Power grid regional load prediction system based on big data

Abstract

The application is suitable for the technical field of load prediction, and particularly relates to a power grid regional load prediction system based on big data, which comprises a searching device, a generating device, an extracting device and an input device; the searching device is used for dividing a power grid load prediction region, marking the region as a total region, collecting historical load data of the total region, configuring a plurality of influence factors, wherein the influence factors at least include weather and temperature, selecting a test time, defining a first power consumption and a first factor, and searching a same period time of previous years of the test time. The application can converge the weight value by establishing a global adversarial network, thereby further improving the accuracy of the weight value, can correct the weight value under the condition of limited sample data by constructing a local adversarial network, thereby reducing noise interference, greatly improving the prediction accuracy of the adversarial network, and effectively guaranteeing the safe and stable operation of the power grid.

Description

Technical Field

[0001] This invention relates to the field of load forecasting technology, and in particular to a power grid regional load forecasting system based on big data. Background Technology

[0002] Since electrical energy cannot be stored in large quantities, it is necessary to maintain a dynamic balance between the power output of power plants and the fluctuations in grid load during the operation of power plants and the dispatch of power. By using big data technology to predict the power load, it is possible to effectively ensure the safe and economical operation of the power grid.

[0003] If the total regional power grid is referred to as the large power grid, then in the sub-regional power grids (small power grids) that are opposite to it, due to the small number of samples, historical load data may be insufficient or incomplete, making data training more difficult and greatly affecting the prediction accuracy of big data. Therefore, "how to perform pre-training on the large power grid and apply it to the small power grid" is the technical problem that this invention needs to solve. Summary of the Invention

[0004] The purpose of this invention is to provide a power grid regional load forecasting system based on big data, in order to solve the problem mentioned in the background art of "how to perform pre-training on a large power grid and apply it to a small power grid".

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A power grid regional load forecasting system based on big data, the system comprising:

[0007] Find devices, generate devices, extract devices, and input devices;

[0008] The search device is used to divide the power grid load prediction area, mark it as the total area, collect historical load data of the total area, configure several influencing factors, wherein the influencing factors include at least: weather and temperature, select the test time, define the first electricity consumption and the first factor, find the same period of previous years of the test time, and define the second electricity consumption and the second factor.

[0009] The generating device is used to construct a global adversarial network consisting of a generator and a discriminator, upload the first factor, the second factor, and the second power consumption to the generator, and synchronize the first power consumption to the discriminator to generate the global adversarial network.

[0010] The extraction device is used to create a weight value calculation formula, construct a loss function, iterate the loss function, converge to obtain the weight value of each influencing factor, input the weight value into the global adversarial network, train the global adversarial network using the historical load data, adjust the weight value, and extract feature data from the trained global adversarial network.

[0011] The input device is used to integrate and generate a local adversarial network using the feature data, select a target sub-region from the total region, create a training set, train the local adversarial network, correct the weight values, collect real-time values ​​of the influencing factors of the target sub-region, input the real-time values ​​into the local adversarial network, and output the predicted value of the power grid load.

[0012] Furthermore, the locating device includes:

[0013] The segmentation module is used to segment the power grid load forecast area, which is marked as the total area. Historical load data of the total area is collected, and several influencing factors are configured, including at least weather and temperature.

[0014] The data preparation module is used to select the test time, define the first power consumption and the first factor, find the same period of previous years for the test time, and define the second power consumption and the second factor.

[0015] Furthermore, the generating device includes:

[0016] The upload module is used to build a global adversarial network consisting of a generator and a discriminator, and upload the first factor, the second factor, and the second power consumption to the generator;

[0017] The synchronization module is used to synchronize the first power consumption to the discriminator to generate a global adversarial network.

[0018] Furthermore, the extraction device includes:

[0019] The input module is used to create a weight value calculation formula, construct a loss function, iterate the loss function, converge to obtain the weight value of each influencing factor, and input the weight value into the global adversarial network.

[0020] The extraction module is used to train the global adversarial network using the historical load data, adjust the weight values, and extract feature data from the trained global adversarial network.

[0021] Furthermore, the input device includes:

[0022] The correction module is used to integrate and generate a local adversarial network using the feature data, select a target sub-region from the total region, create a training set, train the local adversarial network, and correct the weight values.

