Dispatching method and device for discrete action equipment of power distribution network and nonvolatile storage medium

Abstract

The invention discloses a power distribution network discrete action equipment scheduling method and device and a nonvolatile storage medium. The method comprises the following steps: determining load prediction data of a power distribution network in a preset time period and active output prediction data of a distributed power supply of the power distribution network; a discrete action equipment scheduling model is called to process the load prediction data and the active power output prediction data, so that a scheduling scheme of discrete action equipment in the power distribution network is obtained, and training data of the discrete action equipment scheduling model comprises label data and label-free data; the discrete action equipment scheduling model comprises a feature extractor used for identifying data input into the discrete action equipment scheduling model; and adjusting the tap switching position of the discrete action equipment in a preset time period according to the scheduling scheme. According to the method and the device, the technical problem that the discrete action equipment in the power distribution network cannot be automatically dispatched due to the fact that the output result of a discrete action equipment dispatching model in the related technology is inaccurate is solved.

Description

Technical Field

[0001] The present application relates to the field of electrical engineering, and more specifically, to a method and apparatus for scheduling discrete action equipment in a distribution network, and a non-volatile storage medium. Background Art

[0002] In order to ensure the accuracy of the model scheduling results, the discrete action device scheduling model in the related art needs to rely on a large amount of reliable labeled data to train the model during the model training phase, rather than using unlabeled data to train the model. However, there is often a large amount of unlabeled data in the distribution network, which makes it impossible to ensure the accuracy of the output results of the discrete action device scheduling model in the related art, and thus it is impossible to automatically schedule the discrete action devices in the distribution network.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present application provide a method, apparatus and non-volatile storage medium for scheduling discrete action devices in a distribution network, so as to at least solve the technical problem that the discrete action devices in the distribution network cannot be automatically scheduled due to inaccurate output results of the discrete action device scheduling model in the related art.

[0005] According to one aspect of an embodiment of the present application, a method for scheduling discrete action devices in a distribution network is provided, comprising: determining load forecast data of the distribution network within a preset time period, and active output forecast data of distributed power sources in the distribution network; calling a discrete action device scheduling model to process the load forecast data and the active output forecast data, thereby obtaining a scheduling scheme for discrete action devices in the distribution network, wherein the scheduling scheme includes a predicted switching position of a tap of the discrete action device within a preset time period, the training data of the discrete action device scheduling model includes labeled data and unlabeled data, the discrete action device scheduling model includes a feature extractor for identifying data input into the discrete action device scheduling model, and an estimation module for generating a scheduling scheme; and adjusting the tap switching position of the discrete action device within the preset time period according to the scheduling scheme.

[0006] Optionally, calling a discrete action device scheduling model to process load forecast data and active output forecast data to obtain a scheduling plan for discrete action devices in the distribution network includes: extracting data features from the load forecast data and active output forecast data through a feature extractor in the discrete action device scheduling model, and performing data dimension reduction and global average pooling on the data features to obtain a data feature vector of a preset length; determining a position prediction value mapped by the data feature vector through a fully connected layer through an estimation module in the device scheduling model, wherein the position prediction value is used to indicate a predicted switching position.

[0007] Optionally, the discrete action device scheduling model is trained in the following manner: obtaining historical data of the distribution network, wherein the historical data includes historical load forecast data and historical active output forecast data; determining labeled historical data and unlabeled historical data in the historical data; determining the principal component characteristics of the labeled historical data, and filtering the labeled historical data based on the principal component characteristics to obtain a first training data set; using the principal component characteristics of the labeled historical data as reference principal component characteristics of the unlabeled historical data, and clustering and filtering the unlabeled historical data based on the reference principal component characteristics to obtain a second training data set; and using a generative adversarial network to train the discrete action device scheduling model based on the first training data set and the second training data set.

[0008] Optionally, determining the principal component characteristics of labeled historical data includes: standardizing the labeled data so that the characteristic mean of the labeled data is a first preset value and the variance is a second preset value; determining the covariance matrix of the labeled data after standardization, wherein the principal component characteristics include principal component eigenvectors; the covariance matrix is ​​used to reflect the relationship between each feature quantity in the labeled data; performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors corresponding to the eigenvalues; and determining the principal component eigenvectors in the labeled data based on the size of the eigenvalues.

[0009] Optionally, screening the labeled historical data according to the principal component characteristics to obtain the first training data set includes: determining the correlation between the principal component characteristics and the node voltage characteristics of the distribution network; and screening the labeled data according to the correlation to obtain the first training data set.

