Neural network embedding method, device and medium for power distribution network state estimation

Abstract

The application discloses a neural network embedding method and device for power distribution network state estimation, electronic equipment and medium, wherein the method comprises: acquiring an input sequence; embedding time information and node type information into a vector through space-time prior information embedding to obtain a space-time embedding vector; sampling a node of interest according to a power flow direction of optimal power flow and fusing node information to obtain a node embedding vector; using a graph isomorphism neural network to capture the local of a graph and embedding it into a feature vector to obtain a structure embedding vector; fusing the input sequence and the three vectors and inputting them into a graph space-time prediction network to output a prediction result. Through the introduction of space-time prior information, graph node embedding based on the optimal power flow direction and graph structure embedding of the graph isomorphism neural network, the application realizes the modeling of the characteristics of the power distribution network, makes up for the deficiency of the prior art in the specific modeling of the power distribution network and improves the accuracy of the power distribution network state estimation.

Description

TECHNICAL FIELD

[0001] The present application relates to power distribution network state estimation technology, and in particular to a neural network embedding method and device for power distribution network state estimation, an electronic device, and a medium. BACKGROUND

[0002] With the large-scale grid connection of new distributed power sources and electric vehicle charging stations, the power distribution network is facing highly fluctuating charging loads, which can lead to network instability and imbalance, thereby seriously affecting its safe operation. In this context, real-time monitoring and state estimation are crucial for the operation of the power distribution network. However, the spatiotemporal heterogeneity of the power distribution network poses a great challenge to real-time and accurate state estimation. In the spatial dimension, the node states and characteristics of different regions differ significantly; in the temporal dimension, the power load fluctuates greatly in different time periods and seasons. In the temporal dimension of state estimation, traditional methods usually assume that the network state changes slowly, and the measurement noise and time series distribution are ideal. Under this assumption, methods such as least squares and Kalman filtering are commonly used for state estimation. However, actual noise is random and can fluctuate dramatically in a short period of time, making it difficult for traditional methods to cope. In the spatial dimension of state estimation, traditional methods rely on redundant measurement data to correct and estimate the system state through multi-point measurement data. However, due to the lack of real-time synchronous phasor measurement devices in the power distribution network, the estimation of the overall network state becomes more difficult.

[0003] To solve these problems, data-driven neural network techniques model the power distribution network by utilizing historical data, enabling dynamic state estimation of the power distribution network. This method can make up for the lack of real-time data and improve the accuracy and reliability of state estimation. Among them, recurrent neural network (RNN) and its variants can capture the time characteristics of long time series, but ignore the spatial information between nodes. Graph neural network (GNN) and its variants can capture the relationship and graph structure information between nodes, and model complex topological structures, combined with RNN or temporal convolutional network (TCN), can achieve accurate prediction of spatio-temporal information. The general spatio-temporal graph neural network model can perform power distribution network state estimation, but there are still the following problems: (1) Lack of embedding of prior spatio-temporal knowledge. Ignoring time information: existing models often ignore the embedding of time periods and dates, which will appear in the power distribution network. Electricity fluctuation, and there will be different electricity distribution in different seasons. For example, there will be a peak of life electricity during the rest time, and the electricity in the southern region will rise in summer, etc. Ignoring spatial information: existing models generally consider each node as homogeneous, ignoring the differences between nodes, such as distributed power and load, which will have differences in nature. (2) Lack of modeling of the influence of physical properties between nodes. There are complex physical interactions between nodes in the power distribution network, such as power flow direction and other factors that are difficult to accurately capture and model by the model. (3) Lack of embedding of structure information. The structure information of the power distribution network should be fully considered in modeling, including the adjacency relationship between nodes and the overall topological structure, which can reflect the regional information and geographical distribution of the power distribution network, and also affect the overall state analysis. When modeling, a reasonable embedding algorithm needs to be designed to embed these information into the data features of the power distribution system. SUMMARY

[0004] To at least partially solve one of the technical problems existing in the prior art, the purpose of the present application is to provide a neural network embedding method, device, electronic equipment and medium for power distribution network state estimation.

