Topological structure information guided power distribution network measurement data completion method and system and medium

Abstract

The invention discloses a topological structure information guided power distribution network measurement data complementation method and system and a medium, and belongs to the technical field of power distribution networks, and the method comprises the steps: obtaining incomplete historical measurement data under different topological structure sets, and building a measurement data set; building a plurality of conditional auto-encoder models respectively used for missing complementation of different measurement data based on a full-connection neural network; designing a loss function of each conditional auto-encoder model, performing joint training on the plurality of conditional auto-encoder models through the measurement data set, and adjusting hyper-parameters for multiple times to obtain a power distribution network measurement data completion model guided by topological structure information; and inputting incomplete measurement data and a corresponding topological structure into the topological structure information guided power distribution network measurement data completion model, and outputting complete measurement data by the topological structure information guided power distribution network measurement data completion model. According to the invention, the complemented measurement data is controlled to be more authentic and higher in precision by sensing the change of the topological structure.

Description

Technical Field

[0001] The present invention belongs to the technical field of distribution network, and in particular relates to a distribution network measurement data completion method, system and medium guided by topology structure information. Background Art

[0002] With the large-scale access of distributed renewable energy to the distribution network, the structure of the distribution network will become more complex and changeable. The current distribution system is becoming more intelligent, and accurate and rapid perception of the operating status is crucial to the stable operation of the system. However, in the actual operation of the system, due to various factors (such as economic costs, equipment failures, communication failures, etc.), some real-time measurement data in the distribution network may be missing, which brings certain potential risks to the operation and monitoring of the system. Therefore, it is particularly important to study a method that can accurately complete the missing measurement data of the distribution network under complex and changeable topological structures.

[0003] The existing research on the missing data completion technology of distribution network measurement mainly includes traditional methods and artificial intelligence methods. The traditional missing data completion method mainly relies on mathematical statistical methods such as interpolation technology, but it has certain limitations when dealing with the measurement data with high complexity and high representation difficulty generated by the current distribution network with complex and changeable topology, resulting in great difficulty in completion and low accuracy. At present, artificial intelligence methods mainly include machine learning methods and deep learning methods. Machine learning methods such as decision trees and Bayesian networks require certain prior knowledge and are highly dependent on training data and feature engineering, which makes it difficult for the model to represent high-dimensional and complex measurement data. The advantages of deep learning-based methods are its adaptive ability and reliance on data-driven, and its low dependence on physical mathematical models. At present, the commonly used deep learning methods mainly focus on the use of generative adversarial networks, but generative adversarial networks involve the game between generators and discriminators during the training process, and are prone to oscillation and collapse problems. In addition, when generating data, they mainly focus on how to generate realistic data. In the case of complex and changeable topological structures, without the guidance of topological structure information, the generated results may be difficult to control.

[0004] Purpose of the Invention The purpose of the present invention is to address the problems in the above-mentioned prior art and provide a distribution network measurement data completion method, system and medium guided by topology information, which controls the completed measurement data to be more authentic and accurate by sensing the changes in the topology structure, thereby improving the monitoring, analysis and control capabilities of the system operation status.

[0005] In order to achieve the above object, the present invention has the following technical solutions: In a first aspect, a method for completing distribution network measurement data guided by topology information is provided, comprising: Obtain incomplete historical measurement data under different topological structure sets and establish measurement data sets; Based on a fully connected neural network, multiple conditional autoencoder models are constructed for missing data of different measurements. Design the loss function of each conditional autoencoder model, and jointly train multiple conditional autoencoder models through the measurement data set, and adjust the hyperparameters multiple times to obtain the distribution network measurement data completion model guided by the topology structure information; Incomplete measurement data and corresponding topology structures are input into the distribution network measurement data completion model guided by the topology structure information, and the distribution network measurement data completion model guided by the topology structure information outputs complete measurement data.

[0006] As a preferred solution, the step of obtaining incomplete historical measurement data under different topological structure sets and establishing a measurement data set includes: Using simulation software tools, power flow calculations are performed by changing the topology structure, simulating and obtaining complete historical measurement data under different topology sets, and recording the corresponding topology structure information; The complete historical measurement data under different topological structures is introduced into a sample Binary mask matrix of consistent dimension Calculation Sample With the binary mask matrix The Hadamard product between is used to describe the missing measurement data, and Gaussian distribution noise is superimposed to obtain incomplete historical measurement data that is consistent with the actual situation.

[0007] As a preferred solution, the simulation software tool is used to perform flow calculations by changing the topology structure, and the complete historical measurement data under different topology sets are simulated to obtain the corresponding topology structure information. When the wind, light and load are modeled with uncertainty, in the uncertainty modeling process, wind power obeys Weibull distribution, light obeys Beta distribution, and load obeys normal distribution.

[0008] As a preferred solution, the autoencoder model consists of an encoder and a decoder composed of a fully connected neural network; the input layer of the encoder receives missing real-time measurement data and corresponding topological structure information, and the topological structure information is converted into a vector in a latent space as a conditional embedding by designing an embedding layer, and the real-time measurement data and the embedded topological structure information are jointly encoded as latent variables; the input layer of the decoder is used to receive latent variables, and the latent variables generate completed real-time measurement data under the corresponding topological structure through the decoder.