[0010] Optionally, clustering and screening the unlabeled historical data according to the reference principal component characteristics to obtain a second training data set includes: standardizing the unlabeled data so that the characteristic mean of the unlabeled data is a first preset value and the variance is a second preset value; constructing a feature space based on the mutual relationship reflected by the covariance matrix and the standardized unlabeled data; determining the principal component space based on the reference principal component, and processing the feature space in the principal component space to obtain a feature representation corresponding to the unlabeled historical data; performing cluster analysis on the feature representation to obtain the second training data set, and verifying in the process of cluster analysis how and to what extent the discrete action device tap position in the unlabeled data affects the voltage control result.

[0011] Optionally, the method also includes: during the training process, determining the cross entropy loss of the discrete action device scheduling model, and determining the loss function value of the discrete action device scheduling model and the gear switching penalty coefficient of the discrete action device in the distribution network based on the cross entropy loss, wherein the gear switching penalty coefficient is used to adjust the loss function value, and the larger the gear switching penalty coefficient, the larger the adjusted loss function value.

[0012] According to another aspect of an embodiment of the present application, a distribution network discrete action device scheduling device is also provided, including: a first processing module, used to determine the load forecast data of the distribution network within a preset time period, and the active output forecast data of the distributed power source of the distribution network; a second processing module, used to call the discrete action device scheduling model to process the load forecast data and the active output forecast data, so as to obtain a scheduling scheme for the discrete action device in the distribution network, wherein the scheduling scheme includes the predicted switching position of the tap of the discrete action device within the preset time period, the training data of the discrete action device scheduling model includes labeled data and unlabeled data, the discrete action device scheduling model includes a feature extractor for identifying data input into the discrete action device scheduling model, and an estimation module for generating a scheduling scheme; a third processing module, used to adjust the tap switching position of the discrete action device within the preset time period according to the scheduling scheme.

[0013] According to another aspect of an embodiment of the present application, a non-volatile storage medium is further provided, in which a program is stored, wherein when the program is running, a device where the non-volatile storage medium is located is controlled to execute a distribution network discrete action device scheduling method.

[0014] According to another aspect of an embodiment of the present application, there is further provided an electronic device, including: a memory and a processor, the processor being configured to run a program stored in the memory, wherein the method for scheduling discrete action devices in a distribution network is executed when the program is run.

[0015] According to another aspect of an embodiment of the present application, a computer program product is also provided, including a computer program, which implements a method for scheduling discrete action devices in a distribution network when executed by a processor.

[0016] In an embodiment of the present application, load forecast data of a distribution network within a preset time period and active output forecast data of distributed power sources of the distribution network are determined; a discrete action device scheduling model is called to process the load forecast data and the active output forecast data, thereby obtaining a scheduling scheme for discrete action devices in the distribution network, wherein the scheduling scheme includes the predicted switching positions of taps of discrete action devices within a preset time period, the training data of the discrete action device scheduling model includes labeled data and unlabeled data, the discrete action device scheduling model includes a feature extractor for identifying data input into the discrete action device scheduling model, and an estimation module for generating a scheduling scheme; a method of adjusting the switching positions of taps of discrete action devices within a preset time period according to the scheduling scheme is adopted, and the discrete action device scheduling model is trained by using labeled data and unlabeled data together, so as to achieve the purpose of ensuring the accuracy of the model output results when the input data is unlabeled data of low quality, thereby achieving the technical effect of automatic scheduling of discrete action devices, and further solving the technical problem that the discrete action devices in the distribution network cannot be automatically scheduled due to inaccurate output results of the discrete action device scheduling model in the related art. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 is a structural schematic diagram of a computer terminal (mobile terminal) according to an embodiment of the present application;

[0019] Figure 2 It is a flow chart of a method for scheduling discrete action equipment in a distribution network provided according to an embodiment of the present application;

[0020] Figure 3 It is a schematic diagram of a training process of a semi-supervised distribution network discrete action device scheduling model provided according to an embodiment of the present application;