[0005] The first technical solution adopted by the present application is:

[0006] A neural network embedding method for power distribution network state estimation, comprising the following steps:

[0007] Obtain an input sequence, the input sequence including node information (such as voltage, active power, reactive power, etc.), time information, node type, and connection information between nodes in the power distribution network;

[0008] By embedding the spatio-temporal prior information, time information and node type information are embedded into the vector to obtain a spatio-temporal embedding vector;

[0009] According to the power flow direction of the optimal power flow, the node of interest is sampled, and the node information is fused to obtain a node embedding vector;

[0010] The local graph is captured using a graph isomorphism neural network, which is embedded into a feature vector to obtain a structure embedding vector;

[0011] The input sequence, spatio-temporal embedding vector, node embedding vector and structure embedding vector are fused and input into a graph spatio-temporal prediction network to output a prediction result.

[0012] Further, the graph spatio-temporal prediction network can be implemented using an existing graph neural network model.

[0013] Further, the spatio-temporal prior information embedding includes three steps of time embedding, space embedding and spatio-temporal combination:

[0014] For the input sequence, a node data day time period t d , a week day t w , a year day t y and a node type c are obtained after preprocessing;

[0015] Time embedding: considering the time period embedding of data collection, week day embedding and year day embedding, three embedding matrices are constructed, and the corresponding embedding vector is obtained by taking the time as the index;

[0016] Time period embedding: assuming that N d times are sampled per day to obtain a time series, an embedding matrix is randomly initialized, where D is the embedding dimension;

[0017] Week day embedding: the number of days in a week is N w , an embedding matrix is randomly initialized;

[0018] Year day embedding: the number of days in a year is N y , an embedding matrix is randomly initialized;

[0019] Space embedding: considering the node type in the power distribution network, an embedding matrix is constructed, and the corresponding embedding vector is obtained by taking the node type and node number as the index; there are N c node types, and a category embedding matrix

[0020] Spatio-temporal combination: according to the information of each node, four embedding vectors with a dimension of D are obtained by querying the four embedding matrices:

[0021] e d = E d [t d ]

[0022] e w = E w [t w ]

[0023] e y = E y [t y ]

[0024] e c = E c [c]

[0025] The four vectors are connected and passed through a multilayer perceptron (MLP) to obtain a spatio-temporal embedding vector.

[0026] Further, the node types include: 1) load types: residential load nodes, commercial load nodes, electric vehicle charging stations, distributed energy storage nodes; 2) distributed power supply nodes: photovoltaic nodes, wind power nodes.

[0027] Further, the node of interest is sampled according to the power flow direction of the optimal power flow, and the node information is fused to obtain a node embedding vector, including:

[0028] Calculate the optimal power flow of the power distribution network; the optimal power flow refers to the optimal performance index of the power distribution network under the condition of meeting all operating constraints;

[0029] Construct a power distribution network graph: represent the power distribution network as a graph structure, where nodes represent nodes in the power distribution network, and edges are weighted edges, and the weight is the power change of the corresponding line;

[0030] Set the node walk algorithm: the hyperparameter is the number of node walks n w , the return parameter P, and the control parameter q; set the sampling probability, sample the nodes with active power transmission n w times according to the weight, to obtain a node list L;

[0031] Train and embed using a word vector embedding model: treat the node list L as a sequence, treat each node as a word, and treat the node sequence generated during the entire walk process as a sentence. Learn the node vector representation through the word vector embedding model; finally, obtain a high-dimensional vector representation of each node;

[0032] Construct a graph node embedding matrix E N ∈R N×Dwhere N is the total number of nodes of the graph, and D is the embedding dimension; the node embedding vector e is obtained in the embedding matrix with the node sequence number as the index N :

[0033] e N = E N [i].

