A power grid short-term load prediction method, system, computer device and storage medium

Abstract

A power grid short-term load prediction method comprises the following steps: determining a prediction target period; and predicting the power grid short-term load in the prediction target period by using a pre-trained prediction model, wherein the prediction model is: The present application provides a power grid short-term load prediction method, system, computer device and storage medium, comprehensively considers the influence of long-term trend, seasonal factors, special events such as holidays and weather changes on the power grid load, establishes a load prediction model to accurately predict the future load of the power grid.

Description

Technical Field

[0001] This invention relates to the field of power grid management technology, specifically to a method, system, computer equipment, and storage medium for short-term power grid load forecasting. Background Technology

[0002] Currently, the installed capacity of non-fossil energy power generation in my country's power system is gradually increasing. While the increasing proportion of renewable energy sources reduces dependence on fossil fuels, it also brings new problems. The intermittent, random, and fluctuating nature of renewable energy sources makes grid regulation more difficult and rapidly depletes the power system's flexible regulation resources. This is particularly true for large renewable energy bases with weak grid structures and a lack of synchronous power support, where insufficient system support capacity is prevalent, affecting the safe and reliable transmission of renewable energy. Furthermore, although the overall utilization rate of renewable energy nationwide has remained high in recent years, the foundation for its absorption is still not solid, and the problem of wind and solar curtailment remains prominent in certain regions and time periods. In the future, the large-scale, high-proportion development of renewable energy requires a rapid improvement in system regulation capabilities, but the construction of regulatory resources faces many constraints, increasing the risk of efficient absorption of regional renewable energy and hindering its efficient utilization. The "dual high" characteristics of high-proportion renewable energy and high-proportion power electronic equipment are becoming increasingly prominent, posing significant risks and challenges to safe and stable operation. Compared to traditional power systems dominated by synchronous generators, "high-voltage and high-inertia" power systems are characterized by low inertia, low damping, and weak voltage support. Furthermore, my country's power grid exhibits a complex structure with strong AC / DC coupling between the sending and receiving ends and a complex voltage hierarchy, making coordination between the sending and receiving grids and between high- and low-voltage grids challenging. Faults can easily trigger cascading reactions. These challenges will have a profound impact on power system dispatch and operation.

[0003] Electricity load forecasting is a crucial aspect of power system dispatching and operation. Accurate and rapid forecasting methods can significantly improve the safety, stability, and economy of power system operation. Current technologies primarily rely on periodic predictions using historical data. However, grid load is influenced by numerous factors, such as electricity consumption scenarios and meteorological conditions. For large electricity consumption areas, the diverse scenarios and varying meteorological factors affect the overall grid load forecast. First, the complex and variable meteorological conditions within large areas, coupled with significant differences between different meteorological models, prevent the input meteorological conditions for grid load models from encompassing all weather conditions. Second, the diverse electricity consumption scenarios within large areas, including industrial, residential, and commercial use, exhibit inconsistent trends, making grid load model construction difficult and ultimately impacting the accuracy of electricity load forecasting for large areas. These complex factors limit the accuracy of existing forecasting methods. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method, system, computer equipment, and storage medium for short-term power grid load forecasting. It comprehensively considers the impact of long-term power grid load trends, seasonal factors, special events such as holidays, and weather changes on power grid load, and establishes a load forecasting model to accurately predict future power grid load.

[0005] To achieve the above objectives, the specific solution adopted by the present invention is as follows:

[0006] A method for short-term load forecasting of a power grid includes the following steps:

[0007] Determine the target forecast period;

[0008] The short-term load of the power grid during the target period is predicted using a pre-trained prediction model. The prediction model is as follows:

[0009]

[0010] Where α is the weighting coefficient, L 近期 For the near-term load of the power grid, L 去年,同日 Let G be the grid load during the same period corresponding to the target forecast period, and a be the average annual growth rate of the load. T and a H Here, represents the temperature influence coefficient and humidity influence coefficient, respectively; T and H represent the reference temperature and reference humidity for the target forecast period, respectively; c represents the holiday influence coefficient; D represents the holiday factor; e represents the weekend influence coefficient; E represents the weekend factor; f represents the influence coefficient when holidays and weekends overlap; F represents the holiday and weekend overlap factor; and ΔL represents the influence coefficient. 大工业 This refers to the projected load changes reported in advance by large industrial users.

