Lightweight power distribution network pseudo measurement generation method and device based on generative adversarial network

Abstract

This application relates to a lightweight distribution network pseudo-measurement generation method and apparatus based on generative adversarial networks (GANs). The method includes: training a lightweight distribution network pseudo-measurement generation model based on a GAN using training and testing sets to generate a distribution network pseudo-measurement generation model that satisfies preset optimal conditions; inputting data from the testing set into the distribution network pseudo-measurement generation model that satisfies the preset optimal conditions to obtain high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network; and performing inverse normalization processing on the high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network to generate inverse normalized high-frequency pseudo-measurement values ​​for each branch and node of the medium-voltage distribution network. This solves the problems of related technologies typically relying on complex model structures, large parameter counts, low computational efficiency, difficulty in meeting the needs of real-time monitoring of distribution networks, and the lack of accuracy in the generated pseudo-measurement data, failing to accurately reflect the system's operating status.

Description

Technical Field

[0001] This application relates to the field of secondary planning technology for power systems, and in particular to a lightweight method and apparatus for generating pseudo-measurements for distribution networks based on generative adversarial networks. Background Technology

[0002] With the rapid development of smart grids and the energy internet, the distribution network, as a core component, relies heavily on data integrity and accuracy for system state estimation and dispatch decisions. However, the complex network structure and weak data transmission security of the distribution network make measurement data prone to loss or anomalies, severely impacting the reliability of grid operation. Traditional pseudo-measurement generation methods, limited by model complexity and adaptability, struggle to accurately reflect the spatiotemporal characteristics of the data, necessitating innovative technological breakthroughs.

[0003] Among related technologies, it is possible to acquire distribution network topology data and combine it with the power sample set of power grid nodes to perform distribution network power flow calculation, thereby obtaining a trained deep residual network corresponding to each distribution network node without real-time measurement, which improves the accuracy of pseudo-measurement data generation.

[0004] However, related technologies typically rely on complex model structures with a large number of parameters and low computational efficiency, making it difficult to meet the needs of real-time monitoring of distribution networks. Furthermore, they have limited performance in extracting the spatiotemporal characteristics of distribution network data. In particular, when dealing with highly nonlinear and dynamically changing data, the generated pseudo-measurement data lacks accuracy and cannot accurately reflect the system's operating status, which urgently needs improvement. Summary of the Invention

[0005] This application provides a lightweight pseudo-measurement generation method and device for distribution networks based on generative adversarial networks, which solves the problems that related technologies usually rely on complex model structures, have a large number of parameters, low computational efficiency, and are difficult to meet the needs of real-time monitoring of distribution networks. They also have limited performance in extracting the spatiotemporal characteristics of distribution network data. In particular, when dealing with highly nonlinear and dynamically changing data, the generated pseudo-measurement data lacks accuracy and cannot accurately reflect the system operating status.

[0006] The first aspect of this application provides a lightweight pseudo-measurement generation method for distribution networks based on generative adversarial networks (GANs), comprising the following steps: collecting at least one measurement data of voltage, current, power factor, active power, and reactive power from a medium-voltage distribution network; generating normalized historical measurement data based on the at least one measurement data; dividing the normalized historical measurement data into a training set and a test set according to a preset ratio; constructing a generative adversarial network (GAN) model using a generator network and a discriminator network, and performing pruning operations on the GAN model to generate pseudo-measurements that meet preset computational requirements. A generative adversarial network (GAN) is used; a lightweight distribution network pseudo-measurement generation model based on the GAN is trained using the training set and the test set to generate a distribution network pseudo-measurement generation model that meets preset optimal conditions; data from the test set is input into the distribution network pseudo-measurement generation model that meets preset optimal conditions to obtain high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network, and the high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network are denormalized to generate denormalized high-frequency pseudo-measurement values ​​of each branch and node of the medium-voltage distribution network.

[0007] Optionally, in one embodiment of this application, generating normalized measurement data based on the at least one measurement data includes: removing outliers and missing values ​​from the at least one measurement data to obtain new measurement data; filling in missing values ​​in the new measurement data to generate measurement data that meets a preset completeness condition; performing time alignment on the measurement data that meets the preset completeness condition to determine historical measurement data, and normalizing the historical measurement data to generate the normalized historical measurement data.

[0008] Optionally, in one embodiment of this application, the calculation formula for the normalized historical measurement data is:

[0009]

[0010] Where, x ′ k Let x be the k-th normalized historical measurement. k For the k-th original historical measurement, x mink Let x be the minimum value among k sets of data. maxk It represents the maximum value of k sets of data.