[0034] Further, the optimal power flow of the power distribution network is calculated, comprising:

[0035] determining a target function:

[0036]

[0037] where f i (x) is the target function; x is a set of decision variables, including the voltage V, the voltage phase angle θ, the active power P and the reactive power Q of the node; N is the total number of nodes;

[0038] determining a constraint condition:

[0039]

[0040] where P Gi is the active power generation of the i-th node, P Di is the active power load of the i-th node; Q Gi is the reactive power generation of the i-th node, Q Di is the reactive power load of the i-th node; V i is the voltage of the i-th node, V j is the voltage of the j-th node; θ i is the voltage phase angle of the i-th node, θ j is the voltage phase angle of the j-th node; G ij is the real part of the line admittance between node i and node j, B ij is the imaginary part of the line admittance between node i and node j, is the minimum value of the active power generation of the i-th node, is the maximum value of the active power generation of the i-th node, is the minimum value of the reactive power generation of the i-th node, is the maximum value of the reactive power generation of the i-th node, is the minimum allowed value of the voltage amplitude of the i-th node, is the maximum allowed value of the voltage amplitude of the i-th node;

[0041] solving the parameters of the optimal power flow to obtain the power flowing in and out of each node under the optimal power flow condition, thereby obtaining the power flow direction.

[0042] Further, the setting sampling probability comprises:

[0043] Setting the probability of collecting the last sampling node V as When the node connection weight w xi The sampling probability is the weight w xi When the node connection weight w xi The sampling probability is

[0044] Further, the using graph isomorphism neural network to capture the local of the graph, embedding it into a feature vector to obtain a structure embedding vector, comprising:

[0045] Constructing a power distribution network graph, wherein the nodes of the graph are the nodes of the power distribution network;

[0046] Pretreating the power distribution network graph to generate a node feature vector;

[0047] Using a graph isomorphism neural network to perform graph embedding on the power distribution network graph to obtain a structure embedding vector.

[0048] Further, the working mode of the graph isomorphism neural network is as follows:

[0049] The graph isomorphism neural network comprises multiple GIN layers, and each GIN layer is responsible for aggregating neighborhood information of nodes;

[0050] In each GIN layer, the aggregated features of the nodes are calculated, and the specific formula is:

[0051]

[0052] In the formula, V represents the feature vector of node v at the kth layer, MLP represents a multilayer perceptron, and represents a learnable irrational number parameter;

[0053] The output feature vector of the multiple GIN layers is processed through a full connection layer and an activation function to obtain a final structure embedding vector.

[0054] The second technical solution adopted by the present application is:

[0055] An electronic device, comprising a processor and a memory, the memory storing at least one instruction, at least one program, a code set or an instruction set, the at least one instruction, the at least one program, the code set or the instruction set being loaded and executed by the processor to implement a neural network embedding method for power distribution network state estimation as described above.

[0056] The third technical solution adopted by the present application is:

[0057] A computer readable storage medium, the storage medium stores at least one instruction, at least one program, a code set or an instruction set, the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to realize the neural network embedding method for power distribution network state estimation as described above.

[0058] The fourth technical solution adopted by the present application is:

[0059] A computer program product or computer program, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer readable storage medium. The processor of the computer equipment can read the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer equipment executes the above method.

[0060] The beneficial effects of the present application are: the present application realizes the modeling of the characteristics of the power distribution network by introducing the spatio-temporal prior information, the graph node embedding based on the optimal power flow direction and the graph structure embedding of the graph isomorphism neural network, considers the special properties of the power distribution network, makes up for the deficiencies of the prior art in the special modeling of the power distribution network, and effectively improves the accuracy of the power distribution network state estimation. BRIEF DESCRIPTION OF DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following introduces the drawings of the related technical solutions in the embodiments of the present application or the prior art. It should be understood that the drawings in the following introduction are only for the convenience of clearly describing part of the embodiments in the technical solutions of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0062] Figure 1 is the whole flow chart of the power distribution network state estimation model in the embodiment of the present application;

[0063] Figure 2 is the whole flow chart of the spatio-temporal embedding in the embodiment of the present application;

[0064] Figure 3 is the improved node walk probability graph in the embodiment of the present application;

[0065] Figure 4 is the schematic diagram of training the sampled sequence by the word vector embedding model in the embodiment of the present application;

[0066] Figure 5 is the step flow chart of the neural network embedding method for power distribution network state estimation in the embodiment of the present application. DETAILED DESCRIPTION

[0067] Embodiments of the present application are described in detail below with reference to the attached drawing figures, wherein the same or like component have the same or similar designations throughout the attaching figures. The embodiments described below are exemplary in nature, only for the purpose of explanation and are not to be construed as limiting the present application. For the steps in the following embodiments, the numbering is only set for the convenience of explanation, and the order between the steps is not limited in any way, and the execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0068] In the description of the present application, it should be understood that the orientation description, such as the orientation or position relationship indicated by up, down, front, back, left, right, etc. is based on the orientation or position relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, and therefore cannot be understood as limiting the present application. The device or element indicated is not required to have a particular orientation, to be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0069] In the description of the present application, the meaning of several is one or more, and the meaning of multiple is more than two. Greater than, less than, more than, etc. are understood as not including the number, and above, below, etc. are understood as including the number. If it is described as first, second, it is only for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features or the order of indicated technical features.