[0021] Figure 4 It is a structural schematic diagram of a distribution network discrete action equipment scheduling device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0023] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0024] With the gradual advancement of the "carbon peak and carbon neutrality" goals, the penetration rate of distributed power sources, mainly wind and solar, in the distribution network continues to rise, posing severe challenges to the safe and economical operation of the distribution network. In the previous day-ahead scheduling of discrete action devices, a large amount of high-quality labeled data is required. However, many distribution networks have missing or inaccurate records of historical data, resulting in unreasonable day-ahead scheduling results of the model's discrete action devices. The traditional day-ahead scheduling of discrete action devices in distribution networks is usually highly dependent on a large amount of reliable labeled data. Due to the difficulty in obtaining and maintaining data and the difficulty in determining the quality of data, the reliability of the traditional day-ahead scheduling model for discrete action devices in distribution networks is relatively weak. In addition, since different networks have their own unique load characteristics and operating characteristics, the traditional day-ahead scheduling model for discrete action devices in distribution networks has a fixed mode and poor flexibility.

[0025] To solve these problems, the embodiments of the present application provide a distribution network discrete action equipment day-ahead scheduling model and training method based on a semi-supervised adversarial deep learning method. The semi-supervised adversarial deep learning model has low requirements for labeled data and can effectively solve the data requirements and labeling problems of the traditional distribution network discrete action equipment day-ahead scheduling model. In addition, a historical data evaluation module is added to the proposed distribution network discrete action equipment day-ahead scheduling model based on a semi-supervised adversarial deep learning method, which can filter out high-quality data from a large amount of historical data of varying quality for model training, solving the problem of poor model performance caused by the high dependence of model performance on data quality, which is described in detail below.

[0026] According to an embodiment of the present application, a method embodiment of a method for scheduling discrete action equipment in a distribution network is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0027] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 The hardware structure block diagram of a computer terminal (or mobile device) for implementing a method for dispatching discrete action devices in a distribution network is shown. Figure 1 As shown, the computer terminal 10 (or mobile device 10) may include one or more (102a, 102b, ..., 102n are used to illustrate) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0028] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10 (or mobile device). As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0029] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the method for scheduling discrete action devices of the distribution network in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, the above-mentioned method for scheduling discrete action devices of the distribution network is realized. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0030] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0031] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0032] In the above operating environment, the embodiment of the present application provides a method for scheduling discrete action devices in a distribution network, such as Figure 2 As shown, the method comprises the following steps:

[0033] Step S202, determining load forecast data of the distribution network within a preset time period, and active output forecast data of distributed power sources of the distribution network;

[0034] It should be noted that the day-ahead dispatch of discrete action devices is an important technology implemented in the distribution network, which aims to improve voltage quality and reduce losses. Its goal is to minimize active power loss and voltage offset. The minimum active power loss f1 expected by the system can be expressed as follows:

[0035]

[0036] Where N is the total number of nodes in the distribution network; R ij is the resistance of branch ij; n s is the number of scenes; Iij,t is the current flowing through branch ij during period t under scenario s.

[0037] The expected minimum voltage offset f2 can be expressed as follows:

[0038]

[0039] Where U i,s,t represents the voltage amplitude of node i in time period t under scenario s; U i,max , U i,min , U r They are the voltage upper limit, lower limit and rated voltage of node i respectively.

[0040] In some embodiments of the present application, the following constraints should also be met during the day-ahead scheduling of discrete action devices: active power and reactive power balance constraints; voltage and current constraints; substation outlet power constraints; control variable constraints; capacitor bank and on-load tap-changing transformer action times constraints.

[0041] Step S204, calling the discrete action device scheduling model to process the load forecast data and the active output forecast data, so as to obtain a scheduling plan for the discrete action device in the distribution network, wherein the scheduling plan includes the predicted switching position of the tap of the discrete action device within a preset time period, the training data of the discrete action device scheduling model includes labeled data and unlabeled data, the discrete action device scheduling model includes a feature extractor for identifying data input into the discrete action device scheduling model, and an estimation module for generating a scheduling plan;

[0042] It should be noted that the above-mentioned discrete action device scheduling model is a day-ahead scheduling model for discrete action devices. The model mainly includes two modules: a quality assessment and screening module for historical data and a semi-supervised adversarial deep learning module for day-ahead scheduling. The quality assessment and screening module for historical data is used to screen high-quality data from a large amount of historical data for model training. The trained semi-supervised adversarial deep learning module is used to perform day-ahead scheduling for discrete action devices (on-load tap-changing transformers, shunt capacitor groups) in the distribution network based on the predicted data of loads and distributed power sources. The trained semi-supervised adversarial deep learning module includes a feature extractor and an estimation module.