[0023] The prediction module is used to collect real-time values ​​of influencing factors in the target sub-region, input the real-time values ​​into the local adversarial network, and obtain the predicted value of the power grid load.

[0024] Furthermore, the partitioning module includes:

[0025] A traversal unit is used to traverse out additional factors, wherein the additional factors include at least: holidays, events, and population movement;

[0026] The linking unit is used to link the additional factors to the global adversarial network, calculate the corresponding weight values, and create an activation mechanism.

[0027] Furthermore, the data preparation module includes:

[0028] The reference unit is used to establish the mapping between the test time and the weight value, and to construct a reference table;

[0029] The plotting unit is used to plot several weight trend charts with the test time as the horizontal axis and the weight value as the vertical axis, and to configure the correlation relationships, wherein the correlation relationships include: positive, zero and negative.

[0030] Furthermore, the synchronization module includes:

[0031] The segmentation unit is used to integrate global constraints into the global adversarial network, segment the global constraints into several sub-items, and mark the priority corresponding to each sub-item, wherein the priority consists of high, medium and low.

[0032] An insertion unit is used to push the sub-items into the generator in descending order of priority.

[0033] Furthermore, the input module includes:

[0034] The quantization unit is used to quantify the weather, calculate the weather difference between the first factor and the second factor, denoted as W, and calculate the temperature difference between the first factor and the second factor, denoted as T.

[0035] A calculation unit is used to construct a weight value calculation formula, wherein the weight value calculation formula is:

[0036] ;

[0037] in y is the predicted value of the first electricity consumption, α is the weight value corresponding to the weather, and b is the weight value corresponding to the temperature.

[0038] Construct a loss function, which is:

[0039] ;

[0040] Where E is the deviation. This is the first electricity consumption;

[0041] By taking partial derivatives with respect to α and b respectively and performing several iterations, the weight value of each influencing factor is obtained.

[0042] Furthermore, the prediction module includes:

[0043] The corresponding unit is used to generate a label using the weight value and insert the label into the corresponding target sub-region;

[0044] The writing unit is used to record the actual value of the power grid load, calculate the difference between the actual value and the predicted value, and write the difference into a tag.

[0045] Compared with the prior art, the beneficial effects of the present invention are:

[0046] By defining the overall region, power dispatch can be optimized, grid reliability can be improved, and the sources of historical load data can be greatly expanded, providing sufficient training data for adversarial networks. By determining the test time and the same period in previous years, it is easy to calculate the weight values ​​and quantify the impact of weather and temperature on grid load, which greatly improves the accuracy and reliability of load forecasting. By building a global adversarial network, the weight values ​​can be converged, thereby further improving the accuracy of the weight values. By constructing a local adversarial network, the weight values ​​can be corrected with limited sample data, thereby reducing noise interference and greatly improving the prediction accuracy of the adversarial network, effectively ensuring the safe and stable operation of the power grid. Attached Figure Description

[0047] Figure 1 A block diagram of a power grid regional load forecasting system based on big data provided in an embodiment of the present invention;

[0048] Figure 2 A block diagram illustrating the components of a data-based power grid regional load forecasting system provided in this embodiment of the invention.

[0049] Figure 3 A block diagram of the generation equipment in the big data-based power grid regional load forecasting system provided in this embodiment of the invention;

[0050] Figure 4A block diagram illustrating the components of the equipment extracted in the big data-based power grid regional load forecasting system provided in this embodiment of the invention;

[0051] Figure 5 A block diagram illustrating the composition of input devices in a big data-based power grid regional load forecasting system provided in this embodiment of the invention;

[0052] Figure 6 A block diagram illustrating the composition of modules in a big data-based power grid regional load forecasting system provided in an embodiment of the present invention;

[0053] Figure 7 A block diagram illustrating the composition of the data preparation module in a big data-based power grid regional load forecasting system provided in an embodiment of the present invention;

[0054] Figure 8 A block diagram of the synchronization module in a big data-based power grid regional load forecasting system provided in an embodiment of the present invention;

[0055] Figure 9 A block diagram of the input module in a big data-based power grid regional load forecasting system provided in an embodiment of the present invention;

[0056] Figure 10 This is a block diagram of the prediction module in a big data-based power grid regional load prediction system provided in an embodiment of the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0058] Figure 1 The diagram shows the structural composition of a power grid regional load forecasting system based on big data provided in an embodiment of the present invention. The power grid regional load forecasting system 1 based on big data includes: a search device 11, a generation device 12, an extraction device 13, and an input device 14.

