Load prediction method and device, storage medium and electronic equipment

Abstract

The invention discloses a load prediction method and device, a storage medium and electronic equipment, and relates to the technical field of photovoltaic power generation. The method comprises the following steps: acquiring historical load data of a power grid at different moments and photovoltaic permeability of a photovoltaic power supply in a power distribution network; preprocessing the historical load data to obtain preprocessed historical load data; determining to-be-analyzed time sequence data according to the preprocessed historical load data and the photovoltaic permeability; and inputting the time sequence data into a preset load prediction model for load prediction to obtain predicted load data, the preset load prediction model comprising a convolutional network and a residual network. The load prediction precision can be improved.

Description

Technical Field

[0001] The present application relates to the field of photovoltaic power generation technology, and in particular to a load forecasting method, device, storage medium and electronic equipment. Background Art

[0002] With the transformation of my country's energy structure and the rapid development of renewable energy, the penetration rate of photovoltaic power generation in the distribution network has continued to increase. The large-scale access of clean energy generators has changed the distribution network from a single power structure to a multi-power structure, greatly increasing the complexity of the distribution network. The accuracy of load forecasting will directly affect the operating efficiency and reliability of the distribution network. Therefore, how to accurately forecast load is a technical problem that needs to be solved urgently.

[0003] At present, when predicting future load data, usually only historical load data is referenced. However, this method of the prior art does not consider the impact of distributed energy on load forecasting, and its consideration factors are too single, making it difficult to ensure the accuracy of load forecasting. Summary of the invention

[0004] In view of this, the present application provides a load forecasting method, device, storage medium and electronic device, the main purpose of which is to improve the load forecasting accuracy.

[0005] According to a first aspect of the present application, a load forecasting method is provided, the method comprising:

[0006] Obtain historical load data of the power grid at different times, as well as the photovoltaic penetration rate of photovoltaic power sources in the distribution network;

[0007] Preprocessing the historical load data to obtain preprocessed historical load data;

[0008] Determining time series data to be analyzed according to the preprocessed historical load data and the photovoltaic penetration rate;

[0009] The time series data is input into a preset load forecasting model to perform load forecasting to obtain predicted load data, wherein the preset load forecasting model includes a convolutional network and a residual network.

[0010] According to a second aspect of the present application, a load forecasting device is provided, the device comprising:

[0011] An acquisition unit, used to acquire historical load data of the power grid at different times, and the photovoltaic penetration rate of photovoltaic power sources in the distribution network;

[0012] A preprocessing unit, used for preprocessing the historical load data to obtain preprocessed historical load data;

[0013] A determination unit, configured to determine time series data to be analyzed based on the preprocessed historical load data and the photovoltaic penetration rate;

[0014] The prediction unit is used to input the time series data into a preset load prediction model to perform load prediction and obtain predicted load data, wherein the preset load prediction model includes a convolutional network and a residual network.

[0015] According to a third aspect of the present application, a storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned load forecasting method is implemented.

[0016] According to a fourth aspect of the present application, an electronic device is provided, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor implements the above-mentioned load forecasting method when executing the program.

[0017] By means of the above technical scheme, the load forecasting method, device, storage medium and electronic device provided by the present application, compared with the prior art, can perform load forecasting based on time series data and a preset load forecasting model by obtaining historical load data of the power grid at different times, as well as the photovoltaic penetration rate of photovoltaic power sources in the distribution network, and determining the time series data to be analyzed based on the preprocessed historical load data and photovoltaic penetration rate. Since the present application not only considers the historical load data when performing load forecasting, but also fully considers the impact of distributed energy on load forecasting by introducing photovoltaic penetration rate, it can improve the accuracy of load forecasting. At the same time, the present application can further ensure the accuracy of load forecasting results by using a preset load forecasting model including a convolutional network and a residual network for load forecasting.