[0043] In the technical solution provided in step S204, the discrete action device scheduling model is called to process the load forecast data and the active output forecast data, so as to obtain the scheduling scheme of the discrete action device in the distribution network, including: extracting data features from the load forecast data and the active output forecast data through the feature extractor in the discrete action device scheduling model, and performing data dimension reduction processing and global average pooling processing on the data features to obtain a data feature vector of a preset length; determining the position prediction value of the data feature vector mapping through the fully connected layer through the estimation module in the device scheduling model, wherein the position prediction value is used to indicate the predicted switching position.

[0044] As an optional implementation, Figure 3 As shown, the discrete action device scheduling model is trained in the following manner: obtaining historical data of the distribution network, wherein the historical data includes historical load forecast data and historical active output forecast data; determining labeled historical data and unlabeled historical data in the historical data; determining the principal component characteristics of the labeled historical data, and filtering the labeled historical data based on the principal component characteristics to obtain a first training data set; using the principal component characteristics of the labeled historical data as reference principal component characteristics of the unlabeled historical data, and clustering and filtering the unlabeled historical data based on the reference principal component characteristics to obtain a second training data set; and using a generative adversarial network to train the discrete action device scheduling model based on the first training data set and the second training data set.

[0045] In some embodiments of the present application, the step of determining the principal component characteristics of labeled historical data includes: standardizing the labeled data so that the characteristic mean of the labeled data is a first preset value and the variance is a second preset value; determining the covariance matrix of the labeled data after standardization, wherein the principal component characteristics include principal component eigenvectors; the covariance matrix is ​​used to reflect the relationship between each feature quantity in the labeled data; performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors corresponding to the eigenvalues; and determining the principal component eigenvectors in the labeled data based on the size of the eigenvalues.

[0046] As an optional implementation, the step of filtering labeled historical data based on principal component characteristics to obtain a first training data set includes: determining the correlation between the principal component characteristics and the node voltage characteristics of the distribution network; filtering the labeled data based on the correlation to obtain the first training data set.

[0047] Specifically, Figure 3 As shown in the figure, the specific process of filtering labeled historical data is as follows:

[0048] The first step is standardization: standardize the labeled data set so that the mean of each feature is 0 and the variance is 1.

[0049] The second step is to calculate the covariance matrix: the covariance matrix is ​​calculated through the standardized data to obtain the relationship between each feature quantity.

[0050] The third step is eigenvalue decomposition: perform eigenvalue decomposition on the covariance matrix to extract eigenvalues ​​and eigenvectors.

[0051] The fourth step is to select the principal components: filter out the principal components according to the size of the eigenvalues.

[0052] Step 5: Correlation analysis: Analyze the correlation between the principal components and the node voltage characteristics, and select high-quality historical data as labeled data sets for training.

[0053] As an optional implementation, when determining the correlation between the principal component features and the node voltage features of the distribution network, the Pearson correlation coefficient between the principal component features and the node voltage features of the distribution network can be calculated to quantify the degree of association between each principal component feature and the node voltage feature. Then a Pearson correlation coefficient threshold can be set, and labeled data above this threshold is retained in the first training data set. Data with higher correlation can be considered to have higher data quality.

[0054] In some embodiments of the present application, clustering and screening the unlabeled historical data based on the reference principal component characteristics to obtain the second training data set includes: standardizing the unlabeled data so that the characteristic mean of the unlabeled data is a first preset value and the variance is a second preset value; constructing a feature space based on the mutual relationship reflected by the covariance matrix and the standardized unlabeled data; determining the principal component space based on the reference principal component, and processing the feature space in the principal component space to obtain a feature representation corresponding to the unlabeled historical data; performing cluster analysis on the feature representation to obtain the second training data set, and verifying the influence of the discrete action device tap position in the unlabeled data on the voltage control result and the degree of influence during the cluster analysis process.

[0055] Specifically, Figure 3 As shown in Figure 1, the screening process of unlabeled data is as follows:

[0056] The first step is standardization: standardize the unlabeled data set so that the mean of each feature is 0 and the variance is 1.

[0057] The second step is to determine the reference principal component: obtain the principal component analysis results from the labeled dataset.

[0058] The third step is feature relationship mapping: map the feature relationships obtained in the labeled dataset to the unlabeled dataset to form a new feature space.

[0059] The fourth step is to project the data into the principal component space: the data in the unlabeled dataset is analyzed in the principal component space formed by the supervised dataset to obtain a new feature representation.