[0059] The search device 11 is used to divide the power grid load prediction area, mark it as the total area, collect historical load data of the total area, configure several influencing factors, wherein the influencing factors include at least weather and temperature, select the test time, define the first electricity consumption and the first factor, find the same period of previous years of the test time, and define the second electricity consumption and the second factor.

[0060] Based on the grid area division, the grid load forecasting area is determined and marked as the total area. The grid load forecasting area should have a sufficient sample size; for example, the total area can be a city-level grid, or a local grid such as a community, campus, hospital, or commercial center. Historical load data of the total area is collected from the power supply department or publicly available data. Historical load data is historical electricity consumption. The influencing factors of historical load data are determined, including weather and temperature. In this application, grid load under extreme weather conditions is not considered. A test time is selected, such as June 1, 2024, and the electricity consumption on that day is defined as the first electricity consumption. Temperature and weather data for that day are collected and defined as the first factor. The same period of previous years is found, i.e., June 1, 2023. The second electricity consumption and the second factor are determined in the same way. It should be noted that if the temperature difference between the first factor and the second factor is greater than the threshold, the same period of previous years should be reselected until the temperature difference is less than or equal to the threshold. The specific reselection method is to postpone the same period of previous years, and take June 2, 2023 as the same period of previous years.

[0061] In this application, for the purpose of illustrating the specific calculation process, only weather and temperature are used as examples of influencing factors. In the actual forecasting process, more influencing factors should be considered, but the calculation method is the same.

[0062] The generating device 12 is used to construct a global adversarial network consisting of a generator and a discriminator. It uploads the first factor, the second factor, and the second power consumption to the generator and synchronizes the first power consumption to the discriminator to generate the global adversarial network.

[0063] The first factor, the second factor, and the second electricity consumption mentioned above are uploaded to the generator, while the first electricity consumption is uploaded to the discriminator. The discriminator's role is to evaluate the authenticity of the generated result by comparing the load forecast output by the generator with the actual historical load data. The generator continuously adjusts the weight values ​​of the influencing factors to make the forecast result closer to the actual load data. By using the game between the generator and the discriminator, a global adversarial network is constructed. The advantage of this approach is that it can accurately capture the complex factors and potential change patterns affecting the power grid load, thereby improving the accuracy and reliability of load forecasting.

[0064] Continuing with the above example, in the generator, by using the electricity consumption and weather and temperature data of June 1, 2023, combined with the weather and temperature data of June 1, 2024, the electricity consumption of June 1, 2024 is calculated, and the calculated electricity consumption is determined as the predicted value. The predicted value is sent to the discriminator, where the predicted value and the actual electricity consumption are compared to determine whether the deviation between the two is within the control range. The control range is pre-defined by professionals. If it is within the control range, the weight value in the discriminator is output as the result. If it is not within the control range, the generator is used to adjust the weight value in the discriminator until the deviation is within the control range.

[0065] However, the above process is only used to initially determine the weight values. It is also necessary to converge the weight values ​​and train the adversarial network to improve the accuracy of the weight values ​​and the predicted data.

[0066] The extraction device 13 is used to create a weight value calculation formula, construct a loss function, iterate the loss function, converge to obtain the weight value of each influencing factor, input the weight value into the global adversarial network, train the global adversarial network using the historical load data, adjust the weight value, and extract feature data from the trained global adversarial network.

[0067] Create a formula for calculating weight values, i.e. The first factor, the second factor, and the first electricity consumption are substituted into the weight value calculation formula to calculate the predicted electricity consumption value for the test time. The difference between the predicted value and the second electricity consumption is defined as the deviation. The mean square value of the deviation is calculated, and a loss function is constructed. The partial derivatives of α and b are taken, and the function is iterated and converged multiple times to obtain the weight value. After the weight value is calculated, other test times are selected to train the global adversarial network and adjust the weight value. After training, the weight value is defined as feature data.

