Voltage transformer error prediction method and system based on spatio-temporal feature fusion

Abstract

The invention belongs to the technical field of electric power measurement, and provides a voltage transformer error prediction method and system based on spatial-temporal feature fusion, and the method comprises the steps: obtaining the state data of a voltage transformer; respectively extracting spatial features and time features of the obtained state data based on the graph structure to obtain space-time fusion features of the state data; the method comprises the following steps: extracting time sequence features of state data space-time fusion features by adopting a convolutional neural network-long short-term memory network, measuring the importance of the extracted time sequence features based on an attention mechanism to obtain a hidden layer state vector weight, and predicting a voltage transformer error according to the obtained time sequence feature weight and the predicted voltage transformer error. And completing error prediction of the voltage transformer based on spatio-temporal feature fusion.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric power metering, and in particular relates to a voltage transformer error prediction method and system based on spatiotemporal feature fusion. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Voltage transformer is an electric measuring device for measuring system voltage, which is indispensable in substation and even the whole power system. However, the field operation environment of voltage transformer is complex, which is easily affected by factors such as surrounding magnetic field and temperature, and has poor long-term stability. Measuring errors will inevitably occur, which will bring certain potential threats to the economical, safe and stable operation of the power grid. Therefore, error prediction of voltage transformer is particularly important for protecting equipment and ensuring that the transformer maintains a high measurement accuracy.

[0004] In the process of predicting voltage transformer metering errors, the existing technology extracts the theoretical additional errors of the two CVT error influencing factors, temperature and frequency, from the error sequence, predicts the stripped error sequence, and adds the additional errors caused by the current temperature and frequency, thereby achieving accurate prediction of the error; however, it ignores the hidden spatial dependency between the error and each different variable in the time series data. Summary of the invention

[0005] To solve the above problems, the present invention proposes a voltage transformer error prediction method and system based on spatiotemporal feature fusion, which uses a graph convolutional neural network to extract the spatial characteristics of the voltage transformer state data, constructs the spatiotemporal fusion characteristics of the state data, and combines the convolutional neural network-long short-term memory network to extract the temporal characteristics of the spatiotemporal fusion spatial characteristics of the state data. The error prediction of the voltage transformer is completed based on the temporal characteristics and attention mechanism, which improves the prediction accuracy and shortens the prediction time.

[0006] According to some embodiments, a first solution of the present invention provides a voltage transformer error prediction method based on spatiotemporal feature fusion, which adopts the following technical solution: A voltage transformer error prediction method based on spatiotemporal feature fusion, comprising: Get the status data of the voltage transformer; Based on the graph structure, the spatial features and temporal features of the acquired state data are extracted respectively to obtain the spatiotemporal fusion features of the state data; A convolutional neural network-long short-term memory network is used to extract the time series features of the spatiotemporal fusion features of the state data. The importance of the extracted time series features is measured based on the attention mechanism to obtain the weight of the hidden layer state vector. According to the obtained time series feature weights and the predicted voltage transformer error, the error prediction of the voltage transformer based on the fusion of spatiotemporal features is completed.

[0007] As a further technical limitation, in the process of predicting the voltage transformer error based on the extracted timing features and attention mechanism, the hidden layer state matrix of the long short-term memory network is activated through a fully connected network to obtain a hidden layer state vector, and the importance of the obtained hidden layer state vector is measured through the attention mechanism to obtain the hidden layer state vector weight; based on the obtained hidden layer state vector weight and the timing characteristics of the spatiotemporal fusion features of the state data, the voltage transformer error prediction value is obtained.

[0008] As a further technical limitation, in the process of extracting the spatial features of the acquired state data based on the graph structure, a graph structure space is constructed with different state data as nodes, the similarity coefficient between the node and its adjacent nodes is calculated to determine the adjacency relationship between the nodes, the node state is updated according to the adjacency relationship between the nodes, the node features are optimized, and the optimized node features are spliced ​​in time series to obtain the state data spatial sequence, thereby completing the extraction of the state data spatial features.

[0009] As a further technical limitation, in the process of extracting the time features of the acquired state data based on the graph structure, the acquired time series sequence of the state data is divided into several segments, and a graph structure is constructed for each time series sequence from a time perspective to obtain a state data time series that incorporates the time features, thereby completing the extraction of the time features of the state data.

[0010] As a further technical limitation, in the process of extracting the timing features of the spatiotemporal fusion features of the state data, the obtained spatiotemporal fusion features of the state data are convolved through a convolutional neural network, and the convolution output features of the state data are obtained based on the convolution processing of average pooling and maximum pooling. The obtained convolution output features of the state data are input into the long short-term memory network to extract the timing features, so as to obtain the timing features of the spatiotemporal fusion features of the state data.