[0018] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] 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:

[0020] Figure 1 A schematic diagram of a load forecasting method provided in an embodiment of the present application is shown;

[0021] Figure 2 A schematic diagram of the overall architecture of load forecasting provided by an embodiment of the present application is shown;

[0022] Figure 3 A schematic diagram of a process for determining time series data provided by an embodiment of the present application is shown;

[0023] Figure 4 A schematic diagram of the structure of an extended convolutional network provided in an embodiment of the present application is shown;

[0024] Figure 5 A schematic diagram of the structure of a residual network provided in an embodiment of the present application is shown;

[0025] Figure 6 A schematic diagram showing a load forecast result provided by an embodiment of the present application is shown;

[0026] Figure 7 A visualization diagram of the test load data provided by an embodiment of the present application is shown;

[0027] Figure 8 A scatter plot provided by an embodiment of the present application is shown;

[0028] Fig. 9 The KDE contour map provided in the embodiment of the present application is shown;

[0029] Fig.10 A schematic structural diagram of a load forecasting device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0030] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0031] The existing technology does not consider the impact of distributed energy on load forecasting, and its consideration factors are too single, making it difficult to ensure the accuracy of load forecasting.

[0032] In order to solve the above problems, an embodiment of the present invention provides a load forecasting method, such as Figure 1 As shown, the method includes:

[0033] Step 10: Obtain historical load data of the power grid at different times, and the photovoltaic penetration rate of photovoltaic power sources in the distribution network.

[0034] Among them, the historical load data includes the active power and reactive power of the power grid at different times.

[0035] The embodiment of the present invention is mainly applicable to ultra-short-term load forecasting scenarios, that is, forecasting the power load in the next few minutes to hours. The execution subject of the embodiment of the present invention is a device or equipment capable of power load forecasting, such as a server.

[0036] The overall architecture of load forecasting in the embodiment of the present invention is as follows: Figure 2 As shown, the historical load data is first preprocessed, and then the preprocessed historical load data is combined with the photovoltaic penetration rate and input into a preset load forecasting model including a convolutional network and a residual network for load forecasting.

[0037] When carrying out load forecasting, it is necessary to first obtain the historical load data of the power grid at different times. For example, collect the active power and reactive power of the power grid every minute in the past 15 minutes to predict the load data of the power grid in the 16th and 17th minutes in the future. At the same time, obtain the photovoltaic penetration rate of photovoltaic power sources in the distribution network provided by the power grid manufacturer.

[0038] Step 20: preprocess the historical load data to obtain preprocessed historical load data.

[0039] For the embodiment of the present invention, before using historical load data for load forecasting, the historical load data needs to be preprocessed. Regarding the preprocessing process of historical load data, the method includes: filling the historical load data with default values ​​to obtain the filled historical load data; performing abnormal value processing on the filled historical load data to obtain the abnormally processed historical load data; performing standardization processing on the abnormally processed historical load data to obtain the standardized historical load data.

[0040] Specifically, when filling default values ​​for historical load data, the default values ​​can be filled by interpolation, mean or median filling. When processing outliers for historical load data, the Lagrange interpolation method can be used to replace outliers, and the mean or median can also be used to replace outliers. When standardizing historical load data, the maximum and minimum standardization method can be used to scale the historical load data to a specified range. The specific formula is as follows:

[0041] X i_std =(X i -X min ) / (X max -X min )

[0042] Among them, X min and X max is the minimum and maximum historical load data in the historical load data at different times, X i_std is the standardized historical load data.

[0043] Therefore, according to the above method, the default value filling, abnormal value processing and standardization processing of the historical load data can be completed, so as to ensure the quality of the data to be analyzed.

[0044] Step 30: Determine the time series data to be analyzed based on the preprocessed historical load data and the photovoltaic penetration rate.

[0045] For the embodiment of the present invention, after preprocessing the historical load data, it is necessary to combine it with the photovoltaic penetration rate to determine the time series data to be analyzed. For this process, Figure 3 As shown, including:

[0046] Step 31: Calculate the apparent power according to the active power and reactive power in the preprocessed historical load data.

[0047]

[0048] tanθ=Q / P

[0049] Among them, P represents active power, Q represents reactive power, S represents apparent power, θ is the current phase angle, and cosθ represents power factor.

[0050] Step 32: Determine the time series data to be analyzed according to the apparent power and the photovoltaic penetration rate.

[0051] According to the embodiment of the present invention, after the apparent power of the power grid at different times is determined, the photovoltaic penetration rate is introduced, and the time series data to be analyzed is determined according to the apparent power and the photovoltaic penetration rate.

