Distributed power supply type checking method and device and storage medium

The convolutional neural network algorithm uses the power generation and use data of distributed power supply, which solves the problems of low human verification efficiency and data pollution in the existing technology, and realizes efficient and accurate verification of distributed power supply types, which improves the power grid Safety and reliability.

CN118916763APending Publication Date: 2024-11-08STATE GRID LIAONING ELECTRIC POWER CO LTD +5
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411055811.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-02
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing distributed power type calibration method relies on human-based on-site calibration, which is inefficient and difficult to be accurate, and calibration based on individual user data is easily contaminated by data, resulting in poor calibration accuracy.

Method used

The convolutional neural network algorithm is used to process big data. By collecting and processing power generation and consumption data of distributed power supplies, missing values ​​are filled, class normalized and data enhancement are carried out, and the convolutional neural network model is input to improve the verification accuracy.

Benefits of technology

Remotely accurate calibration of self-provided power types is achieved, calibration efficiency and accuracy is improved, manpower visits are reduced, and power grid safety and reliability are enhanced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118916763A_ABST
    Figure CN118916763A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of electrical detection, and particularly relates to a distributed power supply type checking method and device and a storage medium. The method comprises the following steps: collecting power generation and utilization data and label data of a distributed power supply; carrying out power generation and utilization data missing value processing according to the collected data; performing class normalization processing on the processed data, and constructing a data enhancement matrix; and inputting the data subjected to class normalization processing into a convolutional neural network model for checking until a checking result with increased accuracy is obtained. Based on the convolutional neural network, big data is used as the input of the model, the influence of data pollution of individual self-contained power supply users is avoided, the data is processed by using the deep learning algorithm of the convolutional neural network, and the checking accuracy is improved while the checking efficiency is improved. Therefore, the accuracy of the distributed power supply type checking result is obviously improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of electrical detection, and in particular, relates to a distributed power supply type verification method, device and storage medium, and more specifically, to a distributed power supply type verification method, device and storage medium based on a convolutional neural network algorithm. Background Art

[0002] With the comprehensive development of the construction of new power systems, the load-side active trend of the power grid has become more and more obvious. While the customer-side distributed photovoltaic, distributed wind power, electric vehicle charging facilities, distributed energy storage, self-contained power plants and other facilities are widely connected, there are problems such as power generation and consumption units and individuals privately changing the type of power generation equipment, expanding or reducing power generation capacity, which brings new challenges to the power grid in terms of power dispatching. The connection of a large number of power sources that do not match the registration files makes it more difficult to predict the power generation, the quality of power supply and safety hazards control more difficult, and the risk of safe power use on the power user side increases.

[0003] Distributed power sources are devices that are distributed on the user side and can transmit electricity to the grid to meet the needs of power supply, support peak load shifting of the power grid, or meet the requirements of both at the same time. Distributed power sources include: distributed photovoltaic power generation, wind power stations, electric vehicle charging stations, distributed energy storage, gas turbines, diesel generators, etc. According to user needs, distributed power sources can be used as backup power sources, realize new energy consumption, and provide power to isolated islands in remote areas.

[0004] Therefore, in order to ensure the safe and reliable operation of the power grid, it is necessary to monitor, warn and verify the power generation and consumption of distributed power sources. The traditional verification method of distributed power source types mainly relies on on-site verification by power grid operation and maintenance personnel. This verification method requires power grid company operation and maintenance personnel to go deep into the location of distributed power sources, which is greatly restricted by geographical space and time. It not only consumes a lot of manpower and material resources, but also has a limited coverage of distributed power sources.

[0005] The distributed power source calibration technology based on the user power generation and consumption data uploaded by the smart gateway energy meter is an emerging method. By using the user-side smart gateway energy meter, the power grid company can obtain the power generation and consumption data of active customers around the clock, instead of visiting active customers to obtain data manually; and the algorithm-based distributed power source type calibration technology is to automatically identify and alarm the type of distributed power source by transmitting the collected power generation and consumption data to the cloud of the monitoring center using the data monitoring system of the monitoring center. Compared with the traditional on-site calibration method of operation and maintenance personnel, these calibration methods are more convenient, efficient, real-time and economical.