[0070] In the description of the present application, unless otherwise explicitly limited, the words such as setting, installing, connecting, etc. should be broadly understood, and those skilled in the art can reasonably determine the specific meaning of the above words in the present application in combination with the specific content of the technical solution.

[0071] In order to incorporate more information into the graph neural network model of power distribution network state estimation, the present application proposes an information embedding scheme for power distribution network state estimation task. The scheme can effectively embed the spatio-temporal prior information, power flow direction and graph topology structure information. Specifically, the following three main methods are included: (1) Spatio-temporal prior information embedding: embedding time information (time of day, day cycle) and spatial information (such as node type) into the feature vector. (2) Graph node embedding based on optimal power flow power flow direction: using the optimal power flow direction to guide the node embedding process to ensure that the power transmission path and direction information is fully utilized. (3) Graph structure embedding based on graph isomorphism neural network: using graph isomorphism neural network (GIN) to capture the local graph and embed it into the feature vector.

[0072] The application fully considers the specific properties of the power distribution network, and realizes the modeling of the power distribution network characteristics by introducing the space-time prior information, the graph node embedding based on the optimal power flow direction, and the graph structure embedding of the graph isomorphism neural network. First, the space-time information and node type information specific to the power distribution network are added, so that the model can more accurately reflect the actual situation of the power distribution network. Second, it can adapt to existing graph neural network models without making major modifications to the original model architecture, and can directly use the new features for training and inference. In addition, the scheme optimizes the calculation cost, and through the fine design of the embedding strategy and the efficient graph isomorphism neural network algorithm, only a limited amount of calculation is added, which will not cause too much burden on the existing system's computing power requirements.

[0073] Embodiment 1

[0074] As shown in Figure 1 and Figure 5 , the embodiment provides a neural network embedding method for power distribution network state estimation, comprising the following steps:

[0075] S1, obtaining an input sequence, the input sequence comprising node information, time information, node type and connection information between nodes in the power distribution network;

[0076] S2, embedding the time information and the node type information into a vector through space-time prior information embedding to obtain a space-time embedding vector;

[0077] S3, sampling the nodes of interest according to the power flow direction of the optimal power flow and fusing the node information to obtain a node embedding vector;

[0078] S4, using a graph isomorphism neural network to capture the local of the graph and embed it into a feature vector to obtain a structure embedding vector;

[0079] S5, fusing the input sequence, the space-time embedding vector, the node embedding vector and the structure embedding vector, and inputting them into a graph space-time prediction network to output a prediction result.

[0080] Figure 1 The overall process of the embedding method is described. The input data (i.e. the input sequence) includes node information (such as voltage, active power, reactive power, etc.), time information, node type and connection information between nodes in the power distribution network. The space-time prior information embedding embeds the time and node type information into a vector of a specific dimension. The graph node embedding samples the nodes of interest according to the power flow direction of the optimal power flow and fuses the node information into a vector. The graph structure embedding uses a graph isomorphism network to extract graph structure information, ensuring that the subsequent model can distinguish isomorphic power distribution network graph structures. The three modules obtain the embedding vectors and add them to the input sequence as the input of the graph space-time prediction network.

[0081] The above method is explained in detail below in combination with the drawings and specific embodiments.

[0082] (1) Spatio-temporal prior information embedding

[0083] Figure 2 The overall flow of the spatio-temporal embedding method is shown, mainly including two parts of time embedding and space embedding.

[0084] For the input data, a node one data day time period t d , week day t w , year day t y and node type c are obtained through preprocessing.

[0085] Time embedding: considering the time period embedding of data collection, week day embedding, year day embedding to construct three embedding matrices, and taking the time as the index to obtain the corresponding embedding vector.