[0068] Continuing with the example above, suppose that the electricity consumption of a certain sub-region on June 1, 2023 was 1000MW, the temperature was 10℃, and the weather was sunny (sunny days are quantified as 1). On June 1, 2024, the electricity consumption was 1200MW, the temperature was 12℃, and the weather was rainy (sunny days are quantified as 2). In other words, the first electricity consumption is 1200MW, and the first factors are 12℃ and 2; the second electricity consumption is 1000MW, and the second factors are 10℃ and 1. Substituting these data into the weighting formula, we obtain...

[0069]

[0070] Where α is the weight value corresponding to temperature, b is the weight value corresponding to weather, and the deviation is... Let E be the value of the deviation. Calculate the mean square value of the deviation and construct the following loss function:

[0071] ;

[0072] in ;

[0073] The partial derivative with respect to α is:

[0074]

[0075] The partial derivative with respect to b is:

[0076] ;

[0077] Set the learning step size to 0.001 (this step size is pre-defined by professionals), and denote it as... And set the initial value of the weight to 0, that is , Substituting the initial value into the deviation, we get ;

[0078] Calculating the gradient of the above partial derivatives yields:

[0079]

[0080] ;

[0081] Using learning step size, for and Update and get

[0082]

[0083]

[0084] Calculated and Substituting this into E, we continue updating α and b along the direction where the gradient ratio between temperature and weather is 2:1, until... If the value is infinitely close to 1200MW, then α and b at this point are defined as the final weight values.

[0085] The input device 14 is used to integrate and generate a local adversarial network using the feature data, select a target sub-region from the total region, create a training set, train the local adversarial network, correct the weight values, collect real-time values ​​of the influencing factors of the target sub-region, input the real-time values ​​into the local adversarial network, and output the predicted value of the power grid load.

[0086] Feature data, which is also the weight value, is extracted from the adversarial network. This weight value is used to construct a local adversarial network. The total region is divided into several sub-regions, where a sub-region can be a building or a community, etc. The total region and sub-regions are only relative concepts and there is no specific dividing standard. From the sub-regions, the area that needs to be predicted for power grid load is selected, i.e., the target sub-region. The power load data of the target sub-region is read from the power supply department or public data, and the local adversarial network is trained to learn the load change pattern of the target sub-region and adjust the weight values ​​of the influencing factors in the target sub-region.

[0087] Real-time values ​​of influencing factors within the target sub-region are collected. These real-time values ​​are the predicted weather and temperature data for a future day, which can be obtained from meteorological data providers. The real-time values ​​are then input into the generator in the local adversarial network, and the discriminator outputs the predicted power grid load for a future day.

[0088] Figure 2 This diagram illustrates the structural composition of a power grid regional load forecasting system based on big data, as provided in an embodiment of the present invention. The search device 11 includes:

[0089] The segmentation module 111 is used to segment the power grid load forecast area, which is marked as the total area, collect historical load data of the total area, and configure several influencing factors, among which the influencing factors include at least weather and temperature.

[0090] The boundaries of the areas requiring power grid load forecasting are determined, the boundaries are integrated, and a total area is divided, which consists of several sub-areas. Historical daily load data for the total area is retrieved from power supply departments or publicly available data, and the influencing factors for each day are determined.

[0091] The data preparation module 112 is used to select the test time, define the first power consumption and the first factor, find the same period of previous years for the test time, and define the second power consumption and the second factor.

[0092] The number of test periods is not limited. The electricity load during the test period is defined as the first electricity consumption, and the influencing factors during the test period are defined as the first factor. Similarly, the second electricity consumption and the second factor are determined in the same period of previous years.

[0093] Figure 3 This diagram illustrates the structural composition of a big data-based power grid regional load forecasting system provided in an embodiment of the present invention. The generating device 12 includes:

[0094] The upload module 121 is used to construct a global adversarial network consisting of a generator and a discriminator, and upload the first factor, the second factor, and the second power consumption to the generator.