[0009] As a preferred solution, the topological structure information is converted into a vector in a latent space by designing an embedding layer as a conditional embedding, and the steps of jointly encoding the real-time measurement data and the embedded topological structure information into latent variables include: A latent space is introduced, and the input measurement data is mapped to the latent space through the encoder according to the following formula, so that the vector in the latent space contains the characteristics of the input data, and the latent variables of the measurement data are obtained:

[0010] In the formula, is the output vector of the encoder, i.e., the encoded representation of the incomplete measurement data; is the trainable weight parameter of the encoder; is the bias of the encoder; the above parameters are updated and iterated by the optimizer; is the activation function; The topological structure information is embedded into the autoencoder model as conditional information, and the embedding layer is used to map the topological structure information into a vector in the latent space. The dimension is consistent with the dimension of the input feature, and the latent variable of the topological structure information is obtained:

[0011] In the formula, Represents the vector of topological information after being encoded by the embedding layer; input_dim represents the number of labels of topological information, reflecting the number of types of topological structures; output_dim represents the dimension of the latent space, that is, the length of the embedded vector, which is the same as the output vector of the encoder. The vector length of is consistent; Embedding represents the operation of the embedding layer, which is used to map each topological structure information into a vector with a length of output_dim; The latent variables of the measurement data and the latent variables of the topological structure information are jointly encoded to form the overall latent variables. The joint encoding is expressed by the following formula:

[0012] In the formula, A joint encoding vector representing real-time measurement data and topological structure information in the latent space; Represents a vector concatenation operation.

[0013] As a preferred solution, in the step of generating the real-time measurement data completed under the corresponding topological structure by the decoder, the generated joint coding vector is input into the decoder, and the missing reconstruction value of the measurement data is generated by the decoder according to the following formula:

[0014] In the formula, is the output vector of the decoder, i.e., the reconstructed value; is the trainable weight parameter of the decoder; is the bias of the decoder.

[0015] As a preferred solution, in the step of designing the loss function of each conditional autoencoder model and jointly training multiple conditional autoencoder models through a measurement data set, different conditional autoencoder models process missing completion of different measurement data respectively, and after initializing the network parameters of the autoencoder model, the back propagation algorithm is used to train and update the network parameters; The measured data includes node voltage, node injected active power and node injected reactive power. The corresponding loss functions are defined for the node voltage, node injected active power and node injected reactive power conditional autoencoder model as follows:

[0016]

[0017]

[0018] In the formula, , , They are the mean square error (MSE) values ​​when complementing the node voltage, node injected active power, and node injected reactive power, respectively; is the number of samples; , , For sample The true values ​​of node voltage, node injected active power and node injected reactive power; , , For sample The node voltage, node injected active power and node injected reactive power are the complement values; According to the loss function of the designed single conditional autoencoder model, the calculation expression of the total loss function of the model is obtained as follows:

[0019] In the formula, , , The weights of the autoencoder model for node voltage, node injected active power and node injected reactive power conditions are adaptively determined by the training process according to the difficulty of completing the measurement data; the designed model total loss function is used as the optimization target of the distribution network measurement data completion model guided by the topology structure information, the number of training rounds is set, and the adaptive optimization algorithm is used to update the model parameters to obtain the distribution network measurement data completion model guided by the topology structure information.

[0020] As a preferred solution, in the step of adjusting hyperparameters multiple times to obtain a distribution network measurement data completion model guided by topology information, the hyperparameters include a combination of one or more of a learning rate and a regularization coefficient.

[0021] In a second aspect, a distribution network measurement data completion system guided by topology information is provided, comprising: The measurement data set acquisition module is used to acquire incomplete historical measurement data under different topological structure sets and establish measurement data sets; An autoencoder model building module, which is used to build multiple conditional autoencoder models for missing data completion of different measurement data based on a fully connected neural network; The joint training module is used to design the loss function of each conditional autoencoder model, and jointly train multiple conditional autoencoder models through the measurement data set, and adjust the hyperparameters multiple times to obtain the distribution network measurement data completion model guided by the topology structure information; The complete measurement data output module is used to input incomplete measurement data and corresponding topology structure into the distribution network measurement data completion model guided by the topology structure information, and the distribution network measurement data completion model guided by the topology structure information outputs complete measurement data.

[0022] As a preferred solution, the measurement data set acquisition module uses a simulation software tool to perform power flow calculations by changing the topological structure, simulate and acquire complete historical measurement data under different topological structure sets, and record the corresponding topological structure information; The complete historical measurement data under different topological structures is introduced into a sample Binary mask matrix of consistent dimension Calculation Sample With the binary mask matrix The Hadamard product between is used to describe the missing measurement data, and Gaussian distribution noise is superimposed to obtain incomplete historical measurement data that is consistent with the actual situation.

[0023] As a preferred solution, the autoencoder model constructed by the autoencoder model building module consists of an encoder and a decoder composed of a fully connected neural network; the input layer of the encoder receives missing real-time measurement data and corresponding topological structure information, and the topological structure information is converted into a vector in a latent space as a conditional embedding by designing an embedding layer, and the real-time measurement data and the embedded topological structure information are jointly encoded as latent variables; the input layer of the decoder is used to receive latent variables, and the latent variables generate completed real-time measurement data under the corresponding topological structure through the decoder.