[0086] Time period embedding: assuming that N d times are sampled per day to obtain a time series, a randomly initialized embedding matrix is obtained, where D is the embedding dimension.

[0087] Week day embedding: there are seven days in a week, i.e. N w is 7, a randomly initialized embedding matrix

[0088] Year day embedding: there are 366 or 365 days in a year, set N y is 366 or 365, a randomly initialized embedding matrix

[0089] Space embedding: considering the node types in the power distribution network: (1) load type: residential load node, commercial load node, electric vehicle charging station, distributed energy storage node, etc., (2) distributed power node: photovoltaic node, wind power node, etc. An embedding matrix is constructed, and the corresponding embedding vector is obtained by taking the node type and node number as the index.

[0090] There are N c types of node types, a randomly initialized category embedding matrix

[0091] Spatio-temporal combination: according to the information of each node, the four embedding matrices above are queried to obtain four embedding vectors with dimension D

[0092] e d = E d [t d ] (1)

[0093] ew = E w [ t w ] (2)

[0094] e y = E y [ t y ] (3)

[0095] e c = E c [ c ] (4)

[0096] The four vectors are connected and passed through a Multilayer Perceptron (MLP) to obtain a spatio-temporal embedding vector:

[0097] e concat = [ e d ; e w ; e y ; e c ] ∈ R 4D (5)

[0098] e st = MLP( e concat ) ∈ R D (6)

[0099] (2) Graph node embedding based on optimal power flow power flow direction

[0100] The main algorithm improves the walk strategy of the graph node embedding algorithm, so that the embedded nodes can reflect the range of their influence and being influenced. The main steps of the algorithm include:

[0101] Step 1. Calculate the optimal power flow of the distribution network: The optimal power flow refers to the optimal performance of the distribution network under the condition of meeting all operating constraints.

[0102] Objective function (minimize certain cost or loss):

[0103]

[0104] In the formula, f i (x) is the objective function; x is the set of decision variables, including the voltage V, voltage phase angle θ, active power P and reactive power Q of the node; N is the total number of nodes; c i is the weight of the i-th node.

[0105] The constraints include the following:

[0106]

[0107] The constraints include power balance constraints, generator output constraints, node voltage constraints, etc.

[0108] Solve the parameters of optimal power flow using Newton method, interior point method or intelligent optimization algorithm, get the power inflow and outflow of each node under the condition of optimal power flow, and get the power flow direction.

[0109] Step 2. Construct the power distribution network graph: represent the power distribution network as a graph structure, where the node represents the node in the power distribution network, and the edge is the weighted edge, and the weight is the power change of the corresponding line.

[0110] Step 3. Set the node walk algorithm: the hyperparameter is the number of node walks n w , return parameter P, control parameter q. Set the sampling probability, such as Figure 3 , V node is the last sampled node, X is the current sampling node, set the probability of collecting the last sampling node V When the node connection weight w xi is positive, the sampling probability is the weight w xi , when the node connection weight w xi is negative, the sampling probability is

[0111]

[0112] According to the weight, sample the node with active power transmission n w times, get a node list L.

[0113] Step 4. Training and embedding using word vector embedding model: take the node list L as a sequence, take each node as a word, and take the node sequence generated in the whole walk process as a sentence. Through the Skip-gram method of word vector embedding model, learn the node vector representation. The specific process is as follows Figure 4 : take each node and the nodes in its context window as training samples. The size of the context window can be adjusted according to the specific application. Through the neural network model, train the node vector so that the adjacent nodes are closer in the vector space, thereby reflecting the structural and functional relationship between nodes. Finally, get the high-dimensional vector representation of each node.

[0114] Step 5. Construct graph node embedding matrix E N ∈R N×D , where N is the total number of nodes of the graph, and D is the embedding dimension. Take the graph node serial number as the index to get the embedding vector e N in the embedding matrix:

[0115] e N = E N [i] (14)

[0116] Parameter explanation: adjust the number of node walks nw The range of the walk can be controlled, thereby controlling the receptive field of the node. Adjusting the return parameter p can control the tendency to sample the node just visited, and the walk parameter q can control the tendency to sample in the direction of power flow or in the direction of power outflow.