[0006] At present, the existing algorithms are mainly numerical methods and machine learning algorithms, which are difficult to fully utilize big data and have poor calibration accuracy. If only the data of a single user's smart gateway electricity meter is used as the input for calibration, it is easy to be contaminated by data, resulting in calibration errors. Summary of the invention

[0007] In view of the shortcomings of the above-mentioned prior art, the present invention provides a distributed power type verification method, device and storage medium. The purpose is to achieve remote and accurate verification of the type verification efficiency of self-provided power supply. This method uses big data as the input of the model to avoid the influence of data pollution of individual self-provided power supply users, and uses the deep learning algorithm of convolutional neural network to process data, which improves the verification efficiency and the accuracy of the verification.

[0008] The technical solution adopted by the present invention to achieve the above-mentioned purpose is:

[0009] A distributed power source type verification method comprises the following steps:

[0010] Collect power generation and consumption data and label data of distributed power sources;

[0011] According to the collected data, the missing values ​​of power generation and consumption data are processed;

[0012] Normalize the processed data and construct a data enhancement matrix;

[0013] The class-normalized data is input into the convolutional neural network model for verification until a verification result with improved accuracy is obtained.

[0014] Furthermore, the power generation and consumption data of the distributed power source include: only distributed photovoltaic is configured in the distributed photovoltaic power source, only electric vehicle charging piles are configured in the power source, only self-contained power plants are configured in the power source, distributed photovoltaic and electric vehicle charging piles are configured in the power source at the same time, distributed photovoltaic and self-contained power plants are configured in the power source at the same time, electric vehicle charging piles and self-contained power plants are configured in the power source at the same time, and distributed photovoltaic, electric vehicle charging piles and self-contained power plants are configured in the power source at the same time; the label data refers to labeling the power generation and consumption data of different types of distributed power sources with corresponding labels, and collecting row data based on these labels.

[0015] Furthermore, the processing of missing values ​​of power generation and consumption data based on the collected data includes:

[0016] If the distributed generation source is missing only one power generation and consumption data, the general interpolation method of power numbers is used to fill in the missing data;

[0017] If the distributed generation is missing 2 to 4 power generation and consumption data, the power data forward / backward interpolation method is used to fill in the missing data;

[0018] If the distributed power generation is missing more than 4 power generation and consumption data, the power data for that day will be invalidated.

[0019] Furthermore, if the distributed power source is missing only one power generation and consumption data, the missing data is supplemented by using the general interpolation method of power data, including:

[0020] Interpolation is performed on 6 data points, 15 minutes before and after the missing value, to fill in the missing data, as shown in the following formula:

[0021]

[0022]

[0023] In the above formula: For missing values ​​of electricity data The regression function of Complete the formula for missing values; is the maximum allowable calibration deviation remainder; is the first-order difference quotient, is the second-order difference quotient, is the sixth-order difference quotient; f is the item completion function in the missing value completion formula, Sorting missing values ​​in electricity data for time The third power data before, Sorting missing values ​​in electricity data for time The second power data before, Sorting missing values ​​in electricity data for time The third power data after that;

[0024] If the distributed power source is missing only 2 to 4 power generation and consumption data, the missing data is supplemented by the forward / backward interpolation method of power data, which means that the earliest / latest missing data in time is supplemented by the previous / latent data closest to the moment, and then the remaining data is supplemented in sequence by the interpolation method. The formula is as follows:

[0025]

[0026] In the above two formulas, is the data obtained by backward interpolation, is the data obtained by forward interpolation, is the 6th order backward difference, Δ 6 f6 is the sixth-order forward difference, γ is the standard elimination term for distributed power sources, which is 1 when the power generation and consumption data meet the requirements of the power company, and 0 when they do not meet the requirements, f6 is the regression function constructed by the backward interpolation method, f0 is the regression function constructed by the forward interpolation method, and t is the time series value of the missing power value;

[0027] If the distributed power generation is missing more than 4 power generation and consumption data, the power data for this day will be invalidated, where:

[0028] The data obtained by interpolation fills the matrix X to obtain the matrix X′ that conforms to the algorithm network. The formula is as follows:

[0029]

[0030] In the above formula, it is assumed that the original matrix X Missing values, missing values ​​filled by downward interpolation is the data obtained by backward interpolation, It is the missing value of electricity data.