[0060] Step 5: Cluster analysis: Cluster analysis is performed on the new feature representation to verify the effectiveness of discrete action device tap positions in unlabeled datasets in voltage control.

[0061] Specifically, the covariance matrix can be calculated using the standardized unlabeled data, and then the covariance matrix can be decomposed to obtain the eigenvalues ​​and eigenvectors corresponding to the eigenvalues. The eigenvectors are then sorted and screened according to the size of the eigenvalues, and the eigenvectors with a cumulative variance of 90% are selected as the main features. The unlabeled data can then be projected onto the selected principal component direction to obtain the feature representation after dimensionality reduction.

[0062] In some embodiments of the present application, the function of the above-mentioned semi-supervised adversarial deep learning module is to perform day-ahead scheduling of discrete action equipment based on the load forecast data of the 24 hours before the day and the active output forecast data of the distributed power source of the 24 hours before the day. The semi-supervised adversarial deep learning module is mainly composed of two parts: semi-supervised deep learning and adversarial generative network. Semi-supervised deep learning is the main part of the model, which includes a feature extractor for identifying the input day-ahead forecast data and an estimation module for estimating the tap switching position of the discrete action equipment; the adversarial generative network is to enhance the feature extraction capability of the feature extractor and the generalization capability of the model. The specific process of the operation of the semi-supervised deep learning module is as follows: Figure 3 As shown, the following steps are included:

[0063] The first step is to load and preprocess the data. From the reasonable data set selected by the quality assessment and screening module of historical data, the data is divided into labeled data set and unlabeled data set. Data processing is performed to map the tap positions of discrete action devices to the positive integer range, and these integer labels are converted into one-hot encoding format to adapt to the neural network classification task.

[0064] In the second step, the feature extractor is used to extract data features, and three layers of one-dimensional convolutional layers are used to extract the time series features in the input data and then perform data dimensionality reduction. Finally, the features are mapped into a vector of fixed length through a global average pooling layer.

[0065] The third step is to pre-train the model using labeled data. In each iteration, the feature extractor converts the input data into a feature vector, and the estimation module maps the feature vector to the estimated position of the discrete action device through a fully connected layer. Then, the loss and the penalty for gear switching are calculated based on the cross entropy loss.

[0066] The operation process of the adversarial generation network is as follows Figure 3 As shown, the following steps are included:

[0067] The first step is to train the discriminator, which is composed of a fully connected neural network. The data features extracted by the feature extraction module are input into the discriminator, and the discriminator outputs a probability indicating whether the data comes from real data or a generated network.

[0068] The second step is to use unlabeled data and labeled data to train the feature extractor against the discriminator. The feature extractor is responsible for extracting data features from unlabeled data and labeled data, and the discriminator is responsible for judging whether the data comes from unlabeled data or labeled data based on the extracted features. During the training process, the discriminator loss is calculated and the discriminator parameters are updated through back propagation, ultimately achieving the goal of improving the feature extraction capability of the feature extractor.

[0069] As an optional implementation, during the training process, the cross entropy loss of the discrete action device scheduling model can also be determined, and the loss function value of the discrete action device scheduling model and the gear switching penalty coefficient of the discrete action device in the distribution network can be determined based on the cross entropy loss, wherein the gear switching penalty coefficient is used to adjust the loss function value, and the larger the gear switching penalty coefficient, the larger the adjusted loss function value.

[0070] Specifically, during the training process, the action status of the discrete action device corresponding to the output result of the model can be determined according to the output result of the model, and then when the action status meets the constraint conditions or is far away from the control target, a larger penalty coefficient is given to the model to increase the loss function value of the model.

[0071] Step S206: adjusting the tap switching position of the discrete action device within a preset time period according to the scheduling plan.

[0072] By determining the load forecast data of the distribution network within a preset time period and the active output forecast data of the distributed power sources of the distribution network; calling the discrete action device scheduling model to process the load forecast data and the active output forecast data, a scheduling plan for the discrete action devices in the distribution network is obtained, wherein the scheduling plan includes the predicted switching positions of the taps of the discrete action devices within the preset time period, the training data of the discrete action device scheduling model includes labeled data and unlabeled data, the discrete action device scheduling model includes a feature extractor for identifying data input into the discrete action device scheduling model, and an estimation module for generating a scheduling plan; the tap switching positions of the discrete action devices within the preset time period are adjusted according to the scheduling plan, and the discrete action device scheduling model is trained by using labeled data and unlabeled data together, so as to ensure that the accuracy of the model output results can still be guaranteed when the input data is low-quality unlabeled data, thereby achieving the technical effect of automatic scheduling of discrete action devices, and further solving the technical problem that the discrete action devices in the distribution network cannot be automatically scheduled due to inaccurate output results of the discrete action device scheduling model in the related technology.