[0095] An adversarial network is constructed using a generator and a discriminator, where the adversarial network is mainly used to calculate and adjust the weight values.

[0096] The synchronization module 122 is used to synchronize the first power consumption to the discriminator to generate a global adversarial network.

[0097] The first factor, the second factor, and the second electricity consumption are uploaded to the generator, and the first electricity consumption is uploaded to the discriminator. The discriminator is used to compare the predicted data in the generator to determine whether the deviation is within the control range.

[0098] Figure 4 This diagram illustrates the structural composition of a power grid regional load forecasting system based on big data, as provided in an embodiment of the present invention. The extraction device 13 includes:

[0099] Input module 131 is used to create a weight value calculation formula, construct a loss function, iterate the loss function, converge to obtain the weight value of each influencing factor, and input the weight value into the global adversarial network.

[0100] A formula for calculating weight values ​​is constructed, and based on this, a loss function is constructed, and the weight values ​​are calculated. These weight values ​​are then transferred into the global adversarial network.

[0101] The extraction module 132 is used to train the global adversarial network using the historical load data, adjust the weight values, and extract feature data from the trained global adversarial network.

[0102] Historical load data of the total region is used to train the global adversarial network to improve the accuracy of the weight values. After training, the weight values ​​in the global adversarial network are defined as feature data.

[0103] Figure 5 This diagram illustrates the structural composition of a power grid regional load forecasting system based on big data, as provided in an embodiment of the present invention. The input device 14 includes:

[0104] The correction module 141 is used to integrate and generate a local adversarial network using the feature data, select a target sub-region from the total region, create a training set, train the local adversarial network, and correct the weight values.

[0105] Using feature data, a local adversarial network is constructed. From several sub-regions, a target sub-region is selected. Using the power grid load data of the target sub-region, a training set is created to train the local adversarial network, and the weight values ​​are gradually adjusted during the training process.

[0106] The prediction module 142 is used to collect the real-time values ​​of the influencing factors of the target sub-region, input the real-time values ​​into the local adversarial network, and obtain the predicted value of the power grid load.

[0107] The influencing factors of the target sub-region are identified, and the real-time values ​​of these factors are collected on a future day. These real-time values ​​are then input into the local adversarial network to obtain the predicted value of the power grid load.

[0108] Figure 6 This diagram illustrates the structural block diagram of a power grid regional load forecasting system based on big data provided in an embodiment of the present invention. The partitioning module 111 includes:

[0109] Traversal unit 1111 is used to traverse out additional factors, wherein the additional factors include at least: holidays, events and population movement.

[0110] In this embodiment, influencing factors include not only weather and temperature, but also holidays, events, and population movement; however, in the specific data processing process, additional factors need to be quantified.

[0111] Linking unit 1112 is used to link the additional factors to the global adversarial network, calculate the corresponding weight values, and create an activation mechanism.

[0112] After determining the additional factors, link the additional factors in the global adversarial network, and calculate the weight value corresponding to each additional factor according to the weight value calculation method mentioned in extraction device 13; however, in the process of predicting the power grid load at different times, it is not necessary to introduce all additional factors, that is, to introduce additional factors according to the activation mechanism; the activation mechanism can be: when the time for power grid load prediction is a holiday, then introduce the weight value corresponding to the holiday.

[0113] Figure 7 This diagram illustrates the structural composition of a power grid regional load forecasting system based on big data, as provided in an embodiment of the present invention. The data preparation module 112 includes:

[0114] The reference unit 1121 is used to establish the mapping between the test time and the weight value and to construct a reference table.

[0115] Different test times correspond to different weight values. By integrating the test times and their corresponding weight values, a comparison table can be constructed. The advantage of doing this is that it allows for the rapid determination of the weight value for each test time.

[0116] The drawing unit 1122 is used to draw several weight trend charts with the test time as the horizontal axis and the weight value as the vertical axis, and to configure the correlation relationship, wherein the correlation relationship includes: positive, zero and negative.

[0117] Find several adjacent test times and use them as the horizontal axis. Then, plot a weight trend chart with the weight values ​​of the influencing factors corresponding to each test time as the vertical axis. Determine whether there is a correlation between the weather and the weight values. The correlation can be positive, negative, or no correlation.