[0031] Furthermore, the processed data is normalized and a data enhancement matrix is ​​constructed, including:

[0032] The class normalization method is adopted to eliminate the influence of the dimension difference of distributed power sources of different scales on the verification process. The expression is as follows:

[0033]

[0034] In the above formula: Distributed power type DR n The original measured data of power generation and consumption in the tth time period, The missing value of electricity The value after class normalization;

[0035] Then substitute the data enhanced values ​​into the matrix X′ to obtain the data enhanced matrix X″, the formula is as follows:

[0036]

[0037] In the above formula: R nm×288 is a real number matrix of size nm×288 composed of power data after data enhancement; when the convolutional neural network uses the class-normalized data input model to obtain the verification result, the capacity of each distributed power type configuration is obtained by anti-class normalization, and the expression is as follows:

[0038]

[0039] In the above formula: is the actual power generation and consumption prediction value obtained after inverse class normalization, The power DR after class normalization n The power generation and consumption forecast value of the type in the tth time period, is the error deviation, indicating the error range within which the calibration is allowed. The values ​​obtained by denormalization form the feature matrix X″′, X″′ extracted by CNN, as shown in the following formula:

[0040]

[0041] Furthermore, the class-normalized data is input into the convolutional neural network model for verification until a verification result with an increased accuracy is obtained, wherein: the convolutional neural network is set using the obtained data set, using 22 pooling layers of size [288, 1, 1] as the feature extraction layer, after 22 pooling layers of size [144, 1, 1], Relu activation is performed, a maximum pooling layer is added, and 44 convolution layers of size [88, 1, 1], 44 pooling layers of size [24, 1, 1], Relu activation and maximum pooling layers are added in sequence;

[0042] The number of convolutional neural network iterations was set to 864, the learning rate was 0.001007, and the error threshold was 1.125×10 -4 When the number of iterations reaches 764, the learning rate is reduced to 0.00052, and the error threshold is set to 0.610×10 -4 In addition, the Adam gradient descent algorithm is used to optimize each pooling layer in the convolutional neural network;

[0043] The output of the convolutional neural network is 7 fully connected layers. The verification result is decoded by the anti-class normalization formula, and the final verified distributed power type is output. The feature matrix extracted from CNN is stored in the server.

[0044] A distributed power source type verification device, comprising:

[0045] The acquisition module is used to collect the power generation and consumption data and label data of the distributed power source;

[0046] A missing value processing module is used to process missing values ​​of power generation and consumption data based on the collected data;

[0047] The class normalization processing module is used to perform class normalization on the processed data and construct a data enhancement matrix;

[0048] A verification module is used to input the data after class normalization into the convolutional neural network model for verification until a verification result with improved accuracy is obtained;

[0049] The device is used to implement the steps of any one of the methods for verifying the type of distributed power supply.

[0050] Furthermore, the processing of missing values ​​of power generation and consumption data based on the collected data includes:

[0051] If the distributed generation source is missing only one power generation and consumption data, the general interpolation method of power numbers is used to fill in the missing data;

[0052] If the distributed generation is missing 2 to 4 power generation and consumption data, the power data forward / backward interpolation method is used to fill in the missing data;

[0053] If the distributed power generation is missing more than 4 power generation and consumption data, the power data for that day will be invalidated;

[0054] The processed data is normalized and a data enhancement matrix is ​​constructed, including:

[0055] The class normalization method is adopted to eliminate the influence of the dimension difference of distributed power sources of different scales on the verification process. The expression is as follows:

[0056]

[0057] In the above formula: Distributed power type DR n The original measured data of power generation and consumption in the tth time period, The missing value of electricity After class normalization, substitute the data enhancement values ​​into the matrix X′ to obtain the data enhancement matrix X″, the formula is as follows:

[0058]

[0059] In the above formula: R nm×288 is a real number matrix of size nm×288 composed of power data after data enhancement; when the convolutional neural network uses the class-normalized data input model to obtain the verification result, the capacity of each distributed power type configuration is obtained by anti-class normalization, and the expression is as follows:

[0060]

[0061] In the above formula: is the actual power generation and consumption prediction value obtained after inverse class normalization, The power DR after class normalization n The power generation and consumption forecast value of the type in the tth time period, is the error deviation, indicating the error range within which the calibration is allowed. The values ​​obtained by denormalization form the feature matrix X″′, X″′ extracted by CNN, as shown in the following formula:

[0062]

[0063] A computer device comprises a storage medium, a processor and a computer program stored on the storage medium and executable on the processor, wherein when the processor executes the computer program, the steps of any one of the methods for verifying the type of distributed power sources are implemented.

[0064] A computer storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any one of the methods for verifying the type of distributed power supply.

[0065] The present invention has the following beneficial effects and advantages:

[0066] The present invention is based on Convolutional Neural Networks (CNN), which is a currently widely used deep learning algorithm. It is currently commonly used in computer vision fields such as automatic driving and face recognition. Its essence is a classification processing method. Distributed power supply verification mainly involves a variety of power supply types, and different power supply types have their own unique power generation characteristics. Therefore, verification is essentially a multi-classification problem. Convolutional neural networks have extremely excellent data characterization capabilities, so the accuracy of the distributed power supply type verification results of the present invention is significantly improved.