[0024] As a preferred solution, the autoencoder model building module converts the topological structure information into a vector in a latent space by designing an embedding layer as a conditional embedding. In the process of jointly encoding the real-time measurement data and the embedded topological structure information into latent variables, a latent space is introduced. The input measurement data is mapped to the latent space through the encoder according to the following formula, so that the vector in the latent space contains the characteristics of the input data, and the latent variables of the measurement data are obtained:

[0025] In the formula, is the output vector of the encoder, i.e., the encoded representation of the incomplete measurement data; is the trainable weight parameter of the encoder; is the bias of the encoder; the above parameters are updated and iterated by the optimizer; is the activation function; The topological structure information is embedded into the autoencoder model as conditional information, and the embedding layer is used to map the topological structure information into a vector in the latent space. The dimension is consistent with the dimension of the input feature, and the latent variable of the topological structure information is obtained:

[0026] In the formula, Represents the vector of topological information after being encoded by the embedding layer; input_dim represents the number of labels of topological information, reflecting the number of types of topological structures; output_dim represents the dimension of the latent space, that is, the length of the embedded vector, which is the same as the output vector of the encoder. The vector length of is consistent; Embedding represents the operation of the embedding layer, which is used to map each topological structure information into a vector with a length of output_dim; The latent variables of the measurement data and the latent variables of the topological structure information are jointly encoded to form the overall latent variables. The joint encoding is expressed by the following formula:

[0027] In the formula, A joint encoding vector representing real-time measurement data and topological structure information in the latent space; Represents the concatenation operation of vectors; When the latent variable generates the completed real-time measurement data under the corresponding topological structure through the decoder, the generated joint coding vector is input into the decoder, and the missing reconstruction value of the measurement data is generated through the decoder according to the following formula:

[0028] In the formula, is the output vector of the decoder, i.e., the reconstructed value; is the trainable weight parameter of the decoder; is the bias of the decoder.

[0029] As a preferred solution, the joint training module processes the missing completion of different measurement data using different conditional autoencoder models respectively, and after initializing the network parameters of the autoencoder model, the back propagation algorithm is used to train and update the network parameters; The measured data includes node voltage, node injected active power and node injected reactive power. The corresponding loss functions are defined for the node voltage, node injected active power and node injected reactive power conditional autoencoder model as follows:

[0030]

[0031]

[0032] In the formula, , , They are the mean square error (MSE) values ​​when complementing the node voltage, node injected active power, and node injected reactive power, respectively; is the number of samples; , , For sample The true values ​​of node voltage, node injected active power and node injected reactive power; , , For sample The node voltage, node injected active power and node injected reactive power are the complement values; According to the loss function of the designed single conditional autoencoder model, the calculation expression of the total loss function of the model is obtained as follows:

[0033] In the formula, , , The weights of the autoencoder model for node voltage, node injected active power and node injected reactive power conditions are adaptively determined by the training process according to the difficulty of completing the measurement data; the designed model total loss function is used as the optimization target of the distribution network measurement data completion model guided by the topology structure information, the number of training rounds is set, and the adaptive optimization algorithm is used to update the model parameters to obtain the distribution network measurement data completion model guided by the topology structure information.

[0034] According to a third aspect, an electronic device is provided, comprising a processor and a memory, wherein the processor is configured to execute a computer program stored in the memory to implement the distribution network measurement data completion method guided by the topology information.

[0035] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, the distribution network measurement data completion method guided by the topology structure information is implemented.

[0036] Compared with the prior art, the first aspect of the present invention has at least the following beneficial effects: The present invention realizes the completion of active distribution network measurement data under the guidance of topological structure information, and can effectively restore multi-dimensional missing measurement data. Based on a fully connected neural network, multiple conditional autoencoder models are respectively formed for different missing measurement data completion. The data-driven method can be adaptively adjusted to adapt to the diversity of measurement data and can capture the complex nonlinear relationship in the measurement data, which solves the limitation that the traditional distribution network measurement data missing completion method may need to manually adjust the model parameters or use specific completion rules. The present invention embeds the topological structure information corresponding to the measurement data into the neural network as a condition, so that the model training process can learn the specific data patterns and features between the measurement data and the topological structure, so that the data generated by the model is more real and reasonable. At the same time, the conditional autoencoder is usually more stable and converges faster during the training process, which solves the problems such as training collapse faced by the traditional generative adversarial network GAN algorithm, and compared with the unconditional generation of data by the generative adversarial network GAN algorithm, the introduction of conditional information in the present invention improves the accuracy of the completion result. Based on the shortcomings of the existing measured data completion methods, the present invention is based on the current operating conditions of the distribution network with changeable topological structures, introduces conditional autoencoders as a tool for measurement data completion, and embeds the topological structure information as the condition of the autoencoder. The missing measurement data is intelligently filled in according to the existing partial measurement data. At the same time, compared with the existing deep learning methods, the present invention can better control the completed measurement data to be more authentic and more accurate by sensing the changes in the topological structure, thereby improving the monitoring, analysis and control capabilities of the operating status of the distribution system.