[0011] As a further technical limitation, the acquired status data of the voltage transformer includes at least historical errors, temperature, humidity, frequency and secondary load of the voltage transformer; the spatiotemporal fusion characteristics of the status data are the weighted sum of the time characteristics of the status data and the spatial characteristics of the status data.

[0012] According to some embodiments, a second solution of the present invention provides a voltage transformer error prediction system based on spatiotemporal feature fusion, which adopts the following technical solution: A voltage transformer error prediction system based on spatiotemporal feature fusion, comprising: An acquisition module, configured to acquire status data of a voltage transformer; A fusion module, which is configured to extract spatial features and temporal features of the acquired state data based on the graph structure, respectively, to obtain spatiotemporal fusion features of the state data; The prediction module is configured to use a convolutional neural network-long short-term memory network to extract the time series features of the spatiotemporal fusion features of the state data, measure the importance of the extracted time series features based on the attention mechanism, obtain the hidden layer state vector weight, and complete the error prediction of the voltage transformer based on the spatiotemporal feature fusion according to the obtained time series feature weight and the predicted voltage transformer error.

[0013] According to some embodiments, a third solution of the present invention provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium stores a program thereon, which, when executed by a processor, implements the steps in a voltage transformer error prediction method based on spatiotemporal feature fusion as described in the first solution of the present invention.

[0014] According to some embodiments, a fourth solution of the present invention provides an electronic device, which adopts the following technical solution: An electronic device comprises a memory, a processor and a program stored in the memory and running on the processor, wherein when the processor executes the program, the steps in a voltage transformer error prediction method based on spatiotemporal feature fusion as described in the first scheme of the present invention are implemented.

[0015] According to some embodiments, a fifth solution of the present invention provides a computer program product, which adopts the following technical solution: A computer program product includes software code, wherein the program in the software code executes the steps in a voltage transformer error prediction method based on spatiotemporal feature fusion as described in the first embodiment of the present invention.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention adopts a graph convolutional neural network to extract the spatial characteristics of the voltage transformer state data, constructs the spatiotemporal fusion characteristics of the state data, and combines the convolutional neural network-long short-term memory network to extract the temporal characteristics of the spatiotemporal fusion spatial characteristics of the state data. Based on the temporal characteristics and the attention mechanism, the error prediction of the voltage transformer is completed, which improves the prediction accuracy and shortens the prediction time. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings in the specification that constitute a part of this embodiment are used to provide a further understanding of this embodiment. The schematic embodiments of this embodiment and their descriptions are used to explain this embodiment and do not constitute improper limitations on this embodiment.

[0018] Figure 1 This is a flow chart of a voltage transformer error prediction method based on spatiotemporal feature fusion in Embodiment 1 of the present invention; Figure 2 Schematic diagram of the structure of the error prediction model in the first embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the improved LSTM network in the first embodiment of the present invention; Figure 4 This is a schematic diagram of the result of voltage transformer error prediction in the first embodiment of the present invention; Figure 5 It is a structural block diagram of a voltage transformer error prediction system based on spatiotemporal feature fusion in the second embodiment of the present invention. DETAILED DESCRIPTION

[0019] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0020] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0021] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0022] In the present invention, terms such as "upper", "lower", "left", "right", "front", "back", "vertical", "horizontal", "side", "bottom", etc. indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings. They are relational words determined only for the convenience of describing the structural relationships of the various parts or elements of the present invention, and do not specifically refer to any part or element in the present invention and should not be understood as limitations on the present invention.

[0023] In the present invention, terms such as "fixed connection", "connected", "connection", etc. should be understood in a broad sense, indicating that it can be fixedly connected, integrally connected or detachably connected; it can be directly connected or indirectly connected through an intermediate medium. For relevant scientific research or technical personnel in this field, the specific meanings of the above terms in the present invention can be determined according to specific circumstances, and they cannot be understood as limitations on the present invention.

[0024] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0025] Embodiment 1 Embodiment 1 of the present invention introduces a voltage transformer error prediction method based on spatiotemporal feature fusion.

[0026] This embodiment introduces a voltage transformer error prediction method based on spatiotemporal feature fusion, which uses an improved long short-term memory network (LSTM) to extract time series features and a graph convolutional neural network to extract spatial features. By fusing spatiotemporal features, accurate prediction of transformer errors is achieved.