[0011] Optionally, in one embodiment of this application, training a lightweight distribution network pseudo-measurement generation model based on the generative adversarial network using the training set and the test set to generate a distribution network pseudo-measurement generation model that meets preset optimal conditions includes: acquiring at least one input data sample from meteorological data, distribution network load nodes, and branch nodes; training the generator and discriminator alternately based on the at least one input data sample to generate training results; monitoring performance indicators on the validation set based on the training results, and stopping the training of the lightweight distribution network pseudo-measurement generation model when the performance indicators meet preset conditions, so as to generate the distribution network pseudo-measurement generation model that meets preset optimal conditions.

[0012] Optionally, in one embodiment of this application, after generating the actual high-frequency pseudo-measurement values ​​of each branch and node of the medium-voltage distribution network, the method further includes: inputting a random noise vector into the generator to generate pseudo-measurement data; evaluating the pseudo-measurement data to generate an evaluation result, and based on the evaluation result, integrating the pseudo-measurement data into the monitoring system of the medium-voltage distribution network to generate integrated pseudo-measurement data; and updating the generative adversarial network based on the integrated pseudo-measurement data and the high-frequency pseudo-measurement values ​​of the nodes to determine the updated generative adversarial network.

[0013] Optionally, in one embodiment of this application, the formula for calculating the denormalized pseudo-measurement is:

[0014] y m =y n *(y max -y min )+y min ,

[0015] Among them, y n y represents the pseudo-measurement generated by the model. max y min These are the maximum and minimum values ​​of the high-frequency pseudo-quantity measurements of the node, respectively.

[0016] A second aspect of this application provides a lightweight distribution network pseudo-measurement generation device based on a generative adversarial network (GAN), comprising: a data acquisition module for acquiring at least one of the following measurement data from a medium-voltage distribution network: voltage, current, power factor, active power, and reactive power; a data generation module for generating normalized historical measurement data based on the at least one measurement data; a partitioning module for dividing the normalized historical measurement data into a training set and a test set according to a preset ratio; and a construction module for constructing a generative adversarial network (GAN) model using a generator network and a discriminator network, and performing pruning operations on the GAN model to generate pseudo-measurements that satisfy... A generative adversarial network (GAN) with preset computational requirements is included. A training module is used to train a lightweight distribution network pseudo-measurement generation model based on the GAN using the training set and the test set, to generate a distribution network pseudo-measurement generation model that meets preset optimal conditions. A generation module is used to input data from the test set into the distribution network pseudo-measurement generation model that meets preset optimal conditions to obtain high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network, and to perform inverse normalization processing on the high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network to generate inverse normalized high-frequency pseudo-measurement values ​​of each branch and node of the medium-voltage distribution network.

[0017] Optionally, in one embodiment of this application, the data generation module includes: a removal unit, used to remove outliers and missing values ​​from the at least one measurement data to obtain new measurement data; a filling unit, used to fill in missing values ​​in the new measurement data to generate measurement data that meets a preset completeness condition; and a data generation unit, used to time-align the measurement data that meets the preset completeness condition to determine historical measurement data, and normalize the historical measurement data to generate the normalized historical measurement data.

[0018] Optionally, in one embodiment of this application, the calculation formula for the normalized historical measurement data is:

[0019]

[0020] Where, x ′ k Let x be the k-th normalized historical measurement. k For the k-th original historical measurement, x mink Let x be the minimum value among k sets of data. maxk It represents the maximum value of k sets of data.

[0021] Optionally, in one embodiment of this application, the training module includes: an acquisition unit, configured to acquire at least one input data sample from meteorological data, distribution network load nodes, and branch nodes; a training unit, configured to alternately train the generator and discriminator based on the at least one input data sample to generate training results; and a model generation unit, configured to monitor performance indicators on the validation set based on the training results, and stop training the lightweight distribution network pseudo-measurement generation model when the performance indicators meet preset conditions, so as to generate the distribution network pseudo-measurement generation model that meets the preset optimal conditions.

[0022] Optionally, in one embodiment of this application, it further includes: an input module, configured to input a random noise vector into the generator after generating the actual high-frequency pseudo-measurement values ​​of each branch and node of the medium-voltage distribution network to generate pseudo-measurement data; an evaluation module, configured to evaluate the pseudo-measurement data to generate an evaluation result, and based on the evaluation result, integrate the pseudo-measurement data into the monitoring system of the medium-voltage distribution network to generate integrated pseudo-measurement data; and an update module, configured to update the generative adversarial network based on the integrated pseudo-measurement data and the high-frequency pseudo-measurement values ​​of the nodes to determine the updated generative adversarial network.

[0023] Optionally, in one embodiment of this application, the formula for calculating the denormalized pseudo-measurement is:

[0024] y m =y n *(y max -y min )+y min ,

[0025] Among them, y n y represents the pseudo-measurement generated by the model. max y min These are the maximum and minimum values ​​of the high-frequency pseudo-quantity measurements of the node, respectively.