[0037] It can be understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0039] Figure 1 A schematic diagram of the design principle of a method for completing distribution network measurement data guided by topology information according to an embodiment of the present invention; Figure 2 A schematic diagram of the structure of a distribution network measurement data completion model based on a conditional autoencoder according to an embodiment of the present invention; Figure 3 A flowchart of constructing a measurement data set according to an embodiment of the present invention; Figure 4 A flowchart of a conditional autoencoder model constructed in an embodiment of the present invention; Figure 5 A flow chart of training and application of a distribution network measurement data completion model guided by topology information according to an embodiment of the present invention; Figure 6 Overall flow chart of the method for completing distribution network measurement data guided by topology information according to an embodiment of the present invention; Figure 7 A comparison diagram of the effects of the distribution network measurement data completion method guided by topology information in an embodiment of the present invention and the traditional method of completing missing distribution network measurement data using the GAN algorithm. DETAILED DESCRIPTION

[0040] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0041] See also Figure 1 and Figure 6 The embodiment of the present invention proposes a distribution network measurement data completion method guided by topology information, which mainly includes the following steps: S1. Obtain incomplete historical measurement data under different topological structure sets and establish a measurement data set; S2, based on the fully connected neural network, multiple conditional autoencoder models are constructed for missing data completion of different measurement data; S3. Design the loss function of each conditional autoencoder model, and jointly train multiple conditional autoencoder models through the measurement data set, and adjust the hyperparameters multiple times to obtain the distribution network measurement data completion model guided by the topology structure information; S4. Inputting incomplete measurement data and corresponding topology structure into the distribution network measurement data completion model guided by the topology structure information, and the distribution network measurement data completion model guided by the topology structure information outputs complete measurement data.

[0042] See also Figure 3 In a possible implementation, step S1 uses a simulation software tool to perform power flow calculations by changing the topology, simulating and acquiring complete historical measurement data under different topology sets, and recording corresponding topology information. This embodiment uses the MATPOWER software platform in the Matlab tool.

[0043] The measured data include node voltage, node active power, and node reactive power.

[0044] In order to more accurately describe the real environment, step S1 of this embodiment performs uncertainty modeling on wind, light and load. During the uncertainty modeling process, wind power obeys Weibull distribution, light obeys Beta distribution, and load obeys normal distribution.

[0045] The complete historical measurement data under different topological structures is introduced into a sample Binary mask matrix of consistent dimension Calculation Sample With the binary mask matrix The Hadamard product between is used to describe the missing measurement data, and Gaussian distribution noise is superimposed to obtain incomplete historical measurement data that is consistent with the actual situation.

[0046] Furthermore, each set of sample data includes incomplete measurement data such as node voltage amplitude, node active power, node reactive power, etc. and its corresponding complete measurement data and corresponding topology labels.

[0047] See also Figure 2 and Figure 4 In a possible implementation, the autoencoder model of step S2 is composed of an encoder and a decoder composed of a fully connected neural network, and different autoencoders are used to process missing completion tasks for different measurement data.

[0048] The input layer of the encoder is used to receive missing real-time measurement data and corresponding topological structure information, and the topological structure information is converted into a vector in the latent space as a conditional embedding by designing an embedding layer, and the real-time measurement data and the embedded topological structure information are jointly encoded into latent variables; the input layer of the decoder is used to receive latent variables, and the latent variables are used by the decoder to generate the completed real-time measurement data under the corresponding topological structure. The specific process of step S2 is as follows: A latent space is introduced, and the input measurement data is mapped to the latent space through the encoder according to the following formula, so that the vector in the latent space contains the important features of the input data, and the latent variables of the measurement data are obtained:

[0049] In the formula, is the output vector of the encoder, i.e., the encoded representation of the incomplete measurement data; is the trainable weight parameter of the encoder; is the bias of the encoder; the above parameters are updated and iterated by the optimizer; is the activation function.

[0050] The activation function of the embodiment of the present invention uses the parameterized rectified linear unit PReLU function.

[0051] The topological structure information is embedded into the autoencoder model as conditional information, and the embedding layer is used to map the topological structure information into a vector in the latent space. The dimension is consistent with the dimension of the input feature, and the latent variable of the topological structure information is obtained:

[0052] In the formula, Represents the vector of topological information after being encoded by the embedding layer; input_dim represents the number of labels of topological information, reflecting the number of types of topological structures; output_dim represents the dimension of the latent space, that is, the length of the embedded vector, which is the same as the output vector of the encoder. The vector length of is consistent; Embedding represents the operation of the embedding layer, which is used to map each topological structure information into a vector with a length of output_dim; The latent variables of the measurement data and the latent variables of the topological structure information are jointly encoded to form the overall latent variables, ensuring that the conditional information is fully integrated into the latent space. The joint encoding is expressed by the following formula:

[0053] In the formula, A joint encoding vector representing real-time measurement data and topological structure information in the latent space; Represents a vector concatenation operation.

[0054] Furthermore, the generated joint coding vector is input into the decoder, and the missing reconstruction value of the measurement data is generated by the decoder as follows:

[0055] In the formula, is the output vector of the decoder, i.e., the reconstructed value; is the trainable weight parameter of the decoder; is the bias of the decoder.