[0027] like Figure 1 A voltage transformer error prediction method based on spatiotemporal feature fusion is shown, comprising the following steps: Step S100: collecting data such as historical error, historical temperature, humidity, frequency, and secondary load of the voltage transformer; Step S200: Using a graph convolutional neural network to extract spatial and temporal features of the error and its influencing factors to obtain spatiotemporal fusion features; Step S300: using an improved convolutional neural network-long short-term memory network (CNN-LSTM) to extract time series features from the spatiotemporal fusion features and complete the training of the error model; Step S400: predicting the mutual inductor error based on the trained error model.

[0028] In the process of voltage transformer data collection in this embodiment, the collected historical error data of the voltage transformer is recorded as , where T represents the sequence length; considering that the mutual inductor error is affected by factors such as temperature, humidity, grid frequency, and secondary load, the influencing factors are collected to obtain the temperature ,humidity ,frequency , Secondary load ;Build a time series dataset , where J represents the number of variables, here J=5. , where t=1,2,…,T.

[0029] In order to mine error features, this embodiment uses different graph structure feature extraction methods to capture variable error features from spatial and temporal perspectives, and obtains effective error features through feature fusion; first, considering that the error size is affected by factors such as temperature, humidity, frequency, and secondary load, in order to mine the hidden spatial dependency between the error and the influencing variables, a graph structure of the spatial structure is constructed; at the same time, considering the forward and backward dependency of the error in time series, a graph structure of the time structure is constructed to mine the dependency between the previous and next time series points.

[0030] The graph structure of the spatial structure divides the time series of error, temperature, frequency, and secondary load into M segments, and constructs a graph structure for each segment according to the spatial perspective, that is, constructs it with different variables as nodes. This embodiment takes the first segment as an example to explain in detail: ①Constructing graph structure space by ,in , represents a node, here represents each variable; is the set of edges connecting adjacent nodes; The characteristics of are defined as , , , , , which represent the node characteristics of error, temperature, humidity, frequency, and secondary load respectively.

[0031] ②Calculate its adjacency relationship Compute Node Its adjacent nodes The similarity coefficient between ,Right now (1); Among them, || represents the vector concatenation operation, , is the weight parameter; , The current node , neighbor nodes The feature representation of is the activation function; .

[0032] ③Based on the adjacency relationship between nodes, update the node status and optimize the node characteristics Update each feature using formula (2) by updating the adjacent nodes Aggregation is used to propagate node information and update the current node status.

[0033] (2); in, Reflects the importance of adjacent node information to the current node. is the number of nodes in the graph, is a nonlinear activation function. are learnable weight parameters.

[0034] ④ Based on the updated node status, obtain the spatial features of the first segment of the error data .

[0035] The spatial features of each error segment are spliced ​​in time sequence to obtain the error spatial sequence after integrating the spatial features. : (3); in, ; represents the spatial characteristics of the first segment of the error data, is the spatial feature of the Mth segment of error data.

[0036] (2) Graph structure of time structure The error, temperature, frequency, and secondary load are divided into M segments according to the time series, and a graph structure is constructed for each segment from a time perspective. Here, the first segment is taken as an example for explanation.

[0037] ①Constructing graph structure space by ,in , represents a node, where each node is a sample composed of multiple variables; is the set of edges connecting adjacent nodes; The characteristics of , .

[0038] ② Calculate the adjacency matrix (4); (5); in, For sample and The weight of the edge between them; For sample and The Euclidean distance between is the standard deviation of the distance, is a constant term, is the threshold used to ensure the sparsity of the feature matrix,

[0039] ③Based on the adjacency relationship between nodes, update the node status and optimize the node characteristics.

[0040] Update node features using formula (6) (6); in, for The degree matrix of , is the activation function, , To set the weight.

[0041] Considering that each node has different influence on the current node during the node update process, the closer the node is to the current node, the greater the influence on the error characteristics of the current node. Based on this idea, the weight W is designed here to give increasing weight values ​​to the changes in the timing from far to near the current node, so as to consider the change of the error from the perspective of time.

[0042] (7); Based on the updated node status, the error features are extracted and merged to obtain the time features of the first segment of the error data. .

[0043] ④ Splice the time features of each error segment according to the time series to obtain the error time series sequence after integrating the time series features : (8); in, ; Represents the timing characteristics of the first segment of error data, It is the timing characteristics of the Mth segment of error data.

[0044] In this embodiment, the spatiotemporal fusion feature of the state data is the weighted sum of the state data time feature and the state data space feature; the spatiotemporal fusion feature of the state data is (9); in, Set weights for each.