[0026] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the lightweight distribution network pseudo-measurement generation method based on generative adversarial networks as described in the above embodiments.

[0027] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described lightweight distribution network pseudo-measurement generation method based on generative adversarial networks.

[0028] A fifth aspect of this application provides a computer program product that stores a computer program that, when executed by a processor, implements the above-described lightweight distribution network pseudo-measurement generation method based on generative adversarial networks.

[0029] This application's embodiments, through the design of a lightweight GAN architecture, simplify the network structure and the number of parameters, reducing computational complexity and making it suitable for resource-constrained real-time application scenarios. Through adversarial training between the generator and discriminator, the generator is responsible for extracting the spatiotemporal features of distribution network measurement data and generating realistic pseudo-measurement data, while the discriminator optimizes the generator's performance to improve the quality and accuracy of the pseudo-measurements. This lightweight algorithm significantly reduces computational overhead while ensuring high accuracy and robustness, providing strong technical support for real-time state estimation and efficient operation of smart grids. This solves the problems of related technologies typically relying on complex model structures, large parameter counts, and low computational efficiency, making it difficult to meet the needs of real-time monitoring of distribution networks. Furthermore, these technologies have limited performance in extracting the spatiotemporal features of distribution network data, especially when dealing with highly nonlinear and dynamically changing data. The generated pseudo-measurement data lacks accuracy and cannot accurately reflect the system's operating state. Additionally, inaccurate state estimation and scheduling decision failures are caused by missing or abnormal measurement data in the distribution network.

[0030] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0031] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0032] Figure 1 This is a flowchart of a lightweight distribution network pseudo-measurement generation method based on generative adversarial networks provided in an embodiment of this application;

[0033] Figure 2 This is a schematic diagram of the EnerGAN model structure according to an embodiment of this application;

[0034] Figure 3 This is a flowchart illustrating a method for generating pseudo-measurements in a power distribution network according to an embodiment of this application.

[0035] Figure 4 This is a schematic diagram of a lightweight distribution network pseudo-measurement generation device based on generative adversarial networks provided in an embodiment of this application;

[0036] Figure 5 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0037] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0038] The following description, with reference to the accompanying drawings, describes a lightweight distribution network pseudo-measurement generation method and apparatus based on generative adversarial networks (GANs) according to embodiments of this application. Addressing the issues raised in the background section regarding related technologies that typically rely on complex model structures, have a large number of parameters, low computational efficiency, and struggle to meet the needs of real-time distribution network monitoring, this application provides a lightweight distribution network pseudo-measurement generation method based on generative adversarial networks (GANs). This method simplifies the network structure and number of parameters by designing a lightweight GAN architecture, reducing computational complexity and making it suitable for resource-constrained real-time applications. Through adversarial training between the generator and discriminator, the generator extracts the spatiotemporal features of distribution network measurement data and generates realistic pseudo-measurement data, while the discriminator optimizes the generator's performance to improve the quality and accuracy of the pseudo-measurements. This lightweight algorithm significantly reduces computational overhead while maintaining high accuracy and robustness, providing strong technical support for real-time state estimation and efficient operation of smart grids. This solves the problems that related technologies usually rely on complex model structures, have a large number of parameters, low computational efficiency, and are difficult to meet the needs of real-time monitoring of distribution networks. They also have limited performance in extracting the spatiotemporal characteristics of distribution network data. In particular, when dealing with highly nonlinear and dynamically changing data, the generated pseudo-measurement data lacks accuracy and cannot accurately reflect the system's operating status.

[0039] Specifically, Figure 1 This is a flowchart illustrating a lightweight distribution network pseudo-measurement generation method based on generative adversarial networks provided in an embodiment of this application.

[0040] like Figure 1 As shown, the lightweight distribution network pseudo-measurement generation method based on generative adversarial networks includes the following steps:

[0041] In step S101, at least one of the following measurement data of medium-voltage distribution network is collected: voltage, current, power factor, active power and reactive power.

[0042] In actual implementation, the embodiments of this application can utilize data sources such as meteorology, SCAD systems (Supervisory Control And Data Acquisition), and smart meters to collect real measurement data of key indicators such as voltage, current, power factor, active power, and reactive power of medium-voltage distribution networks, thereby significantly improving the quality and efficiency of power grid operation and providing support for subsequent real-time state estimation and efficient operation of smart grids.

[0043] In step S102, normalized historical measurement data is generated based on at least one measurement data point; the normalized historical measurement data is then divided into a training set and a test set according to a preset ratio.

[0044] It is understood that the preset ratio in the embodiments of this application can be 8:2.

[0045] In this embodiment, the above-mentioned measurement data can be processed to generate normalized historical measurement data, thereby accelerating the training and convergence speed of the model. The normalized historical measurement data is divided into training and testing sets in an 8:2 ratio, thereby providing support for the subsequent training of a lightweight distribution network pseudo-measurement generation model, thus optimizing the model parameters and improving the reliability of the model.