[0011] As a further optimization of the above-mentioned short-term load forecasting method for power grids: the method for training the forecasting model includes: acquiring the expected load change of large industrial users and comprehensive historical data within a preset historical period, wherein the comprehensive historical data includes historical load data and historical meteorological data;

[0012] By filtering the comprehensive historical data, relevant basic load data and basic meteorological data are obtained;

[0013] Basic load curves are generated by clustering the basic load data.

[0014] Dynamically acquire the flexible load data of the power grid, and calculate the average flexible load during the same period and the flexible load time index within a preset sample period based on the flexible load data;

[0015] A sample set is constructed based on the expected load change, the basic load curve, the average value of the flexible load during the same period, and the time index of the flexible load; the prediction model is iteratively trained using the sample set.

[0016] As a further optimization of the above-mentioned short-term power grid load forecasting method, methods for obtaining relevant basic load data and basic meteorological data by filtering comprehensive historical data include:

[0017] Multiple load segments and multiple meteorological segments are extracted from historical load data and meteorological data, respectively, and the duration of both load segments and meteorological segments reaches the preset duration threshold.

[0018] Calculate the temporal similarity between load segments and meteorological segments. If the temporal similarity reaches a preset similarity threshold, then the meteorological segment and the load segment are considered to be associated.

[0019] Organize all related load segments and meteorological segments to generate basic load data and basic meteorological data.

[0020] As a further optimization of the above-mentioned short-term load forecasting method for power grids, methods for clustering basic load data include:

[0021] Determine the load level corresponding to the load segment and the meteorological classification corresponding to the meteorological segment;

[0022] Group all load segments of the same load level into one category.

[0023] As a further optimization of the above-mentioned short-term load forecasting method for power grids, the method for calculating the average value of the flexible load during the same period and the flexible load time index within a preset sample period based on flexible load data includes:

[0024] The dynamically acquired elastic load data is segmented based on a preset sample period;

[0025] Calculate the average value of elastic load data for each sample period;

[0026] Calculate the average total value based on the average values ​​of all periods in the same period;

[0027] The ratio of the average value during the same period to the total average value is calculated as the time index of the flexible load.

[0028] As a further optimization of the above-mentioned short-term load forecasting method for power grids, comprehensive historical data within a preset historical period is obtained directly from the power grid's automated dispatching system.

[0029] As a further optimization of the above-mentioned short-term power grid load forecasting method, the flexible load data includes electric vehicle charging and discharging load data and energy storage system load data.

[0030] A short-term load forecasting system for power grids includes:

[0031] Data acquisition device, used to acquire the expected load changes of large industrial users and comprehensive historical data within a preset historical period; data management device, used to build sample sets;

[0032] A forecasting device is used to predict the short-term load of the power grid during the target forecast period using a pre-trained forecasting model.

[0033] A data output device used to output the prediction results of the prediction device.

[0034] Computer equipment, including:

[0035] Memory, which stores computer programs;

[0036] A processor is used to read and execute the computer program to implement the above-described short-term load forecasting method for power grids.

[0037] A storage medium for storing a computer program that, when executed, implements the aforementioned method for short-term load forecasting of a power grid.

[0038] Beneficial effects: This invention comprehensively considers the impact of long-term trends in power grid load, seasonal factors, special events such as holidays, and weather changes on power grid load, establishes a load prediction model, and optimizes the model parameters through data. The trained model can accurately predict the future load of the power grid based on historical power grid load and recent average load, improve the accuracy of power grid load prediction, and provide better guidance for dispatchers to carry out real-time power grid balancing and dispatching. Attached Figure Description

[0039] Figure 1 This is a flowchart of the method of the present invention;

[0040] Figure 2 This is a diagram showing the experimental results in a specific implementation method. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] like Figure 1 As shown, the present invention first provides a method for short-term load forecasting of power grids, including S1 to S2.

[0043] S1. Determine the target forecast period. The length of the target period can be determined according to actual needs, such as one day, one week, or multiple days.