[0052] Since the embodiment of the present invention not only considers historical load data when performing load forecasting, but also fully considers the impact of distributed energy on load forecasting by introducing photovoltaic penetration rate, it is possible to improve the accuracy of load forecasting.

[0053] In some embodiments, in order to further improve the load forecasting accuracy, the embodiments of the present invention also introduce historical meteorological data. Based on this, the method further includes: obtaining the historical meteorological data at different times; determining the time series data to be analyzed according to the apparent power, the photovoltaic penetration rate and the historical meteorological data. The historical meteorological data includes temperature data and light intensity, etc.

[0054] Step 40: Input the time series data into a preset load forecasting model to perform load forecasting and obtain forecasted load data.

[0055] The preset load forecasting model includes a convolutional network and a residual network, the number of convolutional networks and residual networks are multiple, and the input and output lengths of the preset load forecasting model are equal. The preset load forecasting model can be a TCN model or other models, and the embodiment of the present invention does not specifically limit this.

[0056] For the embodiment of the present invention, when using the preset load forecasting model to perform load forecasting, step 40 specifically includes: inputting the time series data into the convolutional network for feature extraction to obtain the time series features corresponding to the time series data; inputting the time series features and the time series data into the residual network for residual calculation to obtain the residual calculation result, and adding the residual calculation result to the input of the next convolutional network; until each convolutional network and each residual network completes data processing, the predicted load data is output.

[0057] Among them, the convolutional network includes an extended convolutional network and an activation function. The extended convolutional network can specifically be a one-dimensional extended convolutional network or an extended convolutional network of other dimensions. The embodiment of the present invention does not make any specific limitation on this.

[0058] When adding the residual calculation result to the input of the next convolutional network, the residual calculation result is first nonlinearly mapped to obtain the mapped residual calculation result, and then the mapped residual calculation result is added to the input of the next convolutional network.

[0059] When using a convolutional network to extract time series features, the time series data is input into the extended convolutional network for sampling to obtain sampled time series data; the sampled time series data is weight normalized to obtain normalized time series data; the normalized time series data is input into the activation function for processing, and a discard operation is performed to obtain the time series features.

[0060] Specifically, the structure of the extended convolutional network is as follows: Figure 4 As shown in Figure 2, the apparent power and photovoltaic penetration rate x0, x1, ..., xn-1 at different historical moments are used as the input xt of the extended convolutional network at time t, which meets the input conditions of time series prediction and the output is the sampled time series data. d is the number of expansions of the output. The extended convolutional network can collect time series data at different paces, and the sampling frequencies of different layers are different. When the sampling frequency is 1, it means that all the data of the lower layer are used; when the sampling frequency is 2, it means that one sample data is taken for every two points. In order to extract richer and more accurate feature information, the input layer often samples all the data.

[0061] By introducing an extended convolutional network, the embodiment of the present invention can process longer time series data dependencies at the same complexity, thereby improving the load prediction accuracy of the model.

[0062] The structure of the residual network is as follows Figure 5As shown, the output of the convolutional network is connected to the input of the residual network, and each convolutional network corresponds to a residual network. The output and input of convolutional networks at different levels are added together to form the input of the residual network, which is output after the activation function. The residual network connection mechanism can improve the feedback and convergence of the network, and can avoid the gradient vanishing and gradient exploding problems of traditional neural networks. The preset load forecasting model of an embodiment of the present invention also includes a 1×1 convolutional layer for nonlinear mapping.

[0063] The input time series data is first subjected to one-dimensional extended causal convolution, and then the weights are normalized, which can solve the gradient explosion and improve the calculation speed. Then the ReLU activation function is used to make the network nonlinear, and finally neurons are randomly discarded to solve the overfitting problem and improve the training speed of the model.

[0064] Finally, based on the input historical load data and PV penetration rate, the preset load forecasting model can output load forecast results, such as Figure 6 shown.

[0065] The prediction evaluation index of the embodiment of the present invention is R 2 , RMSE, MAE, SMAPE. Among them, R 2 The value reaches 0.9922, which is close to 1, a perfect prediction, indicating that the preset load forecasting model can explain 99.22% of the load changes; the RMSE is 188.98MW, and the relative error is only 2.36% relative to the load mean (about 8000MW); the MAE is 106.69MW, indicating that the average absolute deviation between the predicted value and the actual value is small; the SMAPE is 7.14%, reflecting that the relative error level of the prediction is in a low range.