[0046] Optionally, in one embodiment of this application, generating normalized measurement data based on at least one measurement data includes: removing outliers and missing values ​​from at least one measurement data to obtain new measurement data; filling in missing values ​​in the new measurement data to generate measurement data that meets a preset completeness condition; aligning the measurement data that meets the preset completeness condition with time to determine historical measurement data, and normalizing the historical measurement data to generate normalized historical measurement data.

[0047] In practical implementation, this application embodiment can remove outliers and missing values ​​caused by equipment failure, communication interruption, etc., to obtain new measurement data. The KNN (K-Nearest Neighbor) method is used to fill in missing values, ensuring data integrity and generating complete measurement data. Time alignment is then performed to determine historical measurement data. Based on this, the historical measurement data is normalized to the range of [0,1] or [-1,1] to generate normalized historical measurement data, thereby accelerating the training and convergence speed of the model.

[0048] In one embodiment of this application, the formula for calculating the normalized historical measurement data is as follows:

[0049]

[0050] Where, x′ k Let x be the k-th normalized historical measurement. k For the k-th original historical measurement, x mink Let x be the minimum value among k sets of data. maxk It represents the maximum value of k sets of data.

[0051] In step S103, a generative adversarial network (GAN) model is constructed using a generator network and a discriminator network, and a pruning operation is performed on the GAN model to generate a GAN that meets the preset computational requirements.

[0052] It is understood that the generator network in this embodiment adopts a deep convolutional neural network (DCNN) structure to generate a pseudo-measurement data discriminator network based on the input random noise vector z; the discriminator network also adopts a convolutional neural network or a fully connected neural network structure to distinguish between real measurement data and pseudo-measurement data generated by the generator. The network model design is as follows: Figure 2 As shown, the generator part includes an input layer, three one-dimensional convolutional layers, two one-dimensional deconvolutional layers, and an output layer; the discriminator part includes an input total power sequence, two one-dimensional convolutional layers, a fully connected layer, and an output layer.

[0053] Specifically, this embodiment of the application can construct a generative adversarial network (GAN) model using a generator network and a discriminator network, and then perform pruning operations on the GAN model. By analyzing the importance of parameters in each layer of the model, unimportant weights are pruned, thereby reducing the number of parameters in the model. Specifically, by deleting weights at a rate of 10%, and finally optimizing to 30%, the performance of the model is negligible compared to before pruning. After pruning, the model retains its main functions while significantly reducing computational requirements, thus obtaining a GAN that meets certain computational needs.

[0054] This application's embodiments utilize adversarial training of a generative adversarial network (GAN) model to generate pseudo-measurement data that better reflects the actual operating state of the distribution network, thereby improving the reliability of system state estimation. Based on the GAN-based pseudo-measurement generation mechanism, the generator and discriminator continuously optimize the generated pseudo-measurement data through adversarial training, ensuring its accuracy and authenticity.

[0055] In step S104, a lightweight distribution network pseudo-measurement generation model based on generative adversarial networks is trained using the training set and the test set to generate a distribution network pseudo-measurement generation model that meets the preset optimal conditions.

[0056] It is understood that the distribution network pseudo-measurement generation model that meets the preset optimal conditions in the embodiments of this application can be the optimal distribution network pseudo-measurement generation model.

[0057] In actual implementation, the embodiments of this application can use training and testing sets to train a lightweight distribution network pseudo-measurement generation model based on generative adversarial networks. During the training process of the distribution network pseudo-measurement generation model, the parameters and hyperparameters of the model are optimized to obtain the optimal distribution network pseudo-measurement generation model.

[0058] The embodiments of this application can use training and testing sets to train a lightweight distribution network pseudo-measurement generation model based on generative adversarial networks, thereby providing support for improving the accuracy and robustness of pseudo-measurements and enhancing the reliability of distribution network state estimation.

[0059] It should be noted that the preset optimal conditions can be set by those skilled in the art according to the actual situation, and no specific restrictions are imposed here.

[0060] Optionally, in one embodiment of this application, a lightweight distribution network pseudo-measurement generation model based on a generative adversarial network is trained using a training set and a test set to generate a distribution network pseudo-measurement generation model that meets preset optimal conditions. This includes: acquiring at least one input data sample from meteorological data, distribution network load nodes, and branch nodes; training the generator and discriminator alternately based on at least one input data sample to generate training results; monitoring performance indicators on the validation set based on the training results, and stopping the training of the lightweight distribution network pseudo-measurement generation model when the performance indicators meet preset conditions, so as to generate a distribution network pseudo-measurement generation model that meets preset optimal conditions.

[0061] It is understood that the preset conditions in the embodiments of this application can be conditions under which performance no longer improves.