[0056] See also Figure 5 In a possible implementation, step S3 designs the loss function of each conditional autoencoder model. After initializing the network parameters, the back propagation algorithm is used to train and update the parameters to complete the training of the distribution network measurement data completion model guided by topology information. For each conditional autoencoder, the corresponding reconstruction loss should be defined for independent calculation. Assuming there are N autoencoders, the reconstruction loss of each autoencoder is , then the calculation formula for the overall loss is:

[0057] The loss function of each autoencoder model is specified. The optimization goal of the measurement data completion task is to minimize the reconstruction error between the complete measurement data and the decoder output, so that the decoder can restore the complete measurement data as much as possible. The measurement data of this embodiment includes node voltage, node injected active power, and node injected reactive power. The corresponding loss functions of the node voltage, node injected active power, and node injected reactive power condition autoencoder models are defined as follows:

[0058]

[0059]

[0060] In the formula, , , They are the mean square error (MSE) values ​​when complementing the node voltage, node injected active power, and node injected reactive power, respectively; is the number of samples; , , For sample The true values ​​of node voltage, node injected active power and node injected reactive power; , , For sample The node voltage, node injected active power and node injected reactive power are the complement values; According to the loss function of the designed single conditional autoencoder model, a multi-task joint training method is adopted. Therefore, the total loss of the model should be the weighted sum of the losses of each task. Its parameters are adaptively determined by the training process according to the completion difficulty of the measured data. The calculation expression of the total loss function of the model is obtained as follows:

[0061] In the formula, , , They are the weights of the autoencoder model for node voltage, node injected active power, and node injected reactive power conditions, which are adaptively determined by the training process according to the difficulty of completing the measured data.

[0062] The designed model total loss function is used as the optimization target of the distribution network measurement data completion model guided by the topology structure information. In this embodiment, the number of training rounds is set to 300, and the optimizer uses the Adma algorithm to update the model parameters, so that the distribution network measurement data completion model guided by the topology structure information can be obtained.

[0063] In the step of adjusting the hyperparameters multiple times to obtain the distribution network measurement data completion model guided by the topology information, the hyperparameters include a combination of one or more of a learning rate and a regularization coefficient.

[0064] After the training is completed, the model with the best performance is selected and the corresponding model parameters are saved, and the distribution network measurement data completion model guided by the topology information is obtained. For online real-time measurement data completion, the best model parameters are loaded, incomplete measurement data and the corresponding topology are input, and the distribution network measurement data completion model guided by the topology information outputs complete measurement data.

[0065] See also Figure 7 , attached Figure 7It is a calculation example based on IEEE33 nodes, and randomly selects the measurement completion effect diagram under a certain time section, which respectively shows the missing measurement data completion effects of node voltage, node active power, and node reactive power. When the missing ratio of measurement data reaches more than 30%, the completion error of the GAN algorithm is significantly higher than the completion error of the present invention. The present invention can still achieve good completion accuracy in the more extreme case where the missing ratio reaches 75%. Compared with the voltage completion effect, the power completion error data is relatively large. This is because power is more random and volatile than voltage, so it is more challenging in the feature extraction stage of encoder dimensionality reduction. Although the power completion error is large, relative to the GAN algorithm, the completion accuracy of the present invention is still at a relatively high level, which means that the completion method of the present invention is robust and applicable, and can effectively restore multi-dimensional missing data, providing an effective solution to the problem of missing data in actual application scenarios.

[0066] The comparison data of the measurement completion effect of the present invention and the GAN algorithm is shown in the following table:

[0067] Another embodiment of the present invention further provides a topology information-guided distribution network measurement data completion system, comprising: The measurement data set acquisition module is used to acquire incomplete historical measurement data under different topological structure sets and establish measurement data sets; An autoencoder model building module, which is used to build multiple conditional autoencoder models for missing data completion of different measurement data based on a fully connected neural network; The joint training module is used to design the loss function of each conditional autoencoder model, and jointly train multiple conditional autoencoder models through the measurement data set, and adjust the hyperparameters multiple times to obtain the distribution network measurement data completion model guided by the topology structure information; The complete measurement data output module is used to input incomplete measurement data and corresponding topology structure into the distribution network measurement data completion model guided by the topology structure information, and the distribution network measurement data completion model guided by the topology structure information outputs complete measurement data.

[0068] In a possible implementation, the measurement data set acquisition module uses a simulation software tool to perform power flow calculations respectively by changing the topological structure, simulate and acquire complete historical measurement data under different topological structure sets, and record the corresponding topological structure information; The complete historical measurement data under different topological structures is introduced into a sample Binary mask matrix of consistent dimension Calculation Sample With the binary mask matrix The Hadamard product between is used to describe the missing measurement data, and Gaussian distribution noise is superimposed to obtain incomplete historical measurement data that is consistent with the actual situation.

[0069] In one possible implementation, the autoencoder model constructed by the autoencoder model construction module consists of an encoder and a decoder composed of a fully connected neural network; the input layer of the encoder receives missing real-time measurement data and corresponding topological structure information, and the topological structure information is converted into a vector in a latent space as a conditional embedding by designing an embedding layer, and the real-time measurement data and the embedded topological structure information are jointly encoded as latent variables; the input layer of the decoder is used to receive latent variables, and the latent variables generate completed real-time measurement data under the corresponding topological structure through the decoder.

