus i where is the active power of the energy storage at bus i, where indicates that the energy storage is in the charging state. Early power flow calculation methods relied on manual calculation and could only calculate the power flow of small - scale power grids.

[0095] Furthermore, the power balance power flow equation (6) is iteratively solved by the N - R method. In the k - th iteration, let:

[0096]

[0097] where the superscript (k) represents the value of the variable in the k - th iteration; and are the corrected unbalance amounts of active power and reactive power in the k - th iteration respectively.

[0098] Furthermore, the Taylor expansion of Equation (6) is carried out to obtain the corrected equation set as shown in Equation (7):

[0099]

[0100] Among them, and are the voltage magnitude correction and phase angle correction at bus i in the k-th iteration, respectively.

[0101] Furthermore, (7) is simplified into the matrix form as shown in Equation (8):

[0102]

[0103] where J is the Jacobian matrix.

[0104] Furthermore, after calculating the values of each element of the Jacobian matrix, the voltage magnitude correction and phase angle correction of each node are solved, and the new voltage values at each bus are calculated as shown in Equation (9):

[0105]

[0106] Furthermore, after obtaining the new voltage values of each node, they are substituted back into Equation (6) for the (k + 1)-th iteration to obtain the new active power and reactive power correction unbalances and and check whether they meet the convergence condition, that is, check whether they are less than the set error upper limit as shown in Equation (10):

[0107]

[0108] where ε is the set error upper limit.

[0109] If the convergence condition is met, the calculation results are output; if the convergence condition cannot be met after the number of iteration calculation rounds reaches the set upper limit, it is determined that the system power flow does not converge.

[0110] Furthermore, according to the iteration calculation results, the active power (section transmission power) P ij (t) between bus i and bus j in state t is calculated as shown in Equation (11):

[0111]

[0112] where Z ij represents the impedance value of the wire between bus i and bus j.

[0113] Furthermore, based on the load demand, generator output data, and the corresponding section transmission power, a training dataset is constructed.

[0114] As an optional implementation manner, the spatio-temporal feature extraction model includes a convolutional neural network and a long short-term memory network. Then, the spatio-temporal feature information is obtained through the following method:

[0115] Input the node embedding vector into the convolutional neural network to obtain intermediate features;

[0116] Input the intermediate features and the training dataset into the long short-term memory network to obtain spatio-temporal feature information.

[0117] Specifically, in the CNN, taking the node embedding vector as the input, using a one-dimensional convolutional window of size 1×7, and performing a convolutional operation on the node embedding vector in a downward sliding step manner to ensure that the feature matrix of the hidden layer contains the input features of other branches, and finally using the Linear layer for prediction.

[0118] In the LSTM, considering the connection on the historical data time series of the line nodes in the training dataset, using the first layer of LSTM to obtain the hidden features of the line nodes for prediction, and the second layer of LSTM is used to predict the branch parameters.

[0119] In the present invention, 80% of the training dataset is used as the training set, 10% as the validation set, and the last 10% as the test set. To reduce the storage overhead, the batch size of the data fed into the model is set to 4. The model will be trained for 200 generations in the Pytorch environment, using the most popular Adam algorithm as the optimization function, setting the initial learning rate to 0.002, and the learning rate decay is set to 0.9 every 10 generations.

[0120] As an optional implementation manner, the cross-section transmission power probability distribution is shown as the following formula:

[0121]

[0122] where, P ′ (y|x) represents the cross-section transmission power probability distribution; y represents the target vector; x represents the spatio-temporal feature information; π i (x) represents the mixing coefficient; N(y|μ i (x),σ i (x)) represents the conditional density of the target vector of the i-th kernel; g represents the total number of kernels of the mixture density network; μ i (x) represents the center of the i-th kernel; σ i (x) represents the common variance.