[0062] In actual implementation, the embodiments of this application can determine the input data samples, including meteorological data, distribution network load nodes, and branch nodes. The generator and discriminator are trained alternately to generate training results. In each iteration, the generator is first fixed, and the discriminator is trained to maximize its classification accuracy. Then, the discriminator is fixed, and the generator is trained to minimize the difference between the generated data and the real data. The number of training iterations is set to 100, and the loss function is as follows:

[0063] LD=-log(D(X))-log(1-D(Y)),

[0064] Where D(X) is the data generated by the generator, and D(Y) is the actual measurement value.

[0065] Furthermore, embodiments of this application can perform regularization and Dropout: add L1(λ∑∣w∣) and L2(λ∑w) to the hidden layer. 2Regularization terms and a Dropout mechanism (randomly discarding neurons with probability p) are used to prevent model overfitting. Embodiments of this application can utilize early stopping to monitor performance on the validation set (such as the classification accuracy of the discriminator or the MSE between generated and real data). When performance no longer improves, training of the lightweight distribution network pseudo-measurement generation model is stopped to generate an optimal distribution network pseudo-measurement generation model.

[0066] The lightweight design in this embodiment reduces the computational complexity and number of parameters of the model, improves computational efficiency, is suitable for resource-constrained environments and real-time application scenarios, and reduces the consumption of computing resources while ensuring the quality of generated data, thus effectively supporting real-time monitoring and scheduling of power distribution networks.

[0067] It should be noted that the preset conditions and the preset optimal conditions can be set by those skilled in the art according to the actual situation, and no specific restrictions are imposed here.

[0068] In step S105, the data in the test set is input into the distribution network pseudo-measurement generation model that meets the preset optimal conditions to obtain the high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network. The high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network are then denormalized to generate the denormalized high-frequency pseudo-measurement values ​​of each branch and node of the medium-voltage distribution network.

[0069] It is understood that the embodiments of this application are capable of generating and evaluating pseudo-measurement data.

[0070] As one possible implementation, embodiments of this application can input data from the test set into a lightweight distribution network pseudo-measurement generation model based on generative adversarial networks to calculate real-time high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network, and then perform inverse normalization processing on them to finally obtain the actual high-frequency pseudo-measurement values ​​of each branch and node of the medium-voltage distribution network, that is, to obtain the inverse normalized high-frequency pseudo-measurement values ​​of each branch and node of the medium-voltage distribution network.

[0071] The embodiments of this application possess strong adaptability, with optimization strategies enhancing generalization and robustness to ensure the generation of stable and accurate pseudo-measurement data under different environmental and data conditions. This application improves the generalization and robustness of pseudo-measurement generation by optimizing the model structure and enhancing spatiotemporal feature extraction capabilities, achieving efficient and accurate real-time pseudo-measurement generation to meet the operational requirements of smart grids.

[0072] It should be noted that the preset optimal conditions can be set by those skilled in the art according to the actual situation, and no specific restrictions are imposed here.

[0073] In one embodiment of this application, the formula for calculating the denormalized pseudo-measurement value is as follows:

[0074] y m =y n *(y max -y min )+y min ,

[0075] Among them, y n y represents the pseudo-measurement generated by the model. max y min These represent the maximum and minimum values ​​of the high-frequency pseudo-quantity measurements at the nodes, respectively.

[0076] Optionally, in one embodiment of this application, after generating the actual high-frequency pseudo-measurement values ​​of each branch and node of the medium-voltage distribution network, the method further includes: inputting a random noise vector into the generator to generate pseudo-measurement data; evaluating the pseudo-measurement data to generate an evaluation result, and based on the evaluation result, integrating the pseudo-measurement data into the monitoring system of the medium-voltage distribution network to generate integrated pseudo-measurement data; and updating the generative adversarial network based on the integrated pseudo-measurement data and the high-frequency pseudo-measurement values ​​of the nodes to determine the updated generative adversarial network.

[0077] In practical implementation, the embodiments of this application can be integrated and optimized for the generation of pseudo-measurements in power distribution networks. The specific steps are as follows:

[0078] 1) After training, pseudo-measurement data X^ is generated by inputting a random noise vector z into the generator. The authenticity of the generated data can be controlled by adjusting the distribution and dimension of z.

[0079] 2) Conduct quality assessments on the generated pseudo-measurement data, including similarity to real data (e.g., through indicators such as mean squared error (MSE) and Pearson correlation coefficient), data distribution characteristics, and validity verification in practical applications.

[0080] 3) Integrate the generated pseudo-measurement data into the distribution network monitoring system to supplement missing data, improve data density, and support more accurate power grid state estimation, fault detection and location, load forecasting and other applications.

[0081] 4) Continuously optimize the GAN model and update the model regularly based on newly collected real measurement data to maintain the high fidelity of pseudo measurement data.