[0066] The preset load forecasting model provided by the embodiment of the present invention performs well in grasping the overall trend of the load, can accurately capture the periodic changes of the load, and has the best prediction stability and accuracy when the load level is high. During low load periods, the preset load forecasting model still maintains good tracking capabilities. At the same time, at the mutation point of the load, the load forecast value can respond and adjust quickly, thereby reflecting the dynamic adaptability of the load forecasting model.

[0067] In the embodiment of the present invention, before constructing a preset load prediction model and using the preset load prediction model to perform load prediction, it is necessary to first prove that there is a correlation between photovoltaic penetration and load data. For this process, the method includes: obtaining test load data and test photovoltaic penetration; calculating the distance correlation coefficient based on the test load data and the test photovoltaic penetration using a preset distance correlation coefficient algorithm; if the distance correlation coefficient is greater than the preset correlation coefficient, it is determined that the test load data is correlated with the test photovoltaic penetration. The preset correlation coefficient can be set according to actual business needs.

[0068] The specific calculation formula of the distance correlation coefficient is as follows:

[0069]

[0070] Among them, r represents the distance correlation coefficient, X and Y represent the test load data and the test PV penetration rate, respectively, and mx and my represent the mean of load data and the mean of PV penetration rate, respectively.

[0071] For example, the calculated distance correlation coefficient is 0.397, which is greater than the preset correlation coefficient, indicating that there is a medium-strength correlation between the photovoltaic penetration rate and the load data, that is, there is a certain dependence between the two variables, and it can also reflect the overall correlation and nonlinear relationship between the variables.

[0072] Visualize the collected test load data. Figure 7 As shown. Figure 8 The scatter plot shows that most of the data are concentrated in areas with low photovoltaic penetration, and the data are not clearly distributed. Fig. 9 The KDE contour map shown shows that the area with the highest data density is located at the intersection of low penetration and low load. As the photovoltaic penetration rate increases, the load value shows an overall downward trend. This is because when the photovoltaic penetration rate is high, the photovoltaic system can provide more electricity to users, so the grid load demand continues to decrease.

[0073] A load forecasting method provided by an embodiment of the present invention can perform load forecasting based on time series data and a preset load forecasting model by acquiring historical load data of the power grid at different times and the photovoltaic penetration rate of photovoltaic power sources in the distribution network, and determining the time series data to be analyzed based on the preprocessed historical load data and photovoltaic penetration rate. Since the embodiment of the present invention not only considers the historical load data when performing load forecasting, but also fully considers the impact of distributed energy on load forecasting by introducing photovoltaic penetration rate, it is possible to improve the accuracy of load forecasting. At the same time, the present application can further ensure the accuracy of load forecasting results by using a preset load forecasting model including a convolutional network and a residual network to perform load forecasting.

[0074] Further, as Figure 1 and Figure 3 The specific implementation of the method shown in the embodiment provides a load forecasting device, such as Fig.10 As shown, the device includes: an acquisition unit 101, a preprocessing unit 102, a determination unit 103 and a prediction unit 104.

[0075] The acquisition unit 101 may be used to acquire historical load data of the power grid at different times, and the photovoltaic penetration rate of photovoltaic power sources in the distribution network.

[0076] The preprocessing unit 102 may be used to preprocess the historical load data to obtain preprocessed historical load data.

[0077] The determining unit 103 may be configured to determine the time series data to be analyzed according to the preprocessed historical load data and the photovoltaic penetration rate.

[0078] The prediction unit 104 can be used to input the time series data into a preset load prediction model to perform load prediction and obtain predicted load data, wherein the preset load prediction model includes a convolutional network and a residual network.

[0079] In some embodiments, the preprocessing unit 102 can be specifically used to fill the historical load data with default values ​​to obtain filled historical load data; perform outlier processing on the filled historical load data to obtain outlier processed historical load data; perform standardization processing on the outlier processed historical load data to obtain standardized historical load data.

[0080] In some embodiments, the determination unit 103 includes: a calculation module and a determination module.