[0070] In a possible implementation, the autoencoder model building module converts the topological structure information into a vector in a latent space by designing an embedding layer as a conditional embedding. In the process of jointly encoding the real-time measurement data and the embedded topological structure information into latent variables, a latent space is introduced, and the input measurement data is mapped to the latent space through the encoder according to the following formula, so that the vector in the latent space contains the characteristics of the input data, and the latent variables of the measurement data are obtained:

[0071] In the formula, is the output vector of the encoder, i.e., the encoded representation of the incomplete measurement data; is the trainable weight parameter of the encoder; is the bias of the encoder; the above parameters are updated and iterated by the optimizer; is the activation function; The topological structure information is embedded into the autoencoder model as conditional information, and the embedding layer is used to map the topological structure information into a vector in the latent space. The dimension is consistent with the dimension of the input feature, and the latent variable of the topological structure information is obtained:

[0072] In the formula, Represents the vector of topological information after being encoded by the embedding layer; input_dim represents the number of labels of topological information, reflecting the number of types of topological structures; output_dim represents the dimension of the latent space, that is, the length of the embedded vector, which is the same as the output vector of the encoder. The vector length of is consistent; Embedding represents the operation of the embedding layer, which is used to map each topological structure information into a vector with a length of output_dim; The latent variables of the measurement data and the latent variables of the topological structure information are jointly encoded to form the overall latent variables. The joint encoding is expressed by the following formula:

[0073] In the formula, A joint encoding vector representing real-time measurement data and topological structure information in the latent space; Represents the concatenation operation of vectors; When the latent variable generates the completed real-time measurement data under the corresponding topological structure through the decoder, the generated joint coding vector is input into the decoder, and the missing reconstruction value of the measurement data is generated through the decoder according to the following formula:

[0074] In the formula, is the output vector of the decoder, i.e., the reconstructed value; is the trainable weight parameter of the decoder; is the bias of the decoder.

[0075] In a possible implementation, the joint training module processes the missing completion of different measurement data using different conditional autoencoder models respectively, and after initializing the network parameters of the autoencoder model, the back propagation algorithm is used to train and update the network parameters; The measured data includes node voltage, node injected active power and node injected reactive power. The corresponding loss functions are defined for the node voltage, node injected active power and node injected reactive power conditional autoencoder model as follows:

[0076]

[0077]

[0078] In the formula, , , They are the mean square error (MSE) values ​​when complementing the node voltage, node injected active power, and node injected reactive power, respectively; is the number of samples; , , For sample The true values ​​of node voltage, node injected active power and node injected reactive power; , , For sample The node voltage, node injected active power and node injected reactive power are the complement values; According to the loss function of the designed single conditional autoencoder model, the calculation expression of the total loss function of the model is obtained as follows:

[0079] In the formula, , , The weights of the autoencoder model for node voltage, node injected active power and node injected reactive power conditions are adaptively determined by the training process according to the difficulty of completing the measurement data; the designed model total loss function is used as the optimization target of the distribution network measurement data completion model guided by the topology structure information, the number of training rounds is set, and the adaptive optimization algorithm is used to update the model parameters to obtain the distribution network measurement data completion model guided by the topology structure information.

[0080] Another embodiment of the present invention further provides an electronic device, including a processor and a memory, wherein the processor is configured to execute a computer program stored in the memory to implement the topology information-guided distribution network measurement data completion method.

[0081] Another embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, the distribution network measurement data completion method guided by the topology structure information is implemented.

[0082] The computer program includes computer program code, which may be in source code form, object code form, executable file or some intermediate form. The computer-readable storage medium may include: any entity or device, medium, USB flash drive, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electric carrier signal, telecommunication signal and software distribution medium, etc. that can carry the computer program code. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals. For ease of explanation, the above content only shows the part related to the embodiment of the present invention. For specific technical details not disclosed, please refer to the method part of the embodiment of the present invention. The computer-readable storage medium is non-temporary and can be stored in a storage device formed by various electronic devices, which can realize the execution process recorded in the method of the embodiment of the present invention.

[0083] 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 aspects. 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 code.

[0084] 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 process in the flowchart. 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.

[0085] 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.

[0086] 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.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A distribution network measurement data completion method guided by topology information, characterized in that: include: Obtain incomplete historical measurement data under different topological structure sets and establish measurement data sets; Based on a fully connected neural network, multiple conditional autoencoder models are constructed for missing data of different measurements. The loss function of each conditional autoencoder model is designed, and multiple conditional autoencoder models are jointly trained through the measurement data set. The hyperparameters are adjusted multiple times to obtain the distribution network measurement data completion model guided by the topology structure information. Incomplete measurement data and corresponding topology structures are input into the distribution network measurement data completion model guided by the topology structure information, and the distribution network measurement data completion model guided by the topology structure information outputs complete measurement data.