[0082] 5) Utilize transfer learning to apply the updated model to the generation of pseudo-measurements in distribution networks under different topologies and load conditions.

[0083] The enhanced generalization and robustness of the embodiments in this application make the method more adaptable to different scenarios and data distribution variations, ensuring the stability and reliability of pseudo-measurement data. In particular, when facing noise and abnormal data, the system can maintain high accuracy and stability, providing strong support for the efficient operation of smart grids.

[0084] Specifically, it can be combined with Figure 3 As shown, a specific embodiment is used to elaborate in detail on the working principle of the lightweight distribution network pseudo-measurement generation method based on generative adversarial networks in this application.

[0085] like Figure 3 As shown, embodiments of this application may include the following steps:

[0086] Step S301: Obtain relevant data of medium-voltage distribution network, including historical power, voltage amplitude, current amplitude, and meteorological data.

[0087] Step S302: Data missing data completion, alignment, and normalization.

[0088] Step S303: Using the spatiotemporal data of each branch and node as input, construct the training set and the test set.

[0089] Step S304: Offline training of the distribution network pseudo-measurement generation model.

[0090] Step S305: Determine if the model is optimal. If yes, proceed to step S306; otherwise, re-execute step S304.

[0091] Step S306: Determine if the pruning ratio is optimal. If yes, proceed to step S307; otherwise, repeat step S304.

[0092] Step S307: Calculate the high-frequency pseudo-measurement using the online application of the optimal model on the test set.

[0093] Step S308: Inverse normalization process.

[0094] Step S309: Pseudo-measure the actual values ​​of each branch and node.

[0095] The lightweight pseudo-measurement generation method for distribution networks based on generative adversarial networks (GANs) proposed in this application improves the generalization and robustness of pseudo-measurement generation by optimizing the model structure and enhancing spatiotemporal feature extraction capabilities, achieving efficient and accurate real-time pseudo-measurement generation to meet the operational requirements of smart grids. Adversarial training is performed between the generator and discriminator of a GAN to effectively capture the inherent patterns and distribution characteristics of the data. Through deep learning mechanisms, the spatiotemporal features of distribution network measurement data are autonomously extracted to generate high-quality pseudo-measurement data, compensating for the deficiencies of actual measurement data. This scheme not only improves the accuracy and robustness of pseudo-measurements but also enhances the reliability of distribution network state estimation, providing strong support for the efficient operation of smart grids. Therefore, it solves the problems of related technologies that typically rely on complex model structures, have a large number of parameters, low computational efficiency, and are difficult to meet the needs of real-time monitoring of distribution networks. Furthermore, these technologies have limited performance in extracting spatiotemporal features of distribution network data, especially when dealing with highly nonlinear and dynamically changing data, resulting in pseudo-measurement data that lacks accuracy and fails to accurately reflect the system's operating state.

[0096] Next, referring to the accompanying drawings, a lightweight distribution network pseudo-measurement generation device based on generative adversarial networks, according to an embodiment of this application, is described.

[0097] Figure 4 This is a schematic diagram of the lightweight distribution network pseudo-measurement generation device based on generative adversarial networks according to an embodiment of this application.

[0098] like Figure 4 As shown, the lightweight distribution network pseudo-measurement generation device 10 based on generative adversarial networks includes: an acquisition module 100, a data generation module 200, a partitioning module 300, a construction module 400, a training module 500, and a generation module 600.

[0099] Specifically, the acquisition module 100 is used to acquire at least one of the following measurement data from the medium-voltage distribution network: voltage, current, power factor, active power, and reactive power.

[0100] The data generation module 200 is used to generate normalized historical measurement data based on at least one measurement data.

[0101] The partitioning module 300 is used to divide the normalized historical measurement data into training and testing sets according to a preset ratio.

[0102] The construction module 400 is used to build a generative adversarial network model using a generator network and a discriminator network, and to perform pruning operations on the generative adversarial network model to generate a generative adversarial network that meets preset computational requirements.

[0103] Training module 500 is used to train a lightweight distribution network pseudo-measurement generation model based on generative adversarial networks using training and test sets, so as to generate a distribution network pseudo-measurement generation model that meets preset optimal conditions.

[0104] The generation module 600 is used to input the data in the test set into the distribution network pseudo-measurement generation model that meets the preset optimal conditions, so as to obtain the high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network, and to perform inverse normalization processing on the high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network to generate the inverse normalized high-frequency pseudo-measurement values ​​of each branch and node of the medium-voltage distribution network.

[0105] Optionally, in one embodiment of this application, the data generation module includes: a removal unit, used to remove outliers and missing values ​​from at least one measurement data to obtain new measurement data; a filling unit, used to fill in missing values ​​in the new measurement data to generate measurement data that meets a preset completeness condition; and a data generation unit, used to time-align the measurement data that meets the preset completeness condition to determine historical measurement data, and normalize the historical measurement data to generate normalized historical measurement data.