[0044] S2. Use a pre-trained prediction model to predict the short-term load of the power grid during the target period. The prediction model is as follows:

[0045]

[0046] Where α is the weighting coefficient, L 近期 For the near-term load of the power grid, L 去年,同日 Let G be the grid load during the same period corresponding to the target forecast period, and a be the average annual growth rate of the load. T and a H Here, represents the temperature influence coefficient and humidity influence coefficient, respectively; T and H represent the reference temperature and reference humidity for the target forecast period, respectively; c represents the holiday influence coefficient; D represents the holiday factor; e represents the weekend influence coefficient; E represents the weekend factor; f represents the influence coefficient when holidays and weekends overlap; F represents the holiday and weekend overlap factor; and ΔL represents the influence coefficient. 大工业 This refers to the estimated load change reported in advance by large industrial users. This value is determined based on the expected changes in the activities or operations of large industrial electricity customers and is used to adjust the overall load forecast in real time. Large industrial electricity customers are those whose electricity load accounts for more than 3%.

[0047] More specifically, for the holiday factor D, when a certain day is a holiday, D takes the value of 1, otherwise it takes the value of 0. Here, holidays refer to statutory holidays. For the weekend factor E, when a certain day is a weekend, E takes the value of 1, otherwise it takes the value of 0. For the holiday and weekend overlap factor F, when a certain day is both a holiday and a weekend, F takes the value of 1, otherwise it takes the value of 0.

[0048] The methods for training the prediction model include T1 to T6.

[0049] T1. Obtain the projected load changes of large industrial users and comprehensive historical data within a preset historical period. The comprehensive historical data includes historical load data and historical meteorological data. The projected load changes of large industrial users are reported by the users themselves based on their own circumstances. The comprehensive historical data within the preset historical period can be directly obtained from the power grid's distribution automation system. Distribution automation systems are mature existing technologies and will not be elaborated upon here. The preset historical period can be determined according to actual needs, for example, it can be set to a continuous three-year period.

[0050] T2. Filter the comprehensive historical data to obtain relevant basic load data and basic meteorological data. Methods for filtering the comprehensive historical data to obtain relevant basic load data and basic meteorological data include T21 to T23.

[0051] T21. Extract multiple load segments from historical load data and multiple meteorological segments from historical meteorological data. The duration of both load segments and meteorological segments must reach a preset duration threshold. Furthermore, within the same load segment, the difference between the total grid load at any two moments must be less than a preset difference threshold to ensure the grid load status remains stable within the same load segment. Similarly, within the same meteorological segment, the meteorological status must also remain stable; this meteorological status can include rainfall, temperature, and humidity. By ensuring that the duration of both load segments and meteorological segments reaches the preset duration threshold, interference from localized abnormal data can be avoided.

[0052] T22. Calculate the time similarity between load segments and meteorological segments. If the time similarity reaches a preset similarity threshold, the meteorological segment and the load segment are recorded as related. Specifically, if the time similarity between load segments and meteorological segments reaches the preset similarity threshold, it indicates that the power grid load recorded in the load segment is related to the meteorological state, meaning that the meteorological state may have an impact on the power grid load. Therefore, the meteorological segment and the load segment can be recorded as related.

[0053] T23. Organize all related load segments and meteorological segments to generate basic load data and basic meteorological data.

[0054] T3. Generate the basic load curve by clustering the basic load data. The methods for clustering the basic load data include T31 to T32.

[0055] T31. Determine the load level corresponding to the load segment and the weather classification corresponding to the weather segment. The load level can be high load, medium load, and low load, etc., and the weather classification can be sunny, cloudy, and rainy, etc., selected according to the actual situation.

[0056] T32. Group all load segments of the same load level into one category.

[0057] T4. Dynamically acquire the grid's flexible load data and calculate the average flexible load during the same period and the flexible load time index within a preset sample period based on the flexible load data. Specifically, the comprehensive historical data within the preset historical period is directly obtained from the grid's automated dispatch system. The flexible load data includes electric vehicle charging and discharging load data and energy storage system load data. Electric vehicles in the grid participate in V2G dispatch, meaning that during peak load periods, electric vehicles are given priority in dispatching and are used as energy storage devices to discharge to the grid. During off-peak periods, they continue to charge the devices to replenish their own power. When a distributed energy storage system is connected to the grid, it can store and release energy according to the existing load changes in the system. When the grid load is low, the distributed energy storage can act as a load to replenish its own power to fill the trough. When the grid load is high, the energy storage can act as a power source to release electricity to eliminate peaks, achieving the effect of peak shaving and valley filling.