[0067] The present invention modifies the normalization in the convolutional neural network of the existing invention to become a class normalization with data enhancement effect suitable for the power characteristics of the self-contained power supply, and pre-places the class normalization in the convolutional neural network model to improve the type verification efficiency of the self-contained power supply of the present invention.

[0068] In the convolutional neural network architecture, we add the Adam algorithm between each convolution layer for optimization. Compared with the traditional Adam optimization only after all pooling layers, this method can accelerate the convergence of the algorithm and improve the algorithm speed. At the same time, the hidden layer in the traditional convolutional neural network is a 16 / 32-layer architecture, which is suitable for image recognition, while this method improves the hidden layer to a 22 / 44-layer architecture, which is more suitable for 288 time node power data. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0070] Figure 1 is a flow chart of the present invention;

[0071] Figure 2 yes Figure 1 Flowchart of step 2 in ;

[0072] Figure 3 yes Figure 1 Flowchart of step 4 in ;

[0073] Figure 4 It is a schematic diagram of the result of the verification of the present invention. DETAILED DESCRIPTION

[0074] In order to more clearly understand the above-mentioned objectives, features and advantages of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0075] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0076] Refer to the following Figure 1-Figure 4 The technical solutions of some embodiments of the present invention are described.

[0077] Example 1

[0078] The present invention provides an embodiment, which is a distributed power source type verification method.

[0079] When implemented specifically, the present invention comprises the following steps:

[0080] Step 1. Collect power generation and consumption data and tag data of distributed power sources.

[0081] like Figure 1 As shown, Figure 1 It is a flow chart of the present invention. Data preparation, collecting power generation and consumption data of seven types of distributed power sources, namely power data, including: only distributed photovoltaic is configured in the distributed photovoltaic power source, only electric vehicle charging piles are configured in the power source, only self-provided power plant is configured in the power source, distributed photovoltaic and electric vehicle charging piles are configured in the power source at the same time, distributed photovoltaic and self-provided power plant are configured in the power source at the same time, electric vehicle charging piles and self-provided power plant are configured in the power source at the same time, and distributed photovoltaic, electric vehicle charging piles and self-provided power plant are configured in the power source at the same time.

[0082] Specifically, the power generation and consumption data collected in step 1, that is, the data collected every 5 minutes from the smart gateway electricity meter in the park, is used to collect power user data and convert it into a distributed power generation and consumption data verification matrix.

[0083] The 5-minute power generation and consumption data collected by the distributed power source gateway electric energy meter in the park will be used, with a total of 288 nodes, to form the matrix X as shown in formula (1):

[0084]

[0085] In the formula, DRn There are seven types of distributed power sources, among which DR1 represents distributed photovoltaic, DR2 represents electric vehicle charging piles, DR3 represents self-owned power plants, DR4 represents both distributed photovoltaic and electric vehicle charging piles, DR5 represents both distributed photovoltaic and self-owned power plants, DR6 represents both electric vehicle charging piles and self-owned power plants, and DR7 represents distributed photovoltaic, electric vehicle charging piles and self-owned power plants; P is the power of the distributed power source at a time node; m is the time series of the distributed power source at the time node.

[0086] Collecting label data means labeling the power generation and consumption data of the above-mentioned different types of distributed power sources with corresponding labels, and collecting data according to the labels. Among them, the power source is configured with only distributed photovoltaic power generation as 1, the power source is configured with only electric vehicle charging piles as 2, the power source is configured with only self-provided power plants as 3, the power source is configured with both distributed photovoltaic power generation and electric vehicle charging piles as 4, the power source is configured with both distributed photovoltaic power generation and self-provided power plants as 5, the power source is configured with both electric vehicle charging piles and self-provided power plants as 6, and the power source is configured with distributed photovoltaic power generation, electric vehicle charging piles and self-provided power plants as 7.

[0087] Step 2. Fill in the missing values ​​of power generation and consumption data based on the collected data.

[0088] like Figure 2 As shown, Figure 2 yes Figure 1 In the flowchart of step 2, in order to adapt to the structure of the convolutional neural network, the missing data in the power generation and consumption data collected in step 1 are filled by interpolation, and a data set with a time scale of 288 nodes is obtained for training the convolutional neural network and updating the data set. Different interpolation methods are designed to fill the missing values ​​according to the different actual situations of the missing values. In the present invention, the interpolation method is used to fill the missing values ​​in the original power generation data.