[0118] For example, record the daily temperature from June 1, 2024 to July 30, 2024, calculate the corresponding weight value, plot the temperature curve with time as the horizontal axis and temperature as the vertical axis, and plot the weight trend graph with time as the horizontal axis and temperature weight value as the vertical axis. Determine whether the temperature curve and the weight trend graph show a correlation.

[0119] Figure 8 This diagram illustrates the structural composition of a big data-based power grid regional load forecasting system provided in an embodiment of the present invention. The synchronization module 122 includes:

[0120] The segmentation unit 1221 is used to integrate global constraints into the global adversarial network, segment the global constraints into several sub-items, and mark the priority corresponding to each sub-item, wherein the priority consists of high, medium and low.

[0121] To accelerate weight convergence and reduce invalid computations, global constraints are inserted into the global adversarial network. These global constraints are a set of constraints on the data. For example, one constraint might be: if the temperature difference between the test time and the same period in previous years exceeds a threshold, then the same period in previous years should be reselected. Another constraint might be: if the temperature corresponding to a certain test time exceeds the fluctuation range, then the test time should be reselected. The global constraints are then divided into several sub-items, which are the constraints themselves. The priority of each constraint is determined, and the priority can be high, medium, or low.

[0122] Insertion unit 1222 is used to push the sub-items into the generator in descending order of priority.

[0123] Constraints are activated sequentially in descending order of priority.

[0124] Figure 9 This diagram illustrates the structural block diagram of a power grid regional load forecasting system based on big data provided in an embodiment of the present invention. The input module 131 includes:

[0125] The quantization unit 1311 is used to quantify the weather, calculate the weather difference between the first factor and the second factor, denoted as W, and calculate the temperature difference between the first factor and the second factor, denoted as T.

[0126] Calculation unit 1312 is used to construct a weight value calculation formula, wherein the weight value calculation formula is:

[0127]

[0128] in y is the predicted value of the first electricity consumption, α is the weight value corresponding to the weather, and b is the weight value corresponding to the temperature.

[0129] Construct a loss function, which is:

[0130]

[0131] Where E is the deviation. This is the first electricity consumption;

[0132] By taking partial derivatives with respect to α and b respectively and performing several iterations, the weight value of each influencing factor is obtained.

[0133] Following the steps shown in extraction device 13, calculate the weight value of each influencing factor.

[0134] Figure 10 This diagram illustrates the structural composition of a big data-based power grid regional load forecasting system provided in an embodiment of the present invention. The forecasting module 142 includes:

[0135] The corresponding unit 1421 is used to generate a label using the weight value and insert the label into the corresponding target sub-region.

[0136] Insert labels generated from the corresponding weight values ​​into the target sub-region; the advantage of doing this is that the corresponding weight values ​​can be quickly found and power grid load forecasting can be performed.

[0137] The writing unit 1422 is used to record the actual value of the power grid load, calculate the difference between the actual value and the predicted value, and write the difference into a tag.

[0138] The difference between the true value and the predicted value, i.e., the error, may vary in magnitude and distribution characteristics in different target sub-regions. By writing the error into the labels, the generator can adopt differentiated strategies for different regions during training. For example, in regions with larger errors, the weights of key influencing factors (such as temperature and crowd flow) can be adjusted, while in regions with smaller errors, the generalization ability of the model can be optimized. In addition, by inserting labels, the discrimination accuracy of the discriminator can be visually displayed, and which sub-regions have larger errors can be quickly identified in order to optimize the stability of adversarial training.

[0139] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0140] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

[0141] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A power grid regional load forecasting system based on big data, characterized in that, The system includes: a search device, a generation device, an extraction device, and an input device; The search device is used to divide the power grid load prediction area, mark it as the total area, collect historical load data of the total area, configure several influencing factors, wherein the influencing factors include at least: weather and temperature, select the test time, define the first electricity consumption and the first factor, find the same period of previous years of the test time, and define the second electricity consumption and the second factor. The generating device is used to construct a global adversarial network consisting of a generator and a discriminator, upload the first factor, the second factor, and the second power consumption to the generator, and synchronize the first power consumption to the discriminator to generate the global adversarial network. The extraction device is used to create a weight value calculation formula, construct a loss function, iterate the loss function, and converge to obtain the weight value of each influencing factor. The weight value is then input into the global adversarial network. The global adversarial network is trained using the historical load data, the weight value is adjusted, and feature data is extracted from the trained global adversarial network. The feature data refers to the adjusted weight value in the global adversarial network. The input device is used to integrate and generate a local adversarial network using the feature data, select a target sub-region from the total region, create a training set, train the local adversarial network, correct the weight values, collect real-time values ​​of the influencing factors of the target sub-region, input the real-time values ​​into the local adversarial network, and output the predicted value of the power grid load.