[0058] Methods for calculating the average value of elastic load and the time index of elastic load within a preset sample period based on elastic load data include T41 to T44.

[0059] T41. The dynamically acquired elastic load data is segmented based on a preset sample period, which can be set to one week or one month.

[0060] T42. Calculate the average value of elastic load data for each sample period.

[0061] T43. Calculate the average total value based on the average values ​​of all periods.

[0062] T44. Calculate the ratio of the average value of the same period to the total average value as the elastic load time index.

[0063] T5. A sample set is constructed based on the expected load change, the basic load curve, the average value of the flexible load during the same period, and the flexible load time index. In one embodiment of the present invention, two thousand sets of data are collected to form a sample set, and the sample set is divided into a training set and a validation set in an 8:2 ratio, wherein the training set is used to train the model, and the validation set is used to verify the effect of the trained model.

[0064] T6. Use the sample set to iteratively train the prediction model.

[0065] To verify the effectiveness of the present invention, the following experiment was conducted based on data actually obtained from a power grid.

[0066] Table 1 shows the load forecast and actual load comparison for a certain power grid on October 29, 2024, in MW. Where α is taken as 0.9, L... 近期 Take the average load of the power grid from October 21 to 27, 2024, L去年,同日 Using the grid load on October 29, 2023, and G calculated as the average annual load growth of 17.55% based on the maximum load growth of the entire grid from 2015 to 2024, ΔL... 大工业 Based on the estimated load changes reported by large industrial users of the power grid, minus 140MW, the comprehensive load forecast data for December 29, 2023, deviates from the actual load of the power grid by less than 3%.

[0067] Table 1 Experimental Data

[0068]

[0069]

[0070]

[0071] Furthermore, such as Figure 2 As shown, the load data predicted by this invention has a high degree of fit with the actual load data, and this invention can accurately predict the power grid load.

[0072] The present invention further provides a short-term load forecasting system for power grids, including a data acquisition device, a data management device, a forecasting device, and a data output device.

[0073] The data acquisition device is used to obtain the expected load changes of large industrial users and comprehensive historical data within a preset historical period. The data acquisition device can directly connect to the power grid's automated dispatching system and directly acquire data from the automated dispatching system.

[0074] Data management device used to build sample sets.

[0075] A forecasting device is used to predict the short-term load of the power grid during the target forecast period using a pre-trained forecasting model.

[0076] A data output device used to output the prediction results of the prediction device.

[0077] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this disclosure, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division; in actual implementation, there may be other division methods. Furthermore, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or modules may be electrical, mechanical, or other forms.

[0078] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0079] The present invention also provides a computer device, including a memory and a processor.

[0080] A memory that stores computer programs.

[0081] A processor is used to read and execute the computer program to implement the above-described short-term load forecasting method for power grids.

[0082] Finally, the present invention provides a storage medium for storing a computer program that, when executed, implements the above-described method for short-term load forecasting of a power grid.

[0083] Based on this storage medium, the technical solution of this disclosure, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0084] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for short-term load forecasting of a power grid, characterized in that, Includes the following steps: Determine the target forecast period; The short-term load of the power grid during the target period is predicted using a pre-trained prediction model. The prediction model is as follows: ； in, These are the weighting coefficients. For the near-term load of the power grid, G represents the grid load during the same period corresponding to the target forecast period, where G is the average annual growth rate of the load. and Here, represents the temperature influence coefficient and humidity influence coefficient, respectively; T and H represent the reference temperature and reference humidity for the target forecast period, respectively; c represents the holiday influence coefficient; D represents the holiday factor; e represents the weekend influence coefficient; E represents the weekend factor; f represents the influence coefficient when holidays and weekends overlap; and F represents the overlap factor between holidays and weekends. This refers to the projected load changes reported in advance by large industrial users. The methods for training the prediction model include: Acquire the projected load changes of large industrial users and the comprehensive historical data within the preset historical period. The comprehensive historical data includes historical load data and historical meteorological data. By filtering the comprehensive historical data, relevant basic load data and basic meteorological data are obtained; Basic load curves are generated by clustering the basic load data. Dynamically acquire the flexible load data of the power grid, and calculate the average flexible load during the same period and the flexible load time index within a preset sample period based on the flexible load data; A sample set was constructed based on the expected load change, the basic load curve, the average value of the flexible load during the same period, and the time index of the flexible load. The prediction model is trained iteratively using a sample set; The method for calculating the average elastic load during the same period and the elastic load time index within a preset sample period based on elastic load data includes: The dynamically acquired elastic load data is segmented based on a preset sample period; Calculate the average value of elastic load data for each sample period; Calculate the average total value based on the average values ​​of all periods in the same period; The ratio of the average value during the same period to the total average value is calculated as the time index of the flexible load.

