CN-121980456-A - State estimation network construction method and device of power system and running state estimation method

CN121980456ACN 121980456 ACN121980456 ACN 121980456ACN-121980456-A

Abstract

The application discloses a method and a device for constructing a state estimation network of an electric power system and an operation state estimation method, which belong to the field of operation and maintenance of the electric power system, wherein the method comprises the steps of constructing feature vectors of buses according to historical measurement data of buses and bus-related operation equipment of a target electric power system; constructing a plurality of original power diagram data according to the topological structure data of a target power system and the historical measurement data, then performing feature learning on the plurality of original power diagram data by utilizing a generating countermeasure network to generate a plurality of enhanced power diagram data, performing iterative training on a diagram convolution network according to the enhanced power diagram data, and taking the diagram convolution network after training as a state estimation network. By implementing the method and the device, the problem of low efficiency of power system running state evaluation in the prior art can be solved.

Inventors

LI SHIMING
ZHAO RUIFENG
GUO WENXIN
LI HAOBIN
DAI YUE
ZHONG WEI

Assignees

广东电网有限责任公司电力调度控制中心

Dates

Publication Date: 20260505
Application Date: 20260122

Claims (10)

1. A method for constructing a state estimation network of an electric power system, comprising: The method comprises the steps of obtaining topological structure data of a target power system and historical measurement data of each bus and each bus-associated operation device in the target power system, wherein the topological structure data comprises connection relations between each power transmission line and corresponding buses and connection relations between each operation device and corresponding buses; constructing feature vectors of all buses according to the historical measurement data of all buses and all bus-related operation equipment; Constructing a plurality of original power diagram data according to the topological structure data and the historical measurement data, wherein each original power diagram data comprises a plurality of nodes, each node corresponds to one bus in a target power system and is associated with a feature vector of the corresponding bus; Performing feature learning on the plurality of original electric power map data by using a generating countermeasure network to generate a plurality of enhanced electric power map data; Performing iterative training on the graph rolling network according to the enhanced electric diagram data to obtain a trained graph rolling network; and taking the trained graph convolution network as a state estimation network of a target power system.
2. The method for constructing a state estimation network for an electric power system according to claim 1, further comprising, before acquiring the historical measurement data of each bus and each bus-related device in the target electric power system: acquiring original historical measurement data of each busbar and each busbar-associated device in a target power system; Performing missing value complementation on the original historical measurement data to obtain the complemented historical measurement data; and taking the completed historical measurement data as the historical measurement data of each busbar and each busbar-associated device in the target power system.
3. The method for constructing a state estimation network of a power system according to claim 1, wherein the constructing feature vectors of each bus according to the historical measurement data of each bus and each bus-related operation device comprises: and extracting the characteristics of the historical measurement data of each bus and each bus-associated operation device to obtain the characteristic vector of each bus.
4. The method for constructing a state estimation network of a power system according to claim 1, wherein the constructing a plurality of original power map data from the topology data and the history measurement data includes: Taking each busbar in the target power system as a node, and associating the characteristic vector of each busbar to a corresponding node to obtain a node set containing the characteristic vector; According to the connection relation between each power transmission line and the corresponding bus and the connection relation between each operation device and the corresponding bus, taking each power transmission line and each operation device in the target power system as edges for connecting each node; and constructing a plurality of original power diagram data according to the node set and the edges connecting the nodes.
5. The method of constructing a state estimation network for a power system according to claim 1, wherein the feature learning of the plurality of original electrical map data by using the generation countermeasure network, generating a plurality of enhanced electrical map data, comprises: generating a plurality of first random noises; Inputting the first random noise and the plurality of original power map data into a generation countermeasure network to obtain a plurality of enhanced power map data output by the generation countermeasure network; Wherein the training process for generating the countermeasure network comprises: Acquiring electric force diagram sample data; constructing a generated countermeasure network architecture, wherein the generated countermeasure network architecture comprises a generator and a discriminator; generating a plurality of second random noises; Inputting the electric diagram sample data and the second random noise into the generated countermeasure network architecture, so that the generated countermeasure network architecture alternately trains the generator and the discriminator according to the input data until a loss function of the generated countermeasure network architecture reaches a preset convergence condition, and a trained generated countermeasure network architecture is obtained; taking the trained generated countermeasure network architecture as the generated countermeasure network; In each training process, the generator generates simulated power pattern data with similar distribution to the power pattern data according to the power pattern sample data and the random noise, the simulated power pattern data is input to the discriminator, and the discriminator performs discrimination analysis according to the simulated power pattern data and the corresponding power pattern data and returns the discrimination analysis result to the generator.
6. The method for constructing a state estimation network of a power system according to claim 1, wherein the iterative training of the graph rolling network according to the enhanced electric power map data to obtain a trained graph rolling network comprises: The method comprises the steps of obtaining a tag set, wherein the tag set comprises a plurality of tag groups, each tag group comprises a voltage amplitude tag and a phase angle value tag, and different tag groups correspond to different buses in a target power system; And taking the enhanced power diagram data as input data of the graph convolution network, taking the tag set as a supervision target of the graph convolution network, taking the voltage amplitude value and the phase angle value of each bus predicted by the graph convolution network as output data, and performing iterative training on the graph convolution network until a preset training round is reached, so as to obtain the trained graph convolution network.
7. The method for constructing a state estimation network of a power system according to claim 6, wherein the iterative training includes: Calculating the loss function value between the voltage amplitude value and the phase angle value of each bus predicted by the current graph convolution network and the corresponding voltage amplitude value label and phase angle value label; Calculating the gradient of the model parameters of the current graph convolution network by using a back propagation algorithm according to the loss function value, and updating the network parameters of the current graph convolution network by using an optimizer to obtain a graph convolution network with updated parameters; Judging whether the current training round reaches the preset training round, If so, terminating the iterative training, taking the graph rolling network with updated parameters as a trained graph rolling network, If not, updating the current training round, and taking the graph rolling network with updated parameters as the graph rolling network for the next training.
8. A state estimation network construction apparatus of an electric power system, comprising: The system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring topological structure data of a target power system, each bus in the target power system and historical measurement data of each bus-associated operation device, wherein the topological structure data comprises a connection relation between each power transmission line and a corresponding bus and a connection relation between each operation device and the corresponding bus; The characteristic vector construction module is used for constructing characteristic vectors of all buses according to the buses and the historical measurement data of the operation equipment associated with the buses; The first power diagram data construction module is used for constructing a plurality of original power diagram data according to the topological structure data and the history measurement data, wherein each original power diagram data comprises a plurality of nodes, each node corresponds to one bus in a target power system and is associated with a feature vector of the corresponding bus; The second power diagram data construction module is used for performing feature learning on the plurality of original power diagram data by utilizing a generating countermeasure network to generate a plurality of enhanced power diagram data; the training module is used for carrying out iterative training on the graph rolling network according to the enhanced electric diagram data to obtain a trained graph rolling network; and the state estimation network generation module is used for taking the trained graph rolling network as a state estimation network of a target power system.
9. A method of evaluating an operation state of an electric power system, comprising: acquiring topological structure data of a target power system and real-time measurement data of each busbar and each busbar associated operation device in the target power system; extracting features of the real-time measurement data to obtain real-time feature vectors of all buses; constructing target power diagram data according to the topological structure data and the real-time feature vector; Inputting the target power diagram data into a state estimation network to obtain a predicted voltage amplitude value and a predicted phase angle value of each bus output by the state estimation network, wherein the state estimation network is determined by the state estimation network construction method of the power system according to any one of claims 1-7; And according to the predicted voltage amplitude and the predicted phase angle value of each bus, evaluating the running state of each bus in the target power system to obtain an evaluation result.
10. The method for evaluating the operation state of the power system according to claim 9, wherein the evaluating the state of the target power system based on the predicted voltage amplitude and the predicted phase angle value of each bus to obtain the evaluation result includes: For each bus, judging whether the predicted voltage amplitude of the current bus is positioned in a preset bus voltage amplitude safety interval, judging whether the predicted phase angle value of the current bus is positioned in a bus phase angle difference stability threshold value, If the judging results are all yes, judging that the current bus state is normal operation; otherwise, judging the current bus state as dangerous operation.