[0106] Optionally, in one embodiment of this application, the formula for calculating the normalized historical measurement data is:

[0107]

[0108] Where, x ′ k Let x be the k-th normalized historical measurement. k For the k-th original historical measurement, x mink Let x be the minimum value among k sets of data. maxk It represents the maximum value of k sets of data.

[0109] Optionally, in one embodiment of this application, the training module 500 includes: an acquisition unit, a training unit, and a model generation unit.

[0110] The acquisition unit is used to acquire at least one input data sample from meteorological data, distribution network load nodes, and branch nodes.

[0111] A training unit is used to alternately train a generator and a discriminator based on at least one input data sample to generate training results.

[0112] The model generation unit is used to monitor the performance indicators on the validation set based on the training results, and stop the training of the lightweight distribution network pseudo-measurement generation model when the performance indicators meet the preset conditions, so as to generate a distribution network pseudo-measurement generation model that meets the preset optimal conditions.

[0113] Optionally, in one embodiment of this application, the lightweight distribution network pseudo-measurement generation device 10 based on generative adversarial networks further includes: an input module, an evaluation module, and an update module.

[0114] The input module is used to input random noise vectors into the generator after generating the high-frequency pseudo-measurement values ​​of each branch and node of the actual medium-voltage distribution network, so as to generate pseudo-measurement data.

[0115] The evaluation module is used to evaluate pseudo-measurement data to generate evaluation results, and based on the evaluation results, integrate the pseudo-measurement data into the monitoring system of medium-voltage distribution network to generate integrated pseudo-measurement data.

[0116] The update module is used to update the generative adversarial network based on the integrated pseudo-measurement data and the high-frequency pseudo-measurement values ​​of the nodes, so as to determine the updated generative adversarial network.

[0117] Optionally, in one embodiment of this application, the formula for calculating the denormalized pseudo-measurement is:

[0118] y m =y n *(y max -y min )+y min ,

[0119] Among them, y n y represents the pseudo-measurement generated by the model. max y min These represent the maximum and minimum values ​​of the high-frequency pseudo-quantity measurements at the nodes, respectively.

[0120] It should be noted that the foregoing explanation of the embodiment of the lightweight distribution network pseudo-measurement generation method based on generative adversarial networks also applies to the lightweight distribution network pseudo-measurement generation device based on generative adversarial networks in this embodiment, and will not be repeated here.

[0121] The lightweight distribution network pseudo-measurement generation device based on generative adversarial networks (GANs) proposed in this application fully utilizes spatiotemporal feature data and measurement data collected by known sensing devices. It designs a lightweight pseudo-measurement model to improve its generalization and robustness, enhancing the ability of the generated data to reflect the actual operating state of the distribution network. It generates the required pseudo-measurements for the distribution network in real time and accurately. Integrating advanced artificial intelligence technology, it innovatively proposes a GAN-based pseudo-measurement generation method to address the problem of distribution network measurement data. This opens up a new path for the intelligent and precise management of distribution networks and is of great significance for promoting the development of the energy internet and improving the overall operating efficiency of the power grid. Therefore, it solves the problems of related technologies that typically rely on complex model structures, have a large number of parameters, low computational efficiency, and are difficult to meet the needs of real-time monitoring of distribution networks. Furthermore, these technologies have limited performance in extracting the spatiotemporal features of distribution network data, especially when dealing with highly nonlinear and dynamically changing data, resulting in pseudo-measurement data that lacks accuracy and cannot accurately reflect the system's operating state.

[0122] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include:

[0123] The memory 501, the processor 502, and the computer program stored on the memory 501 and capable of running on the processor 502.

[0124] When processor 502 executes the program, it implements the lightweight distribution network pseudo-measurement generation method based on generative adversarial networks provided in the above embodiments.

[0125] Furthermore, electronic devices also include:

[0126] Communication interface 503 is used for communication between memory 501 and processor 502.

[0127] The memory 501 is used to store computer programs that can run on the processor 502.

[0128] The memory 501 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0129] If the memory 501, processor 502, and communication interface 503 are implemented independently, then the communication interface 503, memory 501, and processor 502 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0130] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 503 are integrated on a single chip, then the memory 501, processor 502, and communication interface 503 can communicate with each other through an internal interface.

[0131] Processor 502 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0132] This embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described lightweight distribution network pseudo-measurement generation method based on generative adversarial networks.

[0133] This application also provides a computer program product storing a computer program that, when executed by a processor, implements the above-described lightweight distribution network pseudo-measurement generation method based on generative adversarial networks.