[0089] Step 2.1 If only one power generation and consumption data of the distributed generation is missing, the missing data is supplemented by the general interpolation method of power numbers;

[0090] If there is only one missing value in a certain period of time in the original power generation and consumption data collected in step 1, that is, the distributed power generation data, the general interpolation method is used, that is, 6 data of 15 minutes before and after the missing value are interpolated to fill the missing data. The expression is as follows:

[0091]

[0092] In the above formula: For missing values ​​in electricity data The regression function of Complete the formula for missing values; is the maximum allowable calibration deviation remainder; is the first-order difference quotient, is the second-order difference quotient, is the sixth-order difference quotient; f is the item completion function in the missing value completion formula, Sorting missing values ​​in electricity data for time The third power data before, Sorting missing values ​​in electricity data for time The second power data before, Sorting missing values ​​in electricity data for time The third digit after that is the power data.

[0093] Step 2.2 If the distributed generation is missing only 2 to 4 power generation and consumption data, the power data forward / backward interpolation method is used to fill in the missing data;

[0094] If there are 2 to 4 consecutive missing values ​​in the original power generation and consumption data collected in step 1, that is, the distributed power generation data, the earliest / latest missing data in time is first filled with the previous / next data closest to the moment, and then the remaining data is filled in sequence by interpolation method. The formula is as follows:

[0095]

[0096] In the above two formulas, is the data obtained by backward interpolation, is the data obtained by forward interpolation, is the 6th order backward difference, Δ 6 f6 is the sixth-order forward difference, γ is the standard elimination item of distributed power sources, which is 1 when the power generation and consumption data meet the requirements of the power company, and 0 when they do not meet the requirements, f6 is the regression function constructed by the backward interpolation method, f0 is the regression function constructed by the forward interpolation method, and t is the time series value of the missing power value.

[0097] Step 2.3 If the distributed generation source is missing more than 4 power generation and consumption data, the power data for this day will be invalidated;

[0098] In order to improve accuracy, consecutive missing values ​​are filled using backward / forward interpolation formulas, and if there are 5 or more consecutive days of missing power generation and consumption data, the data for that day will not be used.

[0099] The data obtained by interpolation fills the matrix X to obtain the matrix X′ that conforms to the algorithm network. The formula is as follows:

[0100]

[0101] In the above formula, it is assumed that the original matrix X Missing values, missing values ​​filled by downward interpolation is the data obtained by backward interpolation, It is the missing value of electricity data.

[0102] Step 3. After the completed power generation and consumption data is normalized with data enhancement effect, a data enhancement matrix is ​​constructed.

[0103] Specifically, class normalization is used on the updated power generation and consumption data set in step 2 to highlight the relationship between the power generation and consumption data of different distributed power sources and the time series;

[0104] like Figure 3 As shown, Figure 3 yes Figure 1 In the flowchart of step 4, considering that the power generation and consumption characteristics of distributed power sources have significant time correlation characteristics, in order to improve the robustness and recognition accuracy of the system, we perform data enhancement on the training data. The present invention adopts a class normalization method and eliminates the influence of the dimension difference of distributed power sources of different scales on the verification process. The expression is as follows:

[0105]

[0106] In the above formula: Distributed power type DR n The original measured data of power generation and consumption in the tth time period, The missing value of electricity The values ​​after class normalization.

[0107] Then substitute the data enhanced values ​​into the matrix X′ to obtain the data enhanced matrix X″, the formula is as follows:

[0108]

[0109] In the above formula: R nm×288 is a real number matrix of size nm×288 consisting of the power data after data enhancement.

[0110] When CNN uses the normalized data to input the model to obtain the verification results, the capacity of each type of distributed power supply is obtained by denormalization. The expression is as follows:

[0111]

[0112] In the above formula: is the actual power generation and consumption prediction value obtained after inverse class normalization, The power DR after class normalization n The power generation and consumption forecast value of the type in the tth time period, is the error deviation, indicating the error range within which the calibration is allowed. The values ​​obtained by denormalization form the feature matrix X″′ extracted by CNN. The formula of X″′ is as follows:

[0113]

[0114] Step 4. Train the deep convolutional neural network model; input the class normalized data into the convolutional neural network model for verification until a verification result with improved accuracy is obtained.

[0115] The data augmentation matrix constructed in step 3 is used as the input for training the deep convolutional neural network model in step 4. Its function is to highlight the power characteristics of different self-contained power sources and improve the efficiency of the model in verifying the type of self-contained power source.