[0081] The calculation module can be used to calculate the apparent power according to the active power and reactive power in the preprocessed historical load data.

[0082] The determination module may be used to determine the time series data to be analyzed according to the apparent power and the photovoltaic penetration rate.

[0083] In some embodiments, the determination module may be specifically used to obtain the historical meteorological data at different times; and determine the time series data to be analyzed based on the apparent power, the photovoltaic penetration rate and the historical meteorological data.

[0084] In some embodiments, the prediction unit 104 includes: an extraction module, a residual calculation module and an output module.

[0085] The extraction module can be used to input the time series data into the convolutional network for feature extraction to obtain the time series features corresponding to the time series data.

[0086] The residual calculation module can be used to input the time series features and the time series data into the residual network to perform residual calculation, obtain a residual calculation result, and add the residual calculation result to the input of the next convolutional network.

[0087] The output module can be used to output the predicted load data after each convolutional network and each residual network completes data processing.

[0088] In some embodiments, the residual calculation module can be specifically used to perform nonlinear mapping on the residual calculation result to obtain a mapped residual calculation result; and add the mapped residual calculation result to the input of the next convolutional network.

[0089] In some embodiments, the convolutional network includes an extended convolutional network and an activation function, and the extraction module can be specifically used to input the time series data into the extended convolutional network for sampling to obtain sampled time series data; perform weight normalization on the sampled time series data to obtain normalized time series data; input the normalized time series data into the activation function for processing, and perform a discard operation to obtain the time series features.

[0090] In some embodiments, the device further comprises: an analysis unit.

[0091] The analysis unit can be used to obtain test load data and test photovoltaic penetration rate; based on the test load data and the test photovoltaic penetration rate, a preset distance correlation coefficient algorithm is used to calculate the distance correlation coefficient; if the distance correlation coefficient is greater than the preset correlation coefficient, it is determined that the test load data is correlated with the test photovoltaic penetration rate.

[0092] It should be noted that for other corresponding descriptions of the functional units involved in the load forecasting device provided in this embodiment, please refer to Figure 1 and Figure 3 The corresponding description in will not be repeated here.

[0093] Based on the above Figure 1 and Figure 3 The method shown in the embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the above-mentioned Figure 1 and Figure 3 The load forecasting method shown.

[0094] Based on this understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, USB flash drive, mobile hard disk, etc.), including a number of instructions for enabling an electronic device (which can be a personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of the present application.

[0095] Based on the above Figure 1 and Figure 3 The method shown, and Fig.10 In order to achieve the above-mentioned purpose, the embodiment of the present application also provides an electronic device, which can be a personal computer, a tablet computer, a server, or other network equipment, etc. The device includes a storage medium and a processor; the storage medium is used to store a computer program; the processor is used to execute the computer program to achieve the above-mentioned Figure 1 and Figure 3 The load forecasting method shown.

[0096] Optionally, the above-mentioned physical device may also include a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, etc. The user interface may include a display, an input unit such as a keyboard, etc., and the optional user interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface), etc.

[0097] Those skilled in the art will appreciate that the above-mentioned physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or a combination of certain components, or different arrangements of components.

[0098] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the above-mentioned physical device, and supports the operation of the information processing program and other software and / or programs. The network communication module is used to realize the communication between the components inside the storage medium, and the communication with other hardware and software in the information processing physical device.

[0099] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform, or by hardware.

[0100] The embodiment of the present invention obtains the historical load data of the power grid at different times, as well as the photovoltaic penetration rate of photovoltaic power sources in the distribution network, and determines the time series data to be analyzed based on the pre-processed historical load data and photovoltaic penetration rate, so as to perform load forecasting based on the time series data and the preset load forecasting model. Since the embodiment of the present invention not only considers the historical load data when performing load forecasting, but also fully considers the impact of distributed energy on load forecasting by introducing photovoltaic penetration rate, it is possible to improve the accuracy of load forecasting. At the same time, the embodiment of the present invention can further ensure the accuracy of load forecasting results by using a preset load forecasting model including a convolutional network and a residual network for load forecasting.