[0073] In addition, the distribution network discrete action model scheduling method provided in the embodiment of the present application uses a semi-supervised deep learning algorithm, which reduces the demand for labeled data and uses a large amount of unlabeled data supplemented by a small amount of labeled data for training, which is more suitable for large-scale distribution network day-ahead scheduling with a large amount of data. Moreover, the adversarial generative network is integrated in the model training process, and the model has a stronger feature extraction capability, which is more suitable for distribution networks with complex and changeable operating conditions.

[0074] The present application provides a distribution network discrete action equipment scheduling device, Figure 4 is a schematic diagram of the structure of the device. Figure 4 It can be seen that the device includes: a first processing module 40, which is used to determine the load forecast data of the distribution network within a preset time period, and the active output forecast data of the distributed power source of the distribution network; a second processing module 42, which is used to call the discrete action device scheduling model to process the load forecast data and the active output forecast data, so as to obtain the scheduling plan of the discrete action device in the distribution network, wherein the scheduling plan includes the predicted switching position of the tap of the discrete action device within the preset time period, the training data of the discrete action device scheduling model includes labeled data and unlabeled data, the discrete action device scheduling model includes a feature extractor for identifying data input into the discrete action device scheduling model, and an estimation module for generating a scheduling plan; a third processing module 44, which is used to adjust the tap switching position of the discrete action device within the preset time period according to the scheduling plan.

[0075] In some embodiments of the present application, the second processing module 42 calls the discrete action device scheduling model to process the load forecast data and the active output forecast data, thereby obtaining the scheduling plan for the discrete action device in the distribution network, including: extracting data features from the load forecast data and the active output forecast data through the feature extractor in the discrete action device scheduling model, and performing data dimension reduction and global average pooling on the data features to obtain a data feature vector of a preset length; determining the position prediction value of the data feature vector mapping through the fully connected layer through the estimation module in the device scheduling model, wherein the position prediction value is used to indicate the predicted switching position.

[0076] In some embodiments of the present application, a discrete action device scheduling model is trained in the following manner: historical data of the distribution network is obtained, wherein the historical data includes historical load forecast data and historical active output forecast data; labeled historical data and unlabeled historical data in the historical data are determined; principal component characteristics of the labeled historical data are determined, and the labeled historical data are screened based on the principal component characteristics to obtain a first training data set; the principal component characteristics of the labeled historical data are used as reference principal component characteristics of the unlabeled historical data, and the unlabeled historical data are clustered and screened based on the reference principal component characteristics to obtain a second training data set; and a generative adversarial network is used to train the discrete action device scheduling model based on the first training data set and the second training data set.

[0077] In some embodiments of the present application, determining the principal component features of labeled historical data includes: standardizing the labeled data so that the feature mean of the labeled data is a first preset value and the variance is a second preset value; determining the covariance matrix of the labeled data after standardization, wherein the principal component features include principal component eigenvectors; the covariance matrix is ​​used to reflect the relationship between each feature quantity in the labeled data; performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors corresponding to the eigenvalues; and determining the principal component eigenvectors in the labeled data based on the size of the eigenvalues.

[0078] In some embodiments of the present application, the step of filtering labeled historical data based on principal component characteristics to obtain a first training data set includes: determining the correlation between the principal component characteristics and the node voltage characteristics of the distribution network; and filtering the labeled data based on the correlation to obtain the first training data set.

[0079] In some embodiments of the present application, clustering and screening the unlabeled historical data based on the reference principal component characteristics to obtain the second training data set includes: standardizing the unlabeled data so that the characteristic mean of the unlabeled data is a first preset value and the variance is a second preset value; constructing a feature space based on the mutual relationship reflected by the covariance matrix and the standardized unlabeled data; determining the principal component space based on the reference principal component, and processing the feature space in the principal component space to obtain a feature representation corresponding to the unlabeled historical data; performing cluster analysis on the feature representation to obtain the second training data set, and verifying the influence of the discrete action device tap position in the unlabeled data on the voltage control result and the degree of influence during the cluster analysis process.