[0116] This step is a type verification based on the CNN algorithm. The class-normalized data obtained in step 3 is input into the model, and the initial parameters are manually input into the model. The data results are denormalized and the verification results are output. The parameters are continuously adjusted until the verification accuracy increases.

[0117] Using the data set obtained in step 3, the convolutional neural network is set up. First, 22 pooling layers of size [288, 1, 1] are used as the feature extraction layer, and then 22 pooling layers of size [144, 1, 1] are passed, followed by Relu activation, and then a maximum pooling layer is added. Finally, 44 convolutional layers of size [88, 1, 1], 44 pooling layers of size [24, 1, 1], Relu activation, and maximum pooling layers are added in sequence.

[0118] The number of convolutional neural network iterations was set to 864, the learning rate was 0.001007, and the error threshold was set to 1.125×10 -4 ; When the number of iterations reaches 764, the learning rate is reduced to 0.00052, and the error threshold is set to 0.610×10 -4 In addition, the Adam gradient descent algorithm is used to optimize each pooling layer in the convolutional neural network.

[0119] The output result of the convolutional neural network of the present invention is 7 fully connected layers, and the formula decoding verification result after the anti-class normalization is as shown in formula (13), and the final verified distributed power type is output. The final verification result is as shown in Figure 4 As shown, finally, the feature matrix extracted from CNN is stored in the server.

[0120] Example 2

[0121] The present invention further provides an embodiment, which is a distributed power type verification method. In order to cooperate with the power grid company to accurately verify the distributed power configuration of active customers, the present invention proposes a distributed power configuration verification method based on convolutional neural networks (CNN).

[0122] When implementing it, Figure 1 As shown, in the first stage of the present invention, a distributed power type verification method is provided, and the steps are as follows:

[0123] Step 1. Collect power generation and consumption data of seven types of distributed power sources, including: only distributed photovoltaics are configured in the power source, only electric vehicle charging piles are configured in the power source, only self-provided power plants are configured in the power source, distributed photovoltaics and electric vehicle charging piles are configured in the power source at the same time, distributed photovoltaics and self-provided power plants are configured in the power source at the same time, electric vehicle charging piles and self-provided power plants are configured in the power source at the same time, and distributed photovoltaics, electric vehicle charging piles and self-provided power plants are configured in the power source at the same time;

[0124] Step 2. Label the types of the above-mentioned distributed power sources to form a distributed power generation and consumption data verification matrix, as shown in formula (1) in Example 1; design different interpolation methods to fill in the missing values ​​according to the actual situation of the missing values. If the distributed power source is only missing 1 power generation and consumption data, use the general interpolation method of power data to fill in the missing values; if the distributed power source is only missing 2 to 4 power generation and consumption data, use the forward / backward interpolation method of power data to fill in the missing values; if the distributed power source is only missing 5 or more power generation and consumption data, the power data for that day will be invalidated;

[0125] Step 3. Construct a data set that meets the format requirements and matches the network structure according to the missing values ​​filled in step 2, and convert it into a missing value filling matrix as shown in formula (10) in Example 1;

[0126] Step 4. In order to enhance the difference between power generation and consumption data of different distributed power sources, the data class is normalized and updated to a data enhancement matrix, as shown in formula (12) in Example 1;

[0127] Step 5. Use the normalized data set to set up the convolutional neural network. First, use 22 convolutional layers as feature extraction layers, and then use 22 pooling layers to compress features and extract more important features; then perform Relu activation, and then add a maximum pooling layer; finally, add 44 convolutional layers and 44 pooling layers in sequence;

[0128] Step 6. Each pooling layer or convolution layer is optimized by the Adam algorithm after output and sent to the next layer;

[0129] Step 7. The output layer adds 7 fully connected layers to output the type of distributed power source to be determined. The verification result is decoded by the denormalization formula, and the final verified distributed power source type is output in text form. Finally, the feature matrix extracted from CNN is stored in the server, as shown in formula (14) in Example 1.

[0130] Example 3

[0131] The present invention further provides an embodiment, which is a distributed power source type verification device, comprising:

[0132] The acquisition module is used to collect the power generation and consumption data and label data of the distributed power source;

[0133] A missing value processing module is used to process missing values ​​of power generation and consumption data based on the collected data;

[0134] The class normalization processing module is used to perform class normalization on the processed data and construct a data enhancement matrix;

[0135] The verification module is used to input the data after class normalization into the convolutional neural network model for verification until a verification result with improved accuracy is obtained.