[0101] Those skilled in the art will appreciate that the accompanying drawings are only schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily necessary for implementing the present application. Those skilled in the art will appreciate that the modules in the devices in the implementation scenario can be distributed in the devices of the implementation scenario according to the description of the implementation scenario, or can be changed accordingly and located in one or more devices different from the present implementation scenario. The modules of the above-mentioned implementation scenario can be combined into one module, or can be further split into multiple submodules.

[0102] The above serial numbers of this application are only for description and do not represent the advantages and disadvantages of the implementation scenarios. The above disclosure is only a few specific implementation scenarios of this application, but this application is not limited to them, and any changes that can be thought of by technicians in this field should fall within the scope of protection of this application.

Claims

1. A load forecasting method, characterized in that: include: Obtain historical load data of the power grid at different times, as well as the photovoltaic penetration rate of photovoltaic power sources in the distribution network; Preprocessing the historical load data to obtain preprocessed historical load data; Determining time series data to be analyzed according to the preprocessed historical load data and the photovoltaic penetration rate; The time series data is input into a preset load forecasting model to perform load forecasting to obtain predicted load data, wherein the preset load forecasting model includes a convolutional network and a residual network.

2. The method according to claim 1, characterized in that: The preprocessing of the historical load data to obtain the preprocessed historical load data includes: Filling the historical load data with default values ​​to obtain filled historical load data; Performing outlier processing on the filled historical load data to obtain outlier processed historical load data; The historical load data after the abnormal processing is standardized to obtain standardized historical load data.

3. The method according to claim 1, characterized in that The step of determining the time series data to be analyzed based on the pre-processed historical load data and the photovoltaic penetration rate includes: Calculating apparent power based on the active power and reactive power in the preprocessed historical load data; The time series data to be analyzed is determined according to the apparent power and the photovoltaic penetration rate.

4. The method according to claim 3, characterized in that The step of determining the time series data to be analyzed according to the apparent power and the photovoltaic penetration rate comprises: Obtaining historical meteorological data at different times; The time series data to be analyzed is determined according to the apparent power, the photovoltaic penetration rate and the historical meteorological data.

5. The method according to claim 1, characterized in that: The step of inputting the time series data into a preset load forecasting model to perform load forecasting to obtain forecasted load data includes: Inputting the time series data into the convolutional network to extract features, and obtaining time series features corresponding to the time series data; Inputting the time series features and the time series data into the residual network to perform residual calculation, obtaining a residual calculation result, and adding the residual calculation result to the input of the next convolutional network; After each convolutional network and each residual network completes data processing, the predicted load data is output.

6. The method according to claim 5, characterized in that The adding the residual calculation result to the input of the next convolutional network includes: Performing nonlinear mapping on the residual calculation result to obtain a mapped residual calculation result; Adding the mapped residual calculation result to the input of the next convolutional network; and / or The convolution network includes an extended convolution network and an activation function, and the time series data is input into the convolution network for feature extraction to obtain the time series features corresponding to the time series data, including: Inputting the time series data into the extended convolutional network for sampling to obtain sampled time series data; Performing weight normalization processing on the sampled time series data to obtain normalized time series data; The normalized time series data is input into the activation function for processing, and a discard operation is performed to obtain the time series features.

7. The method according to any one of claims 1 to 6, characterized in that: Before acquiring the historical load data of the power grid at different times and the photovoltaic penetration rate of the photovoltaic power source in the distribution network, the method further includes: Obtain test load data and test PV penetration; Calculate the distance correlation coefficient using a preset distance correlation coefficient algorithm according to the test load data and the test photovoltaic penetration rate; If the distance correlation coefficient is greater than the preset correlation coefficient, it is determined that the test load data is correlated with the test photovoltaic penetration rate.

8. A load forecasting device, characterized in that: include: An acquisition unit, used to acquire historical load data of the power grid at different times, and the photovoltaic penetration rate of photovoltaic power sources in the distribution network; A preprocessing unit, used for preprocessing the historical load data to obtain preprocessed historical load data; A determination unit, configured to determine time series data to be analyzed based on the preprocessed historical load data and the photovoltaic penetration rate; The prediction unit is used to input the time series data into a preset load prediction model to perform load prediction and obtain predicted load data, wherein the preset load prediction model includes a convolutional network and a residual network.

9. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. An electronic 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 method according to any one of claims 1 to 7 is implemented.

Images