[0045] This embodiment uses a convolutional neural network-long short-term memory network to extract the time series features of the spatiotemporal fusion features of the state data, and measures the importance of the extracted time series features based on the attention mechanism to obtain the hidden layer state vector weight. According to the obtained time series feature weight and the predicted voltage transformer error, the error prediction of the voltage transformer based on the spatiotemporal feature fusion is completed; specifically: Using CNN to analyze the error space features Convolution is performed. In order to extract more features, average pooling and maximum pooling are used. To keep the length of the time dimension unchanged, a padding strategy is used during the convolution process.

[0046] (10); in, is the output of the CNN network, represents the Hadamard product, is the Sigmoid activation function; Indicates that the convolution kernel size is Two-dimensional convolution of Express and To splice; like Figure 2 As shown, the convolutional features Input LSTM to extract time series features and complete error prediction. In order to reduce the time of prediction training, the LSTM model is optimized and only one gate is used to control the recursive self-connection of input and cell state. Input the time series data of YX as follows Figure 3 In the improved LSTM shown, the temporal features of the data are extracted; that is, (11); Use the attention mechanism to complete error prediction; the LSTM hidden layer state matrix Through the fully connected network, and use Activate and get the new hidden layer state vector C; then use the attention mechanism to measure the importance of the hidden layer state vector and get the weight D; and get the weighted sum of D and H , and then combined with the The error prediction value can be obtained by summing : (12); in, for Activation function; are the weight matrices, are biased respectively.

[0047] During the training process, formula (13) is used as the loss function: (13); Where N is the number of prediction samples, is the predicted value of the nth sample, is the true error value of the nth sample.

[0048] This embodiment predicts the voltage transformer error, and compares the error directly predicted by LSTM, the estimated error obtained by the prediction in this embodiment, and the actual error result. Figure 4As shown, it can be seen that the error prediction model of this embodiment can realize the effective prediction of the mutual inductor error value.

[0049] This embodiment uses a graph convolutional neural network to extract the spatial characteristics of the voltage transformer state data, constructs the spatiotemporal fusion characteristics of the state data, and combines the convolutional neural network-long short-term memory network to extract the temporal characteristics of the spatiotemporal fusion spatial characteristics of the state data. Based on the temporal characteristics and the attention mechanism, the error prediction of the voltage transformer is completed, which improves the prediction accuracy and shortens the prediction time.

[0050] Embodiment 2 Embodiment 2 of the present invention introduces a voltage transformer error prediction system based on spatiotemporal feature fusion.

[0051] like Figure 5 A voltage transformer error prediction system based on spatiotemporal feature fusion is shown, comprising: An acquisition module, configured to acquire status data of a voltage transformer; A fusion module, which is configured to extract spatial features and temporal features of the acquired state data based on the graph structure, respectively, to obtain spatiotemporal fusion features of the state data; The prediction module is configured to use a convolutional neural network-long short-term memory network to extract the time series features of the spatiotemporal fusion features of the state data, measure the importance of the extracted time series features based on the attention mechanism, obtain the hidden layer state vector weight, and complete the error prediction of the voltage transformer based on the spatiotemporal feature fusion according to the obtained time series feature weight and the predicted voltage transformer error.

[0052] The detailed steps are the same as those of the voltage transformer error prediction method based on spatiotemporal feature fusion provided in Example 1, and will not be repeated here.

[0053] Embodiment 3 Embodiment 3 of the present invention provides a computer-readable storage medium.

[0054] A computer-readable storage medium stores a program thereon, which, when executed by a processor, implements the steps in a voltage transformer error prediction method based on spatiotemporal feature fusion as described in Embodiment 1 of the present invention.

[0055] The detailed steps are the same as those of the voltage transformer error prediction method based on spatiotemporal feature fusion provided in Example 1, and will not be repeated here.

[0056] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The solutions in the embodiments of the present invention may be implemented in various computer languages, for example, object-oriented programming language Java and interpreted scripting language JavaScript, etc.

[0057] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0058] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0059] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0060] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0061] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

[0062] Embodiment 4 A fourth embodiment of the present invention provides an electronic device.

[0063] An electronic device comprises a memory, a processor and a program stored in the memory and running on the processor, wherein when the processor executes the program, the steps in a voltage transformer error prediction method based on spatiotemporal feature fusion as described in Embodiment 1 of the present invention are implemented.

[0064] The detailed steps are the same as those of the voltage transformer error prediction method based on spatiotemporal feature fusion provided in Example 1, and will not be repeated here.

[0065] Embodiment 5 Embodiment 5 of the present invention provides a computer program product.

[0066] A computer program product includes software code, wherein the program in the software code executes the steps in a voltage transformer error prediction method based on spatiotemporal feature fusion as described in Embodiment 1 of the present invention.