[0136] The distributed power source type verification device is used to implement the steps of any one of the distributed power source type verification methods described in Example 1 or 2.

[0137] Example 4

[0138] Based on the same inventive concept, an embodiment of the present invention further provides a computer device, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor. When the processor executes the computer program, the steps of any one of the distributed power type verification methods described in Embodiment 1 or 2 are implemented.

[0139] Example 5

[0140] Based on the same inventive concept, an embodiment of the present invention further provides a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the distributed power source type verification methods described in Embodiment 1 or 2 are implemented.

[0141] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0142] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0143] 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.

[0144] 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. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0145] 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 distributed power source type verification method, characterized by: The following steps are involved: Collect power generation and consumption data and label data of distributed power sources; According to the collected data, the missing values ​​of power generation and consumption data are processed; Normalize the processed data and construct a data enhancement matrix; The class-normalized data is input into the convolutional neural network model for verification until a verification result with improved accuracy is obtained.

2. A distributed power source type verification method according to claim 1, characterized in that: The power generation and consumption data of the distributed power sources include: only distributed photovoltaics are configured in the distributed photovoltaic power source, only electric vehicle charging piles are configured in the power source, only self-contained power plants are configured in the power source, distributed photovoltaics and electric vehicle charging piles are configured in the power source at the same time, distributed photovoltaics and self-contained power plants are configured in the power source at the same time, electric vehicle charging piles and self-contained power plants are configured in the power source at the same time, and distributed photovoltaics, electric vehicle charging piles and self-contained power plants are configured in the power source at the same time; the label data refers to labeling the power generation and consumption data of different types of distributed power sources with corresponding labels, and collecting row data based on the labels.

3. A distributed power source type verification method according to claim 1, characterized in that: The processing of missing values ​​of power generation and consumption data based on the collected data includes: If the distributed generation source is missing only one power generation and consumption data, the general interpolation method of power numbers is used to fill in the missing data; If the distributed generation is missing 2 to 4 power generation and consumption data, the power data forward / backward interpolation method is used to fill in the missing data; If the distributed power generation is missing more than 4 power generation and consumption data, the power data for that day will be invalidated.

4. A distributed power source type verification method according to claim 3, characterized in that: If the distributed power source is missing only one power generation and consumption data, the missing data is supplemented by the general interpolation method of power data, including: Interpolation is performed on 6 data points, 15 minutes before and after the missing value, to fill in the missing data, as shown in the following formula: In the above formula: For missing values ​​in electricity data The regression function of Complete the formula for missing values; is the maximum allowable calibration deviation remainder; is the first-order difference quotient, is the second-order difference quotient, is the sixth-order difference quotient; f is the item completion function in the missing value completion formula, Sorting missing values ​​in electricity data for time The third power data before, Sorting missing values ​​in electricity data for time The second power data before, Sorting missing values ​​in electricity data for time The third power data after that; If the distributed power source is missing only 2 to 4 power generation and consumption data, the missing data is supplemented by the forward / backward interpolation method of power data, which means that the earliest / latest missing data in time is supplemented by the previous / latent data closest to the moment, and then the remaining data is supplemented in sequence by the interpolation method. The formula is as follows: In the above two formulas, is the data obtained by backward interpolation, is the data obtained by forward interpolation, is the 6th order backward difference, Δ 6 f6 is the sixth-order forward difference, γ is the standard elimination term for distributed power sources, which is 1 when the power generation and consumption data meet the requirements of the power company, and 0 when they do not meet the requirements, f6 is the regression function constructed by the backward interpolation method, f0 is the regression function constructed by the forward interpolation method, and t is the time series value of the missing power value; If the distributed power generation is missing more than 4 power generation and consumption data, the power data for this day will be invalidated, where: The data obtained by interpolation fills the matrix X to obtain the matrix X′ that conforms to the algorithm network. The formula is as follows: In the above formula, it is assumed that the original matrix X Missing values, missing values ​​filled by downward interpolation is the data obtained by backward interpolation, It is the missing value of electricity data.