2. The method for completing distribution network measurement data guided by topology information according to claim 1, characterized in that: The steps of obtaining incomplete historical measurement data under different topological structure sets and establishing a measurement data set include: Using simulation software tools, power flow calculations are performed by changing the topology structure, simulating and obtaining complete historical measurement data under different topology sets, and recording the corresponding topology structure information; The complete historical measurement data under different topological structures is introduced into a sample Binary mask matrix of consistent dimension Calculation Sample With the binary mask matrix The Hadamard product between is used to describe the missing measurement data, and Gaussian distribution noise is superimposed to obtain incomplete historical measurement data that is consistent with the actual situation.

3. The method for completing distribution network measurement data guided by topology information according to claim 2, characterized in that: The simulation software tool is used to perform flow calculations by changing the topology structure, and complete historical measurement data under different topology sets are obtained by simulation. When the corresponding topology structure information is recorded, uncertainty modeling is performed on wind, light, and load. In the uncertainty modeling process, wind power obeys Weibull distribution, light obeys Beta distribution, and load obeys normal distribution.

4. The method for completing distribution network measurement data guided by topology information according to claim 1, characterized in that: The autoencoder model consists of an encoder and a decoder composed of a fully connected neural network; the input layer of the encoder receives missing real-time measurement data and corresponding topological structure information, and the topological structure information is converted into a vector in a latent space as a conditional embedding by designing an embedding layer, and the real-time measurement data and the embedded topological structure information are jointly encoded as latent variables; the input layer of the decoder is used to receive latent variables, and the latent variables generate the completed real-time measurement data under the corresponding topological structure through the decoder.

5. The method for completing distribution network measurement data guided by topology information according to claim 4 is characterized in that: The topological structure information is converted into a vector in a latent space by designing an embedding layer as a conditional embedding, and the steps of jointly encoding the real-time measurement data and the embedded topological structure information into latent variables include: A latent space is introduced, and the input measurement data is mapped to the latent space through the encoder according to the following formula, so that the vector in the latent space contains the characteristics of the input data, and the latent variables of the measurement data are obtained: In the formula, is the output vector of the encoder, i.e., the encoded representation of the incomplete measurement data; is the trainable weight parameter of the encoder; is the bias of the encoder; the above parameters are updated and iterated by the optimizer; is the activation function; The topological structure information is embedded into the autoencoder model as conditional information, and the embedding layer is used to map the topological structure information into a vector in the latent space. The dimension is consistent with the dimension of the input feature, and the latent variable of the topological structure information is obtained: In the formula, Represents the vector of topological information after being encoded by the embedding layer; input_dim represents the number of labels of topological information, reflecting the number of types of topological structures; output_dim represents the dimension of the latent space, that is, the length of the embedded vector, which is the same as the output vector of the encoder. The vector length of is consistent; Embedding represents the operation of the embedding layer, which is used to map each topological structure information into a vector with a length of output_dim; The latent variables of the measurement data and the latent variables of the topological structure information are jointly encoded to form the overall latent variables. The joint encoding is expressed by the following formula: In the formula, A joint encoding vector representing real-time measurement data and topological structure information in the latent space; Represents a vector concatenation operation.

6. The method for completing distribution network measurement data guided by topology information according to claim 5 is characterized in that: In the step of generating the complementary real-time measurement data under the corresponding topological structure through the decoder, the generated joint coding vector is input into the decoder, and the missing reconstruction value of the measurement data is generated through the decoder according to the following formula: In the formula, is the output vector of the decoder, i.e., the reconstructed value; is the trainable weight parameter of the decoder; is the bias of the decoder.

7. The method for completing distribution network measurement data guided by topology information according to claim 1, characterized in that: In the step of designing the loss function of each conditional autoencoder model and jointly training multiple conditional autoencoder models through the measurement data set, different conditional autoencoder models process the missing completion of different measurement data respectively, and after initializing the network parameters of the autoencoder model, the back propagation algorithm is used to train and update the network parameters; The measured data includes node voltage, node injected active power and node injected reactive power. The corresponding loss functions are defined for the node voltage, node injected active power and node injected reactive power conditional autoencoder model as follows: In the formula, , , They are the mean square error (MSE) values ​​when complementing the node voltage, node injected active power, and node injected reactive power, respectively; is the number of samples; , , For sample The true values ​​of node voltage, node injected active power and node injected reactive power; , , For sample The node voltage, node injected active power and node injected reactive power are the complement values; According to the loss function of the designed single conditional autoencoder model, the calculation expression of the total loss function of the model is obtained as follows: In the formula, , , The weights of the autoencoder model for node voltage, node injected active power and node injected reactive power conditions are adaptively determined by the training process according to the difficulty of completing the measurement data; the designed model total loss function is used as the optimization target of the distribution network measurement data completion model guided by the topology structure information, the number of training rounds is set, and the adaptive optimization algorithm is used to update the model parameters to obtain the distribution network measurement data completion model guided by the topology structure information.

8. The method for completing distribution network measurement data guided by topology information according to claim 1, characterized in that: In the step of adjusting the hyperparameters multiple times to obtain the distribution network measurement data completion model guided by the topology information, the hyperparameters include a combination of one or more of a learning rate and a regularization coefficient.