[0067] The detailed steps are the same as those of the voltage transformer error prediction method based on spatiotemporal feature fusion provided in Example 1, and will not be repeated here.

[0068] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The solutions in the embodiments of the present invention may be implemented in various computer languages, for example, object-oriented programming language Java and interpreted scripting language JavaScript, etc.

[0069] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0070] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0071] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0072] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0073] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

[0074] The above description is only a preferred embodiment of the present embodiment and is not intended to limit the present embodiment. For those skilled in the art, the present embodiment may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present embodiment shall be included in the protection scope of the present embodiment.

Claims

1. A voltage transformer error prediction method based on spatiotemporal feature fusion, characterized in that: include: Get the status data of the voltage transformer; Based on the graph structure, the spatial features and temporal features of the acquired state data are extracted respectively to obtain the spatiotemporal fusion features of the state data; A convolutional neural network-long short-term memory network is used to extract the time series features of the spatiotemporal fusion features of the state data. The importance of the extracted time series features is measured based on the attention mechanism to obtain the weight of the hidden layer state vector. According to the obtained time series feature weights and the predicted voltage transformer error, the error prediction of the voltage transformer based on the fusion of spatiotemporal features is completed.

2. A voltage transformer error prediction method based on spatiotemporal feature fusion as described in claim 1, characterized in that: In the process of predicting the voltage transformer error based on the extracted timing features and attention mechanism, the hidden layer state matrix of the long short-term memory network is activated through a fully connected network to obtain the hidden layer state vector. The importance of the obtained hidden layer state vector is measured by the attention mechanism to obtain the hidden layer state vector weight. According to the obtained hidden layer state vector weight and the timing characteristics of the spatiotemporal fusion features of the state data, the predicted value of the voltage transformer error is obtained.

3. A voltage transformer error prediction method based on spatiotemporal feature fusion as described in claim 1, characterized in that: In the process of extracting the spatial features of the acquired state data based on the graph structure, a graph structure space is constructed with different state data as nodes, the similarity coefficient between the node and its adjacent nodes is calculated to determine the adjacency relationship between the nodes, the node state is updated according to the adjacency relationship between the nodes, the node features are optimized, and the optimized node features are spliced ​​in time series to obtain the state data space sequence, thereby completing the extraction of the state data space features.

4. A voltage transformer error prediction method based on spatiotemporal feature fusion as claimed in claim 1, characterized in that: In the process of extracting the time features of the acquired state data based on the graph structure, the time series of the acquired state data is divided into several segments, and a graph structure is constructed for each time series segment from a time perspective to obtain a state data time series that incorporates the time features, thereby completing the extraction of the time features of the state data.

5. A voltage transformer error prediction method based on spatiotemporal feature fusion as claimed in claim 1, characterized in that: In the process of extracting the timing features of the spatiotemporal fusion features of the state data, the obtained spatiotemporal fusion features of the state data are convolved through a convolutional neural network, and the convolution output features of the state data are obtained based on the convolution processing of average pooling and maximum pooling. The obtained convolution output features of the state data are input into the long short-term memory network to extract the timing features, so as to obtain the timing features of the spatiotemporal fusion features of the state data.

6. A voltage transformer error prediction method based on spatiotemporal feature fusion as claimed in claim 1, characterized in that: The acquired state data of the voltage transformer includes at least historical errors, temperature, humidity, frequency and secondary load of the voltage transformer; the spatiotemporal fusion characteristics of the state data are the weighted sum of the time characteristics of the state data and the spatial characteristics of the state data.

7. A voltage transformer error prediction system based on spatiotemporal feature fusion, characterized in that: include: An acquisition module, configured to acquire status data of a voltage transformer; A fusion module, which is configured to extract spatial features and temporal features of the acquired state data based on the graph structure, respectively, to obtain spatiotemporal fusion features of the state data; The prediction module is configured to use a convolutional neural network-long short-term memory network to extract the time series features of the spatiotemporal fusion features of the state data, measure the importance of the extracted time series features based on the attention mechanism, obtain the hidden layer state vector weight, and complete the error prediction of the voltage transformer based on the spatiotemporal feature fusion according to the obtained time series feature weight and the predicted voltage transformer error.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of a voltage transformer error prediction method based on spatiotemporal feature fusion as described in any one of claims 1 to 6 are implemented.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the program, the steps of a voltage transformer error prediction method based on spatiotemporal feature fusion as described in any one of claims 1-6 are implemented.

10. A computer program product comprising software code, characterized in that The program in the software code executes the steps of a voltage transformer error prediction method based on spatiotemporal feature fusion as described in any one of claims 1-6.

Images