2. The power grid regional load forecasting system based on big data according to claim 1, characterized in that, The search device includes: The segmentation module is used to segment the power grid load forecast area, which is marked as the total area. Historical load data of the total area is collected, and several influencing factors are configured, including at least weather and temperature. The data preparation module is used to select the test time, define the first power consumption and the first factor, find the same period of previous years for the test time, and define the second power consumption and the second factor.

3. The power grid regional load forecasting system based on big data according to claim 1, characterized in that, The generating device includes: The upload module is used to build a global adversarial network consisting of a generator and a discriminator, and upload the first factor, the second factor, and the second power consumption to the generator; The synchronization module is used to synchronize the first power consumption to the discriminator to generate a global adversarial network.

4. The power grid regional load forecasting system based on big data according to claim 3, characterized in that, The extraction device includes: The input module is used to create a weight value calculation formula, construct a loss function, iterate the loss function, converge to obtain the weight value of each influencing factor, and input the weight value into the global adversarial network. The extraction module is used to train the global adversarial network using the historical load data, adjust the weight values, and extract feature data from the trained global adversarial network.

5. The power grid regional load forecasting system based on big data according to claim 1, characterized in that, The input device includes: The correction module is used to integrate and generate a local adversarial network using the feature data, select a target sub-region from the total region, create a training set, train the local adversarial network, and correct the weight values. The prediction module is used to collect real-time values ​​of influencing factors in the target sub-region, input the real-time values ​​into the local adversarial network, and obtain the predicted value of the power grid load.

6. The power grid regional load forecasting system based on big data according to claim 2, characterized in that, The partitioning module includes: A traversal unit is used to traverse out additional factors, wherein the additional factors include at least: holidays, events, and population movement; The linking unit is used to link the additional factors to the global adversarial network, calculate the corresponding weight values, and create an activation mechanism.

7. The power grid regional load forecasting system based on big data according to claim 2, characterized in that, The data preparation module includes: The reference unit is used to establish the mapping between the test time and the weight value, and to construct a reference table; The plotting unit is used to plot several weight trend charts with the test time as the horizontal axis and the weight value as the vertical axis, and to configure the correlation relationships, wherein the correlation relationships include: positive, zero and negative.

8. The power grid regional load forecasting system based on big data according to claim 3, characterized in that, The synchronization module includes: The segmentation unit is used to integrate global constraints into the global adversarial network, segment the global constraints into several sub-items, and mark the priority corresponding to each sub-item, wherein the priority consists of high, medium and low. An insertion unit is used to push the sub-items into the generator in descending order of priority.

9. The power grid regional load forecasting system based on big data according to claim 4, characterized in that, The input module includes: The quantization unit is used to quantify the weather, calculate the weather difference between the first factor and the second factor, denoted as W, and calculate the temperature difference between the first factor and the second factor, denoted as T. A calculation unit is used to construct a weight value calculation formula, wherein the weight value calculation formula is: ； in y is the predicted value of the first electricity consumption, α is the weight value corresponding to the weather, and b is the weight value corresponding to the temperature. Construct a loss function, which is: ； Where E is the deviation. This is the first electricity consumption; By taking partial derivatives with respect to α and b respectively and performing several iterations, the weight value of each influencing factor is obtained.

10. The power grid regional load forecasting system based on big data according to claim 5, characterized in that, The prediction module includes: The corresponding unit is used to generate a label using the weight value and insert the label into the corresponding target sub-region; The writing unit is used to record the actual value of the power grid load, calculate the difference between the actual value and the predicted value, and write the difference into a tag.

Images