[0080] In some embodiments of the present application, the distribution network discrete action equipment scheduling device is also used to: during the training process, determine the cross entropy loss of the discrete action equipment scheduling model, and determine the loss function value of the discrete action equipment scheduling model based on the cross entropy loss, as well as the gear switching penalty coefficient of the discrete action equipment in the distribution network, wherein the gear switching penalty coefficient is used to adjust the loss function value, and the larger the gear switching penalty coefficient, the larger the adjusted loss function value.

[0081] It should be noted that the various modules in the above-mentioned distribution network discrete action equipment scheduling device can be a program module (for example, a set of program instructions to implement a certain specific function) or a hardware module. For the latter, it can be expressed in the following forms, but is not limited to this: the expression form of each of the above-mentioned modules is a processor, or the functions of each of the above-mentioned modules are implemented by a processor.

[0082] According to an embodiment of the present application, a non-volatile storage medium is also provided, in which a program is stored, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute the following distribution network discrete action device scheduling method: determine the load forecast data of the distribution network within a preset time period, and the active output forecast data of the distributed power sources of the distribution network; call the discrete action device scheduling model to process the load forecast data and the active output forecast data, so as to obtain a scheduling plan for the discrete action devices in the distribution network, wherein the scheduling plan includes the predicted switching position of the tap of the discrete action device within the preset time period, the training data of the discrete action device scheduling model includes labeled data and unlabeled data, the discrete action device scheduling model includes a feature extractor for identifying data input into the discrete action device scheduling model, and an estimation module for generating a scheduling plan; adjust the tap switching position of the discrete action device within the preset time period according to the scheduling plan.

[0083] According to an embodiment of the present application, an electronic device is also provided, comprising: a memory and a processor, the processor being used to run a program stored in the memory, wherein when the program is run, the following distribution network discrete action device scheduling method is executed: determining the load forecast data of the distribution network within a preset time period, and the active output forecast data of the distributed power sources of the distribution network; calling the discrete action device scheduling model to process the load forecast data and the active output forecast data, so as to obtain a scheduling scheme for discrete action devices in the distribution network, wherein the scheduling scheme includes the predicted switching position of the tap of the discrete action device within the preset time period, the training data of the discrete action device scheduling model includes labeled data and unlabeled data, the discrete action device scheduling model includes a feature extractor for identifying data input into the discrete action device scheduling model, and an estimation module for generating a scheduling scheme; adjusting the tap switching position of the discrete action device within the preset time period according to the scheduling scheme.

[0084] According to an embodiment of the present application, a computer program product is also provided, including a computer program, which implements the following distribution network discrete action device scheduling method when executed by a processor: determining the load forecast data of the distribution network within a preset time period, and the active output forecast data of the distributed power sources of the distribution network; calling the discrete action device scheduling model to process the load forecast data and the active output forecast data, so as to obtain a scheduling plan for the discrete action devices in the distribution network, wherein the scheduling plan includes the predicted switching position of the tap of the discrete action device within the preset time period, the training data of the discrete action device scheduling model includes labeled data and unlabeled data, the discrete action device scheduling model includes a feature extractor for identifying data input into the discrete action device scheduling model, and an estimation module for generating a scheduling plan; adjusting the tap switching position of the discrete action device within the preset time period according to the scheduling plan.

[0085] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0086] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0087] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0088] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0089] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the relevant technology or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc. Various media that can store program codes.

[0090] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for dispatching discrete action equipment in a distribution network, characterized in that: include: Determine load forecast data of the distribution network within a preset time period, and active output forecast data of distributed power sources of the distribution network; Calling a discrete action device scheduling model to process the load forecast data and the active output forecast data, thereby obtaining a scheduling scheme for discrete action devices in the distribution network, wherein the scheduling scheme includes the predicted switching position of the tap of the discrete action device within the preset time period, the training data of the discrete action device scheduling model includes labeled data and unlabeled data, the discrete action device scheduling model includes a feature extractor for identifying data input into the discrete action device scheduling model, and an estimation module for generating the scheduling scheme; The tap switching position of the discrete action device within the preset time period is adjusted according to the scheduling scheme.

2. The method for dispatching discrete action equipment in a distribution network according to claim 1, characterized in that: Calling a discrete action device scheduling model to process the load forecast data and the active output forecast data, thereby obtaining a scheduling scheme for discrete action devices in the distribution network includes: Extracting data features from the load forecast data and the active output forecast data through a feature extractor in the discrete action device scheduling model, and performing data dimension reduction processing and global average pooling processing on the data features to obtain a data feature vector of a preset length; The position prediction value of the data feature vector mapping is determined by the estimation module in the equipment scheduling model through a fully connected layer, wherein the position prediction value is used to indicate the predicted switching position.