2. The short-term load forecasting method for power grids as described in claim 1, characterized in that, Methods for filtering comprehensive historical data to obtain relevant basic load data and basic meteorological data include: Multiple load segments and multiple meteorological segments are extracted from historical load data and meteorological data, respectively, and the duration of both load segments and meteorological segments reaches the preset duration threshold. Calculate the temporal similarity between load segments and meteorological segments. If the temporal similarity reaches a preset similarity threshold, then the meteorological segment and the load segment are considered to be associated. Organize all related load segments and meteorological segments to generate basic load data and basic meteorological data.

3. The short-term load forecasting method for power grids as described in claim 1, characterized in that, Methods for clustering basic load data include: Determine the load level corresponding to the load segment and the meteorological classification corresponding to the meteorological segment; Group all load segments of the same load level into one category.

4. The short-term load forecasting method for power grids as described in claim 1, characterized in that, When processing comprehensive historical data within a preset historical period, the data is obtained directly from the power grid's automated dispatching system.

5. The short-term load forecasting method for power grids as described in claim 1, characterized in that, Flexible load data includes electric vehicle charging and discharging load data and energy storage system load data.

6. A short-term load forecasting system for power grids, characterized in that, include: Data acquisition devices are used to acquire the expected load changes of large industrial users and comprehensive historical data within a preset historical period; Data management device, used to build sample sets; The forecasting device is used to predict the short-term load of the power grid during the target forecast period using a pre-trained forecasting model; the forecasting model is as follows: ； in, These are the weighting coefficients. For the near-term load of the power grid, G represents the grid load during the same period corresponding to the target forecast period, where G is the average annual growth rate of the load. and Here, represents the temperature influence coefficient and humidity influence coefficient, respectively; T and H represent the reference temperature and reference humidity for the target forecast period, respectively; c represents the holiday influence coefficient; D represents the holiday factor; e represents the weekend influence coefficient; E represents the weekend factor; f represents the influence coefficient when holidays and weekends overlap; and F represents the overlap factor between holidays and weekends. This refers to the projected load changes reported in advance by large industrial users. The methods for training the prediction model include: Acquire the projected load changes of large industrial users and the comprehensive historical data within the preset historical period. The comprehensive historical data includes historical load data and historical meteorological data. By filtering the comprehensive historical data, relevant basic load data and basic meteorological data are obtained; Basic load curves are generated by clustering the basic load data. Dynamically acquire the flexible load data of the power grid, and calculate the average flexible load during the same period and the flexible load time index within a preset sample period based on the flexible load data; A sample set was constructed based on the expected load change, the basic load curve, the average value of the flexible load during the same period, and the time index of the flexible load. The prediction model is trained iteratively using a sample set; The method for calculating the average elastic load during the same period and the elastic load time index within a preset sample period based on elastic load data includes: The dynamically acquired elastic load data is segmented based on a preset sample period; Calculate the average value of elastic load data for each sample period; Calculate the average total value based on the average values ​​of all periods in the same period; The ratio of the average value during the same period to the total average value is calculated as the time index of the flexible load. A data output device used to output the prediction results of the prediction device.

7. A computer device, characterized in that, include: Memory, which stores computer programs; A processor for reading and executing the computer program to implement a short-term power grid load forecasting method as described in any one of claims 1-5.

8. A storage medium, characterized in that, Used to store a computer program that, when executed, implements a short-term power grid load forecasting method as described in any one of claims 1-5.

Images