9. A distribution network measurement data completion system guided by topology information, characterized in that: include: The measurement data set acquisition module is used to acquire incomplete historical measurement data under different topological structure sets and establish measurement data sets; An autoencoder model building module, which is used to build multiple conditional autoencoder models for missing data completion of different measurement data based on a fully connected neural network; The joint training module is used to design the loss function of each conditional autoencoder model, and jointly train multiple conditional autoencoder models through the measurement data set, and adjust the hyperparameters multiple times to obtain the distribution network measurement data completion model guided by the topology structure information; The complete measurement data output module is used to input incomplete measurement data and corresponding topology structure into the distribution network measurement data completion model guided by the topology structure information, and the distribution network measurement data completion model guided by the topology structure information outputs complete measurement data.

10. The distribution network measurement data completion system guided by topology information according to claim 9, characterized in that: The measurement data set acquisition module uses a simulation software tool to perform power flow calculations by changing the topological structure, simulate and acquire complete historical measurement data under different topological structure sets, and record the corresponding topological structure information; The complete historical measurement data under different topological structures is introduced into a sample Binary mask matrix of consistent dimension Calculation Sample With the binary mask matrix The Hadamard product between is used to describe the missing measurement data, and Gaussian distribution noise is superimposed to obtain incomplete historical measurement data that is consistent with the actual situation.

11. The distribution network measurement data completion system guided by topology information according to claim 9, characterized in that: The autoencoder model constructed by the autoencoder model construction module consists of an encoder and a decoder composed of a fully connected neural network; the input layer of the encoder receives missing real-time measurement data and corresponding topological structure information, and the topological structure information is converted into a vector in a latent space as a conditional embedding by designing an embedding layer, and the real-time measurement data and the embedded topological structure information are jointly encoded as latent variables; the input layer of the decoder is used to receive latent variables, and the latent variables generate the completed real-time measurement data under the corresponding topological structure through the decoder.

12. The distribution network measurement data completion system guided by topology information according to claim 11, characterized in that: The autoencoder model building module converts the topological structure information into a vector in a latent space by designing an embedding layer as a conditional embedding. In the process of jointly encoding the real-time measurement data and the embedded topological structure information into latent variables, a latent space is introduced. The input measurement data is mapped to the latent space through the encoder according to the following formula, so that the vector in the latent space contains the characteristics of the input data, and the latent variables of the measurement data are obtained: In the formula, is the output vector of the encoder, i.e., the encoded representation of the incomplete measurement data; is the trainable weight parameter of the encoder; is the bias of the encoder; the above parameters are updated and iterated by the optimizer; is the activation function; The topological structure information is embedded into the autoencoder model as conditional information, and the embedding layer is used to map the topological structure information into a vector in the latent space. The dimension is consistent with the dimension of the input feature, and the latent variable of the topological structure information is obtained: In the formula, Represents the vector of topological information after being encoded by the embedding layer; input_dim represents the number of labels of topological information, reflecting the number of types of topological structures; output_dim represents the dimension of the latent space, that is, the length of the embedded vector, which is the same as the output vector of the encoder. The vector length of is consistent; Embedding represents the operation of the embedding layer, which is used to map each topological structure information into a vector with a length of output_dim; The latent variables of the measurement data and the latent variables of the topological structure information are jointly encoded to form the overall latent variables. The joint encoding is expressed by the following formula: In the formula, A joint encoding vector representing real-time measurement data and topological structure information in the latent space; Represents the concatenation operation of vectors; When the latent variable generates the completed real-time measurement data under the corresponding topological structure through the decoder, the generated joint coding vector is input into the decoder, and the missing reconstruction value of the measurement data is generated through the decoder according to the following formula: In the formula, is the output vector of the decoder, i.e., the reconstructed value; is the trainable weight parameter of the decoder; is the bias of the decoder.

13. The distribution network measurement data completion system guided by topology information according to claim 9, characterized in that: The joint training module processes the missing completion of different measurement data by different conditional autoencoder models respectively, and after initializing the network parameters of the autoencoder model, uses the back propagation algorithm to train and update the network parameters; The measured data includes node voltage, node injected active power and node injected reactive power. The corresponding loss functions are defined for the node voltage, node injected active power and node injected reactive power conditional autoencoder model as follows: In the formula, , , They are the mean square error (MSE) values ​​when complementing the node voltage, node injected active power, and node injected reactive power, respectively; is the number of samples; , , For sample The true values ​​of node voltage, node injected active power and node injected reactive power; , , For sample The node voltage, node injected active power and node injected reactive power are the complement values; According to the loss function of the designed single conditional autoencoder model, the calculation expression of the total loss function of the model is obtained as follows: In the formula, , , The weights of the autoencoder model for node voltage, node injected active power and node injected reactive power conditions are adaptively determined by the training process according to the difficulty of completing the measurement data; the designed model total loss function is used as the optimization target of the distribution network measurement data completion model guided by the topology structure information, the number of training rounds is set, and the adaptive optimization algorithm is used to update the model parameters to obtain the distribution network measurement data completion model guided by the topology structure information.

14. An electronic device, characterized in that: It comprises a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement the distribution network measurement data completion method guided by topology structure information as claimed in any one of claims 1 to 8.

15. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, the method for completing distribution network measurement data guided by topology information as claimed in any one of claims 1 to 8 is implemented.

Images