[0117] (3) Graph structure embedding based on graph isomorphism neural network

[0118] The purpose of graph structure embedding is to add the structural features of the power distribution network into the features. The graph isomorphism neural network has the function of distinguishing isomorphic graphs, so it is used by us to extract the structural features of the power distribution network. The graph structure embedding is realized through the following steps:

[0119] Step 1. Construct a power distribution network graph, where the nodes are the nodes of the power distribution network.

[0120] Step 2. Preprocess the power distribution network graph to generate node feature vectors.

[0121] Step 3. Use the graph isomorphism neural network to perform graph embedding on the power distribution network graph.

[0122] As an optional implementation, step 3 includes the following steps:

[0123] 1) Define the graph isomorphism neural network model structure, including multiple GIN layers, each GIN layer being responsible for aggregating the neighborhood information of the node.

[0124] 2) In each GIN layer, the aggregated features of the node are calculated. The specific formula is:

[0125]

[0126] where represents the feature vector of node v at the kth layer, represents the set of neighbor nodes of node v, and MLP represents a multi-layer perceptron.

[0127] 3) The output feature vector of the multi-layer GIN layer is processed through a fully connected layer and an activation function to obtain the final structure embedding vector.

[0128] In summary, the existing power distribution network state estimation method based on graph neural network mainly relies on the powerful feature extraction ability of neural network and large-scale high-quality power distribution network data, and has the characteristics of high accuracy compared with traditional state estimation methods. However, these models all use general graph structure for modeling, ignoring the unique properties of the power distribution network, and lack of specific modeling of the power distribution network. The embedding algorithm used by the existing model is only a general graph space-time embedding, which cannot effectively reflect the characteristics of the power distribution network. The application designs a method of embedding the properties of the power distribution network into the power distribution network graph, which includes the embedding of space-time prior information, graph node embedding based on optimal power flow direction and graph structure embedding based on graph isomorphism neural network, providing additional data feature support for the existing model. Specifically, a time coding module is designed to embed week and date information, and combine with a space coding module to embed node type information, and learn these category information through an embedding layer; a graph node embedding algorithm based on optimal power flow direction is used to walk strategy, to embed each node and make full use of the information of power transmission direction in the power distribution network; a graph isomorphism network is used to embed the local graph structure, better capturing the structure information of the power distribution network. Through these improvements, the application scheme can improve the accuracy of power distribution network state estimation and make up for the shortcomings of the prior art in specific modeling of the power distribution network.

[0129] Embodiment 2

[0130] The embodiment of the application also provides an electronic device, which comprises a processor and a memory, and the memory stores at least one instruction, at least one program, a code set or an instruction set, which is loaded and executed by the processor to implement the neural network embedding method for power distribution network state estimation as shown in Figure 1 or Figure 5 .

[0131] It can be understood that the memory can include a random access memory (RAM) and a read-only memory (ROM). Optionally, the memory includes a non-transitory computer-readable storage medium. The memory can be used to store instructions, programs, codes, code sets or instruction sets. The memory can include a program storage area and a data storage area, wherein the program storage area can store instructions for implementing an operating system, instructions for at least one function, instructions for implementing each of the above methods, etc.; and the data storage area can store data created according to the use of the server, etc.

[0132] The processor can include one or more processing cores. The processor connects various parts within the entire server by various interfaces and lines, executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory, and calling data stored in the memory. Optionally, the processor can be implemented in at least one of a hardware form of a digital signal processing (DSP), a field-programmable gate array (FPGA), and a programmable logic array (PLA). The processor can be integrated with one or several combinations of a central processing unit (CPU) and a modem. Among them, the CPU mainly processes operating systems and application programs; the modem is used to process wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor, but be implemented by a separate chip.

[0133] Since the electronic device is an electronic device corresponding to the neural network embedding method for power distribution network state estimation according to the embodiments of the present application, and the principle of solving problems of the electronic device is similar to that of the method, the implementation of the electronic device can be referred to the implementation process of the above-mentioned method embodiments, and the repeated parts will not be repeated.