Description

State estimation network construction method and device of power system and running state estimation method Technical Field The present invention relates to the field of operation and maintenance of power systems, and in particular, to a method and apparatus for constructing a state estimation network of a power system, and a method for evaluating an operation state. Background The safe and stable operation of the power system is a precondition for guaranteeing the social production and living order, and at present, the state estimation means of the power system is mainly developed and judged by combining manual experience analysis with offline data, and a large amount of scattered SCADA measurement data is required to be arranged and analyzed one by means of manual work, so that the steady state operation state and potential risk points of the power grid are judged. However, the state estimation method has poor adaptability to the power grid topology change, the manual data processing efficiency is low, the rapid analysis requirement of mass measurement data is difficult to deal with, the risk evolution trend in the power grid operation process is difficult to effectively and rapidly capture, the randomness and the fluctuation of the new energy output further aggravate the complexity of the power grid operation state under the complex power grid scene of high-proportion new energy access, the hysteresis and the limitation of the manual analysis method are further amplified, the linkage fault is extremely easy to be caused due to the fact that the state estimation is not timely and inaccurate, and the power grid safety is seriously threatened. Disclosure of Invention The invention provides a method and a device for constructing a state estimation network of a power system and a method for estimating the running state, which can solve the problem of low efficiency of estimating the running state of the power system in the prior art. In order to solve the above technical problems, an embodiment of the present invention provides a method for constructing a state estimation network of an electric power system, including: The method comprises the steps of obtaining topological structure data of a target power system and historical measurement data of each bus and each bus-associated operation device in the target power system, wherein the topological structure data comprises connection relations between each power transmission line and corresponding buses and connection relations between each operation device and corresponding buses; constructing feature vectors of all buses according to the historical measurement data of all buses and all bus-related operation equipment; Constructing a plurality of original power diagram data according to the topological structure data and the historical measurement data, wherein each original power diagram data comprises a plurality of nodes, each node corresponds to one bus in a target power system and is associated with a feature vector of the corresponding bus; Performing feature learning on the plurality of original electric power map data by using a generating countermeasure network to generate a plurality of enhanced electric power map data; Performing iterative training on the graph rolling network according to the enhanced electric diagram data to obtain a trained graph rolling network; and taking the trained graph convolution network as a state estimation network of a target power system. Further, before acquiring the historical measurement data of each busbar and each busbar-associated device in the target power system, the method further comprises: acquiring original historical measurement data of each busbar and each busbar-associated device in a target power system; Performing missing value complementation on the original historical measurement data to obtain the complemented historical measurement data; and taking the completed historical measurement data as the historical measurement data of each busbar and each busbar-associated device in the target power system. Further, the constructing the feature vector of each busbar according to the historical measurement data of each busbar and each busbar associated operation device includes: and extracting the characteristics of the historical measurement data of each bus and each bus-associated operation device to obtain the characteristic vector of each bus. Further, the constructing a plurality of original power map data according to the topology structure data and the historical measurement data includes: Taking each busbar in the target power system as a node, and associating the characteristic vector of each busbar to a corresponding node to obtain a node set containing the characteristic vector; According to the connection relation between each power transmission line and the corresponding bus and the connection relation between each operation device and the corresponding bus, taking each power transmission line