[0123] As an optional implementation manner, the expression of N(y|μ i (x),σ i (x)) is shown as the following formula:

[0124]

[0125] where, y represents the target vector; x represents the spatio-temporal feature information; π i(x) represents the mixing coefficient; N(y|μ i (x),σ i (x)) represents the conditional density of the target vector of the i-th kernel; μ i (x) represents the center of the i-th kernel; σ i (x) represents the common variance; c represents the dimension of the target vector.

[0126] It should be noted that the mixture density network is a neural network used to model complex probability distributions. It is mainly used to handle multi-modal (multiple peaks) or non-Gaussian distribution situations. Different from traditional neural networks with a single output node, the output of the mixture density network is a mixture distribution composed of multiple distributions, which makes the mixture density network perform well in dealing with tasks with uncertainty. The cross-section transmission power belongs to uncertain data, which means that there may be multiple different power flow situations at future moments. The mixture density network has the ability of multi-modal modeling and outputs a probability density function instead of a single point prediction, which can better capture multiple possibilities and adapt to situations with different probability distributions. Therefore, the embodiments of the present invention adopt the mixture density network for probabilistic power flow prediction.

[0127] The mixture density network represents the probability density of the target vector y as a linear combination of kernel functions. By parameterizing the output layer, the mixture normal distribution is used to describe the probability distribution. Considering a mixture Gaussian distribution with g components, the probability density function of each component can be expressed as shown in Equation (12):

[0128]

[0129] Among them, P ′ (y|x) represents the cross-section transmission power probability distribution; y represents the target vector (the target vector refers to the vector composed of the active power of different buses); x represents the spatio-temporal feature information; π i (x) represents the mixing coefficient, representing the weight of the i-th component (the "i-th component" is the "i-th kernel function"), and can be regarded as the prior probability of y conditional on x. And, π i (x) is a function of x and satisfies N(y|μ i (x),σ i (x)) represents the conditional density of the target vector of the i-th kernel; g represents the total number of kernels of the mixture density network; μ i (x) represents the center of the i-th kernel; σ i (x) represents the common variance.

[0130] Furthermore, N(y|μ i (x),σ i(x)) is the conditional density of the target vector y of the i-th kernel. There can be various choices for the kernel function. In the embodiments of the present invention, the normal distribution is selected, as shown in Equation (13):

[0131]

[0132] where y represents the target vector; x represents spatio-temporal feature information; π i (x) represents the mixing coefficient; N(y|μ i (x), σ i (x)) represents the conditional density of the target vector of the i-th kernel; g represents; μ i (x) represents the center of the i-th kernel; σ i (x) represents the common variance; c represents the dimension of the target vector.

[0133] Compared with the prior art, the cross-section transmission power probability prediction method provided by the embodiments of the present invention inputs the power grid topological structure, takes the buses in the power grid topological structure as nodes, takes the lines as edges, and converts the power grid topological structure into a power system graph; inputs the power system graph into a graph neural network to obtain node embedding vectors; generates initial values based on the power system graph, performs power system power flow calculations on the initial values to construct a training data set; wherein the initial values include load demand and unit output data, and the training data set includes the initial values and the corresponding cross-section transmission power; inputs the node embedding vectors into a spatio-temporal feature extraction model to obtain spatio-temporal feature information; inputs the spatio-temporal feature information and the training data set into a mixture density network to obtain the cross-section transmission power probability distribution. Compared with the prior art, the embodiments of the present invention can improve the accuracy and efficiency of power probability prediction.

[0134] See Figure 4 , the embodiments of the present invention also provide a cross-section transmission power probability prediction device 10, including:

[0135] A graph construction module 11, configured to input a power grid topological structure, take the buses in the power grid topological structure as nodes, take the lines as edges, and convert the power grid topological structure into a power system graph;

[0136] A graph feature extraction module 12, configured to input the power system graph into a graph neural network to obtain node embedding vectors;

[0137] A training data generation module 13, configured to generate initial values based on the power system graph, perform power system power flow calculations on the initial values to construct a training data set; wherein the initial values include load demand and unit output data, and the training data set includes the initial values and the corresponding cross-section transmission power;

[0138] The spatio-temporal feature information extraction module 14 is configured to input the node embedding vector into a spatio-temporal feature extraction model to obtain spatio-temporal feature information;

[0139] The power probability prediction module 15 is configured to input the spatio-temporal feature information and the training data set into a mixture density network to obtain a cross-section transmission power probability distribution.