[0134] Embodiment 3

[0135] The embodiments of the present application also provide a computer readable storage medium, wherein at least one instruction, at least one program, a code set or an instruction set are stored in the storage medium, and the at least one instruction, the at least one program, the code set or the instruction set are loaded and executed by a processor to implement a neural network embedding method for power distribution network state estimation as shown in Figure 1 or Figure 5 .

[0136] Those skilled in the art can understand that all or part of the steps of various methods in the above embodiments can be instructed by programs to relevant hardware, and the programs can be stored in a computer readable storage medium, including Read-Only Memory (ROM), Random Access Memory (RAM), Programmable Read-only Memory (PROM), Erasable Programmable Read Only Memory (EPROM), One-time Programmable Read-Only Memory (OTPROM), Electrically-Erasable Programmable Read-Only Memory (EEPROM), Compact Disc Read-Only Memory (CD-ROM) or other optical disk storage, magnetic disk storage, magnetic tape storage, or any other medium that can be used to carry or store data which can be read by a computer.

[0137] Since the storage medium is a storage medium corresponding to the neural network embedding method for power distribution network state estimation of the embodiments of the present application, and the principle of solving problems of the storage medium is similar to the method, the implementation of the storage medium can refer to the implementation process of the above method embodiments, and the repeated parts will not be described again.

[0138] Embodiment 4

[0139] In some possible implementation manners, various aspects of the method of the embodiments of the present application can also be implemented in the form of a program product, which includes program codes for causing a computer device to execute the steps of the neural network embedding method for power distribution network state estimation according to various exemplary embodiments of the present application described above in the specification when the program product is run on the computer device. Wherein, the executable computer program codes or "codes" for executing various embodiments can be written in high-level programming languages such as C, C++, C#, Smalltalk, Java, JavaScript, Visual Basic, Structured Query Language (for example, Transact-SQL), Perl, or in various other programming languages.

[0140] It should be understood that various aspects of the application can be implemented in hardware, software, firmware or a combination of them. In the above embodiments, various steps or methods can be implemented in software or firmware which is stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, and in another embodiment, any of the following technologies, known in the art, can be used to implement the hardware: discrete logic circuitry having logic gates for implementing logic functions upon data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and so forth, or a combination of them.

[0141] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Also, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, a person skilled in the art can combine and combine the different embodiments or examples described in the specification and the features of the different embodiments or examples without contradiction.

[0142] The above embodiments are only for the purpose of illustrating the technical concept and characteristics of the present application, and the purpose is to enable those skilled in the art to understand the content of the present application and to implement it, and cannot limit the protection scope of the present application. Any equivalent changes or modifications made according to the essence of the present application should be covered within the protection scope of the present application.

Claims

1. A neural network embedding method for power distribution network state estimation, characterized in that, The method comprises the following steps: obtaining an input sequence comprising node information, time information, node type and connection information between nodes in a power distribution network; embedding the time information and the node type information into a vector through space-time prior information embedding to obtain a space-time embedding vector; sampling a node of interest according to a power flow direction of optimal power flow and fusing node information to obtain a node embedding vector; using a graph isomorphism neural network to capture the local of the graph and embed it into a feature vector to obtain a structure embedding vector; fusing the input sequence, the space-time embedding vector, the node embedding vector and the structure embedding vector and inputting them into a graph space-time prediction network to output a prediction result; the space-time prior information embedding comprises three steps of time embedding, space embedding and space-time combination: For the input sequence, a node one piece of data of the day time period is obtained through preprocessing , the middle of the week date , the middle of the year date , and the node type ; time embedding: considering time period embedding, week day embedding and year day embedding to construct three embedding matrices, and obtaining the corresponding embedding vector by taking the time as an index; Time period embedding: sample every day Obtain time series, randomly initialize an embedding matrix where is the embedding dimension; Weekday embedding: the day of the week is , a random initialization of an embedding matrix ; Mid-year date embedding: the number of days in each year is , a random initialization of an embedding matrix ; Space embedding: considering the node types in the power distribution network, an embedding matrix is constructed, and the corresponding embedding vector is obtained with the node type and node number as the index; a total of One kind of node type, a category embedding matrix is randomly initialized ; Space-time combination: According to the information of each node, four embedding matrices are queried to obtain four embedding vectors with dimensions of ​ connecting the four vectors and passing them through a multi-layer perception to obtain a space-time embedding vector.