5. The method for verifying the type of distributed power source according to claim 1, characterized in that: The processed data is normalized and a data enhancement matrix is ​​constructed, including: The class normalization method is adopted to eliminate the influence of the dimension difference of distributed power sources of different scales on the verification process. The expression is as follows: In the above formula: Distributed power type DR n The original measured data of power generation and consumption in the tth time period, The missing value of electricity The value after class normalization; Then substitute the data enhanced values ​​into the matrix X′ to obtain the data enhanced matrix X″, the formula is as follows: In the above formula: R nm×288 is a real number matrix of size nm×288 composed of power data after data enhancement; when the convolutional neural network uses the class-normalized data input model to obtain the verification result, the capacity of each distributed power type configuration is obtained by anti-class normalization, and the expression is as follows: In the above formula: is the actual power generation and consumption prediction value obtained after inverse class normalization, The power DR after class normalization n The power generation and consumption forecast value of the type in the tth time period, is the error deviation, indicating the error range within which the calibration is allowed. The values ​​obtained by denormalization form the feature matrix X″′, X″′ extracted by CNN, as shown in the following formula:

6. A distributed power source type verification method according to claim 1, characterized in that: The data after class normalization is input into the convolutional neural network model for verification until a verification result with an increased accuracy is obtained, wherein: the convolutional neural network is set using the obtained data set, and 22 pooling layers of size [288, 1, 1] are used as feature extraction layers. After 22 pooling layers of size [144, 1, 1], Relu activation is performed, and a maximum pooling layer is added, and 44 convolution layers of size [88, 1, 1], 44 pooling layers of size [24, 1, 1], Relu activation, and maximum pooling layers are added in sequence; The number of convolutional neural network iterations was set to 864, the learning rate was 0.001007, and the error threshold was 1.125×10 -4 When the number of iterations reaches 764, the learning rate is reduced to 0.00052, and the error threshold is set to 0.610×10 -4 In addition, the Adam gradient descent algorithm is used to optimize each pooling layer in the convolutional neural network; The output of the convolutional neural network is 7 fully connected layers. The verification result is decoded by the anti-class normalization formula, and the final verified distributed power type is output. The feature matrix extracted from CNN is stored in the server.

7. A distributed power source type verification device, characterized by: include: The acquisition module is used to collect the power generation and consumption data and label data of the distributed power source; A missing value processing module is used to process missing values ​​of power generation and consumption data based on the collected data; The class normalization processing module is used to perform class normalization on the processed data and construct a data enhancement matrix; A verification module is used to input the data after class normalization into the convolutional neural network model for verification until a verification result with improved accuracy is obtained; The device is used to implement the steps of a distributed power source type verification method as described in any one of claims 1-6.

8. A distributed power source type verification device according to claim 7, characterized in that: The processing of missing values ​​of power generation and consumption data based on the collected data includes: If the distributed generation source is missing only one power generation and consumption data, the general interpolation method of power numbers is used to fill in the missing data; If the distributed generation is missing 2 to 4 power generation and consumption data, the power data forward / backward interpolation method is used to fill in the missing data; If the distributed power generation is missing more than 4 power generation and consumption data, the power data for that day will be invalidated; The processed data is normalized and a data enhancement matrix is ​​constructed, including: The class normalization method is adopted to eliminate the influence of the dimension difference of distributed power sources of different scales on the verification process. The expression is as follows: In the above formula: Distributed power type DR n The original measured data of power generation and consumption in the tth time period, The missing value of electricity After class normalization, substitute the data enhancement values ​​into the matrix X′ to obtain the data enhancement matrix X″, the formula is as follows: In the above formula: R nm×288 is a real number matrix of size nm×288 composed of power data after data enhancement; when the convolutional neural network uses the class-normalized data input model to obtain the verification result, the capacity of each distributed power type configuration is obtained by anti-class normalization, and the expression is as follows: In the above formula: is the actual power generation and consumption prediction value obtained after inverse class normalization, The power DR after class normalization n The power generation and consumption forecast value of the type in the tth time period, is the error deviation, indicating the error range within which the calibration is allowed. The values ​​obtained by denormalization form the feature matrix X″′, X″′ extracted by CNN, as shown in the following formula:

9. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that: When the processor executes the computer program, the steps of a distributed power source type verification method described in any one of claims 1 to 6 are implemented.

10. A computer storage medium, characterized in that: The computer storage medium stores a computer program, and when the computer program is executed by a processor, the steps of a distributed power source type verification method described in any one of claims 1 to 6 are implemented.

Citation Information

Patent Citations

  • Time series data complement method, device and electronic device

    CN109460398A

  • Smart grid load prediction method

    CN111178618A

  • Distributed new energy power distribution network multi-objective optimization scheduling method and system

    CN114142521A

  • Power distribution network reconstruction method considering time-varying characteristics of load and distributed power supply

    CN118399501A

  • System and method for energy management of a power pack for an electric vehicle including photovoltaic charging

    US20240181927A1