[0134] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0135] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0136] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0137] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0138] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0139] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0140] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0141] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A lightweight method for generating pseudo-measurements in a distribution network based on generative adversarial networks, characterized in that, Includes the following steps: Collect at least one of the following measurement data from the medium-voltage distribution network: voltage, current, power factor, active power, and reactive power; Generate normalized historical measurement data based on at least one of the measurement data; The normalized historical measurement data is divided into a training set and a test set according to a preset ratio; A generative adversarial network (GAN) model is constructed using a generator network and a discriminator network, and the GAN model is pruned to generate a GAN that meets preset computational requirements. The training set and the test set are used to train a lightweight distribution network pseudo-measurement generation model based on the generative adversarial network, so as to generate a distribution network pseudo-measurement generation model that meets the preset optimal conditions. The data in the test set is input into the distribution network pseudo-measurement generation model that meets the preset optimal conditions to obtain high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network. The high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network are then denormalized to generate denormalized high-frequency pseudo-measurement values ​​of each branch and node of the medium-voltage distribution network.

2. The method according to claim 1, characterized in that, The step of generating normalized measurement data based on the at least one measurement data includes: Remove outliers and missing values ​​from the at least one measurement data item to obtain new measurement data; Fill in the missing values ​​in the new measurement data to generate measurement data that meets the preset completeness conditions; The measurement data that meet the preset completeness conditions are time-aligned to determine the historical measurement data, and the historical measurement data is normalized to generate the normalized historical measurement data.

3. The method according to claim 1, characterized in that, The formula for calculating the normalized historical measurement data is as follows: Where, x ′ k Let x be the k-th normalized historical measurement. k For the k-th original historical measurement, x mink Let x be the minimum value among k sets of data. maxk It represents the maximum value of k sets of data.

4. The method according to claim 1, characterized in that, The step of training a lightweight distribution network pseudo-measurement generation model based on the generative adversarial network using the training set and the test set to generate a distribution network pseudo-measurement generation model that satisfies preset optimal conditions includes: Acquire at least one input data sample from meteorological data, distribution network load nodes, and branch nodes; Based on the at least one input data sample, the generator and discriminator are trained alternately to generate training results; Based on the training results, the performance indicators on the validation set are monitored, and when the performance indicators meet the preset conditions, the training of the lightweight distribution network pseudo-measurement generation model is stopped, so as to generate the distribution network pseudo-measurement generation model that meets the preset optimal conditions.

5. The method according to claim 1, characterized in that, After generating the actual high-frequency pseudo-measurement values ​​for each branch and node of the medium-voltage distribution network, the following is also included: A random noise vector is input into the generator to generate pseudo-measurement data; The pseudo-measurement data is evaluated to generate an evaluation result, and based on the evaluation result, the pseudo-measurement data is integrated into the monitoring system of the medium-voltage distribution network to generate integrated pseudo-measurement data; Based on the integrated pseudo-measurement data, the generative adversarial network is updated according to the high-frequency pseudo-measurement values ​​of the nodes to determine the updated generative adversarial network.

6. The method according to claim 1, characterized in that, The formula for calculating the pseudo-measured value after inverse normalization is: and m / and n *(and max -and min )+and min , Among them, y n y represents the pseudo-measurement generated by the model. max y min These are the maximum and minimum values ​​of the high-frequency pseudo-quantity measurements of the node, respectively.

7. A lightweight pseudo-measurement generation device for distribution networks based on generative adversarial networks, characterized in that, include: The data acquisition module is used to acquire at least one of the following measurement data from medium-voltage distribution networks: voltage, current, power factor, active power, and reactive power. The data generation module is used to generate normalized historical measurement data based on the at least one measurement data. The partitioning module is used to divide the normalized historical measurement data into a training set and a test set according to a preset ratio. A construction module is used to build a generative adversarial network model using a generator network and a discriminator network, and to perform pruning operations on the generative adversarial network model to generate a generative adversarial network that meets preset computational requirements. The training module is used to train a lightweight distribution network pseudo-measurement generation model based on the generative adversarial network using the training set and the test set, so as to generate a distribution network pseudo-measurement generation model that meets the preset optimal conditions. The generation module is used to input the data in the test set into the distribution network pseudo-measurement generation model that meets the preset optimal conditions, so as to obtain the high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network, and to perform inverse normalization processing on the high-frequency pseudo-measurements of each branch and node of the medium-voltage distribution network to generate the inverse normalized high-frequency pseudo-measurement values ​​of each branch and node of the medium-voltage distribution network.

8. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, the processor executing the program to implement the lightweight distribution network pseudo-measurement generation method based on generative adversarial networks as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the lightweight distribution network pseudo-measurement generation method based on generative adversarial networks as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, The computer program is executed to implement the lightweight distribution network pseudo-measurement generation method based on generative adversarial networks as described in any one of claims 1-6.

Images