2. The neural network embedding method for power distribution network state estimation according to claim 1, wherein, the node type comprises: 1) load type: residential load node, commercial load node, electric vehicle charging station and distributed energy storage node; 2) distributed power supply node: photovoltaic node and wind power node.

3. The neural network embedding method for power distribution network state estimation according to claim 1, wherein, the sampling of the node of interest according to the power flow direction of the optimal power flow and the fusion of the node information to obtain the node embedding vector comprise: calculating the optimal power flow of the power distribution network; constructing a power distribution network graph: representing the power distribution network as a graph structure, wherein the nodes represent the nodes in the power distribution network and the edges are weighted edges, and the weight is the power change of the corresponding line; Set the node walk algorithm: hyperparameters are the number of node walks , return parameters , control parameters ; Set the sampling probability to sample the nodes with power transmission according to the weight times, get a node list L; training and embedding using a word vector embedding model: regarding the node list L as a sequence, regarding each node as a word, regarding the node sequence generated in the whole walk process as a sentence, and learning the node vector representation through the word vector embedding model; finally, obtaining a high-dimensional vector representation of each node; Constructing graph node embedding matrix from high-dimensional vector representation wherein is the total number of nodes of the graph, is the embedding dimension; the node embedding vector is obtained in the embedding matrix with the graph node serial number as the index : 。 4. The neural network embedding method for power distribution network state estimation according to claim 3, wherein, the calculation of the optimal power flow of the power distribution network comprises: determining the objective function: wherein is the objective function; is a set of decision variables including the voltage , the voltage phase angle , the active power and the reactive power at the nodes; is the total number of nodes; determining the constraint condition: In the formula, For the first Active power generation of each node For the first The active power load of each node; For the first Reactive power generation of each node For the first The reactive power load of each node; For the first The voltage of each node, For the first The voltage of each node; For the first The voltage phase angle of each node, For the first Voltage phase angle at each node; For nodes With nodes The real part of the line admittance between them For nodes With nodes The imaginary part of the line admittance between them For the first The minimum active power generation of each node. For the first The maximum active power generation of each node. For the first The minimum reactive power generation of each node. For the first The maximum reactive power generation of each node. For the first The minimum allowable value of voltage amplitude at each node. For the first The maximum allowable voltage amplitude of each node; solving the parameters of the optimal power flow to obtain the power inflow and outflow of each node under the optimal power flow condition, thereby obtaining the power flow direction.

5. The neural network embedding method for power distribution network state estimation according to claim 3, wherein, the setting of the sampling probability comprises: Let V be the last sampled node, X be the current sampled node, set the probability of collecting the last sampled node V as When the node connection weight is positive, the sampling probability is the weight When the node connection weight is negative, the sampling probability is The specific formula is as follows: 。 6. The neural network embedding method for power distribution network state estimation according to claim 1, wherein, the use of the graph isomorphism neural network to capture the local of the graph and embed it into a feature vector to obtain a structure embedding vector comprises: constructing a power distribution network graph, wherein the nodes of the graph are the nodes of the power distribution network; preprocessing the power distribution network graph to generate a node feature vector; using a graph isomorphism neural network to perform graph embedding on the power distribution network graph to obtain a structure embedding vector.

7. The neural network embedding method for power distribution network state estimation according to claim 6, wherein, the working mode of the graph isomorphism neural network is as follows: the graph isomorphism neural network comprises multiple GIN layers, and each GIN layer is responsible for aggregating the neighborhood information of the nodes; in each GIN layer, the aggregated features of the nodes are calculated, and the specific formula is: wherein a node In a first characteristic vector of a layer, a node a set of neighbor nodes of a node, MLP denotes a multi-layer perceptron, denotes a learnable irrational number parameter; processing the output feature vector of the multiple GIN layers through a full connection layer and an activation function to obtain the final structure embedding vector.

8. An electronic device, comprising: The electronic device comprises a processor and a memory, and the memory stores at least one program, which is loaded and executed by the processor to implement the method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The storage medium stores at least one program, which is loaded and executed by the processor to implement the method according to any one of claims 1 to 7.

Images