[0140] The cross-section transmission power probability prediction device provided by the embodiment of the present invention can implement all the process steps of the cross-section transmission power probability prediction method described in the above embodiment. The functions and achieved technical effects of each module and unit in the device respectively correspond to the functions and achieved technical effects of the cross-section transmission power probability prediction method described in the above embodiment, and the specific implementation manners are not elaborated herein.

[0141] See Figure 5 , the embodiment of the present invention further provides a cross-section transmission power probability prediction device 20, including a processor 21, a memory 22, and a computer program stored in the memory 22 and configured to be executed by the processor 21. When the processor 21 executes the computer program, the steps in the embodiment of the above cross-section transmission power probability prediction method are implemented, such as Figure 1 the steps S1 to S6 described in

[0142] The cross-section transmission power probability prediction device may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The cross-section transmission power probability prediction device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the schematic diagram is only an example of the cross-section transmission power probability prediction device, and does not constitute a limitation on the cross-section transmission power probability prediction device. It may include more or fewer components than shown, or combine certain components, or different components. For example, the cross-section transmission power probability prediction device may further include an input / output device, a network access device, a bus, etc.

[0143] The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the cross-section transmission power probability prediction device, and connects various parts of the entire cross-section transmission power probability prediction device through various interfaces and lines.

[0144] The memory can be used to store the computer programs and / or modules. The processor realizes various functions of the cross-section transmission power probability prediction device by running or executing the computer programs and / or modules stored in the memory, and by calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function, etc.; the data storage area can store data created according to the use of the controller, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disks, memory, plug-in hard disks, Smart Media Cards (SMCs), Secure Digital (SD) cards, Flash Cards, at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage devices.

[0145] Among them, if the modules integrated in the cross-section transmission power probability prediction device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-mentioned embodiment methods of the present invention, it can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0146] Compared with the prior art, the cross-section transmission power probability prediction device, equipment and storage medium provided by the embodiments of the present invention input the power grid topology structure, use the buses in the power grid topology structure as nodes and the lines as edges, and convert the power grid topology structure into a power system graph; input the power system graph into a graph neural network to obtain node embedding vectors; generate initial values based on the power system graph and perform power system power flow calculations on the initial values to construct a training data set; among them, the initial values include load demand and unit output data, and the training data set includes the initial values and the corresponding cross-section transmission power; input the node embedding vectors into a spatio-temporal feature extraction model to obtain spatio-temporal feature information; input the spatio-temporal feature information and the training data set into a mixture density network to obtain the cross-section transmission power probability distribution. Compared with the prior art, the embodiments of the present invention can improve the accuracy and efficiency of power probability prediction.

[0147] The above is the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements are also regarded as the protection scope of the present invention.

Claims

1. A method for predicting cross-section transmission power probability, characterized in that: include: Inputting a power grid topology structure, taking buses in the power grid topology structure as nodes and lines as edges, converting the power grid topology structure into a power system diagram; Inputting the power system graph into a graph neural network to obtain a node embedding vector; Generate initial values ​​based on the power system diagram, and perform power system flow calculation on the initial values ​​to construct a training data set; wherein the initial values ​​include load demand and unit output data, and the training data set includes the initial values ​​and corresponding section transmission power; Inputting the node embedding vector into a spatiotemporal feature extraction model to obtain spatiotemporal feature information; The spatiotemporal feature information and the training data set are input into a mixed density network to obtain a cross-sectional transmission power probability distribution.