3. The method for dispatching discrete action equipment in a distribution network according to claim 1, characterized in that: The discrete action device scheduling model is trained in the following way: Acquiring historical data of the distribution network, wherein the historical data includes historical load forecast data and historical active output forecast data; Determine labeled historical data and unlabeled historical data in the historical data; Determine the principal component characteristics of the labeled historical data, and filter the labeled historical data according to the principal component characteristics to obtain a first training data set; Using the principal component features of the labeled historical data as reference principal component features of the unlabeled historical data, and performing clustering screening on the unlabeled historical data according to the reference principal component features to obtain a second training data set; The discrete action device scheduling model is trained using a generative adversarial network based on the first training data set and the second training data set.

4. The method for dispatching discrete action equipment in a distribution network according to claim 3, characterized in that: Determining the principal component features of the labeled historical data includes: Standardizing the labeled data so that a feature mean of the labeled data is a first preset value and a variance is a second preset value; Determine a covariance matrix of the labeled data after standardization, wherein the principal component features include principal component eigenvectors; the covariance matrix is ​​used to reflect the relationship between each feature quantity in the labeled data; Performing eigenvalue decomposition processing on the covariance matrix to obtain eigenvalues ​​and eigenvectors corresponding to the eigenvalues; The principal component eigenvector in the labeled data is determined according to the size of the eigenvalue.

5. The method for dispatching discrete action equipment in a distribution network according to claim 3, characterized in that: The labeled historical data is screened according to the principal component features to obtain a first training data set including: Determining the correlation between the principal component characteristics and the node voltage characteristics of the distribution network; The labeled data is screened according to the correlation to obtain the first training data set.

6. The method for dispatching discrete action equipment in a distribution network according to claim 5, characterized in that: The unlabeled historical data is clustered and screened according to the reference principal component features to obtain a second training data set including: Standardizing the unlabeled data so that a feature mean of the unlabeled data is a first preset value and a variance is a second preset value; Constructing a feature space based on the mutual relationship embodied by the covariance matrix and the standardized unlabeled data; Determining a principal component space according to the reference principal component, and processing the feature space in the principal component space to obtain a feature representation corresponding to the unlabeled historical data; The feature representation is subjected to cluster analysis processing to obtain the second training data set, and in the process of cluster analysis processing, the influence mode and influence degree of the discrete action device tap position in the unlabeled data on the voltage control result are verified.

7. The method for dispatching discrete action equipment in a distribution network according to claim 3, characterized in that: The method further comprises: During the training process, the cross entropy loss of the discrete action device scheduling model is determined, and the loss function value of the discrete action device scheduling model and the gear switching penalty coefficient of the discrete action device in the distribution network are determined based on the cross entropy loss, wherein the gear switching penalty coefficient is used to adjust the loss function value, and the larger the gear switching penalty coefficient, the larger the adjusted loss function value.

8. A distribution network discrete action equipment dispatching device, characterized in that: include: A first processing module, used to determine load forecast data of the distribution network within a preset time period, and active output forecast data of the distributed power sources of the distribution network; A second processing module is used to call a discrete action device scheduling model to process the load forecast data and the active output forecast data, so as to obtain a scheduling scheme for discrete action devices in the distribution network, wherein the scheduling scheme includes the predicted switching position of the tap of the discrete action device within the preset time period, the training data of the discrete action device scheduling model includes labeled data and unlabeled data, the discrete action device scheduling model includes a feature extractor for identifying data input into the discrete action device scheduling model, and an estimation module for generating the scheduling scheme; The third processing module is configured to adjust the tap switching position of the discrete action device within the preset time period according to the scheduling scheme.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a program, wherein when the program is executed, the device where the non-volatile storage medium is located is controlled to execute the method for scheduling discrete action devices in a distribution network according to any one of claims 1 to 7.

10. An electronic device, characterized in that: include: A memory and a processor, wherein the processor is used to run a program stored in the memory, wherein the program, when running, executes the distribution network discrete action device scheduling method described in any one of claims 1 to 7.

11. A computer program product, characterized in that It comprises a computer program, which, when executed by a processor, implements the method for dispatching discrete action devices in a distribution network according to any one of claims 1 to 7.

Images