2. The method for predicting cross-section transmission power probability according to claim 1, characterized in that: The spatiotemporal feature extraction model includes a convolutional neural network and a long short-term memory network, and the spatiotemporal feature information is obtained by: Inputting the node embedding vector into the convolutional neural network to obtain intermediate features; The intermediate features and the training data set are input into the long short-term memory network to obtain spatiotemporal feature information.

3. The method for predicting cross-section transmission power probability according to claim 1, characterized in that: The cross-section transmission power probability distribution is shown as follows: Among them, P ′ (y|x) represents the probability distribution of cross-sectional transmission power; y represents the target vector; x represents the spatiotemporal characteristic information; π i (x) represents the mixing coefficient; N(y|μ i (x),σ i (x)) represents the conditional density of the target vector of the i-th kernel; g represents the total number of kernels in the mixed density network; μ i (x) represents the center of the i-th core; σ i (x) represents the common variance.

4. The method for predicting cross-section transmission power probability according to claim 3, characterized in that: N(y|μ i (x),σ i The expression of (x)) is as follows: Among them, y represents the target vector; x represents the spatiotemporal feature information; π i (x) represents the mixing coefficient; N(y|μ i (x),σ i (x)) represents the conditional density of the target vector of the i-th kernel; μ i (x) represents the center of the i-th core; σ i (x) represents the common variance; c represents the dimension of the target vector.

5. The method for predicting cross-section transmission power probability according to claim 1, characterized in that: The training dataset is constructed in the following way: Initialize the load demand power of each edge in the power system diagram and the unit output data of each bus node; wherein the unit output data includes the active power and reactive power injected by the generator into the bus node; Constructing a power balance equation based on the unit output data and the load demand power; Iteratively solving the power balance equation to obtain the cross-sectional transmission power in a steady state; The unit output data, the load demand power and the corresponding section transmission power are added as a group of data into a training data set.

6. The method for predicting cross-section transmission power probability according to claim 5, characterized in that: In each round of iteration, the voltage amplitude correction amount and the voltage phase angle correction amount of each bus node are calculated, and based on the voltage amplitude correction amount and the voltage phase angle correction amount, the voltage amplitude and the voltage phase angle of each bus node are corrected, and then a new round of iteration is entered; When the voltage amplitude correction amount and the voltage phase angle correction amount are both less than the preset error upper limit, the cross-sectional transmission power between the bus nodes is calculated based on the current voltage amplitude of each bus node and the current voltage phase angle difference between the bus nodes.

7. The method for predicting cross-section transmission power probability according to claim 1, characterized in that: The step of inputting the power system graph into a graph neural network to obtain a node embedding vector comprises: Inputting an initial characteristic matrix and an adjacency matrix of the power system diagram; Based on the adjacency matrix, nodes in the power system graph are aggregated to obtain hidden layer features; Perform linear regression prediction on the hidden layer features to obtain a node embedding vector.

8. A device for predicting cross-section transmission power probability, characterized in that: include: A graph construction module, for inputting a power grid topology structure, taking buses in the power grid topology structure as nodes and lines as edges, and converting the power grid topology structure into a power system graph; A graph feature extraction module, used for inputting the power system graph into a graph neural network to obtain a node embedding vector; A training data generation module, used to generate initial values ​​based on the power system diagram, and perform power system flow calculation on the initial values ​​to construct a training data set; wherein the initial values ​​include load demand and unit output data, and the training data set includes the initial values ​​and the corresponding cross-section transmission power; A spatiotemporal feature information extraction module, used for inputting the node embedding vector into a spatiotemporal feature extraction model to obtain spatiotemporal feature information; The power probability prediction module is used to input the spatiotemporal feature information and the training data set into a mixed density network to obtain a cross-sectional transmission power probability distribution.

9. A device for predicting cross-section transmission power probability, characterized in that: The invention comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for predicting the probability of cross-section transmission power according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the section transmission power probability prediction method according to any one of claims 1 to 7.

Images