Load data expansion method and system based on improved expansion domain
By improving the load data expansion method in the expansion domain, the problem of insufficient data samples in medium and long-term grid load prediction is solved, and the accurate prediction of the grid load capacity is achieved, which improves the stability and accuracy of the prediction model.
Patent Information
- Application Number
- CN202411732655.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-29
- Publication Date
- 2025-05-13
AI Technical Summary
The existing technology faces the problem of insufficient historical data volume in medium- and long-term power grid load prediction, especially the monthly data scale limits the revelation of subtle dynamics and potential laws of load changes, and cannot meet the demand of modern prediction algorithms for large data volume.
The load data expansion method based on the improved expansion domain is adopted. By identifying similar domains and similar week domains, data points are randomly selected to form the load power data on the expansion day, ensuring that the expansion data is consistent with the working week load change rules of the actual data, and quality evaluation and verification are carried out.
It effectively solves the problem of insufficient data samples in medium and long-term load prediction, significantly enhances the breadth and depth of the grid load-voltage prediction data set, improves the stability and accuracy of the prediction model, and provides strong technical support for power system scheduling and energy management.
Smart Images
Figure CN119990384A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of electric load forecasting, and in particular to a load data expansion method and system based on an improved expansion domain. Background Art
[0002] In the practice of power system analysis and grid load forecasting, accurately grasping the changing rules of load and electricity is crucial to optimizing grid operation and improving energy management efficiency. Traditional research and application often rely on long-term accumulated historical data to build prediction models. However, in actual operation, medium- and long-term load forecasting faces the challenge of insufficient historical data, especially when only relying on natural monthly electricity data for analysis. This limitation is particularly prominent. The monthly data scale limits our in-depth exploration of the subtle dynamics and potential rules of load changes.
[0003] The load of power grid is affected by many factors, including but not limited to meteorological conditions, socio-economic activities, holiday effects, etc. These factors show complex interactions and periodic changes in different time periods. Therefore, in order to accurately capture these complex dynamics, the existing monthly data sets are insufficient to fully reveal the full picture of power changes, and cannot meet the needs of modern prediction algorithms for large data volumes, because these algorithms usually require rich data to learn more sophisticated patterns and features, reduce prediction errors, and improve the stability and reliability of the model.
[0004] In view of the above difficulties, there is an urgent need to develop a new data expansion method to scientifically simulate and expand historical data and make up for the technical problem of insufficient original data volume, so as to effectively increase the density and diversity of historical data while maintaining the authenticity and representativeness of the data, and ensure that the expanded data set can more accurately reflect the inherent changes in the power grid load and provide sufficient and high-quality training materials for the algorithm model. Summary of the invention
[0005] In view of the shortcomings of the above-mentioned prior art, the present invention proposes a load data expansion method and system based on an improved expansion domain, which is particularly suitable for accurate prediction of medium- and long-term power grid load and electricity. The method aims to efficiently utilize historical data resources, expand the amount of existing data, and enhance the diversity and representativeness of the model training set, thereby improving the accuracy and generalization ability of the prediction model. The method effectively solves the problem of insufficient sample data in medium- and long-term load forecasting.
[0006] In order to achieve the above object, the present invention provides the following technical solutions:
[0007] A load data expansion method based on improved expansion domain comprises the following steps:
[0008] Acquire historical power grid load power data, wherein the historical power grid load power data at least includes monthly power grid power records for multiple historical years;
[0009] Mark the week type of the extended day according to the predetermined week attribute classification rules;
[0010] Identify similar domains in historical power grid load data, where the dates of the similar domains include M days before and after the expansion date and the same date as the expansion date, and the dates of the similar domains have similar meteorological conditions and social and economic activities as the expansion date;
[0011] Screening out similar weekday domains that match the week type of the expanded day from the similar domains;
[0012] Randomly select data points from the load and power data of similar week domains to form the load and power data of the extended day. The working week extended by the data points must maintain consistency with the load variation law of the actual working week of the extended day data.
[0013] Until the expanded data is generated to form the expanded days for the whole year, a new expanded year data set is formed.
[0014] Furthermore, in step 2, the week types are scheduled to include Monday to Friday, Saturday, and Sunday.
[0015] Furthermore, the similar week domain in step 4 is the load and electricity load on the date with the same week type as the expanded day in the similar domain.
[0016] Furthermore, the step five also includes that the load curve shape of the data point must match the expected load pattern of the expansion day.
[0017] Furthermore, the method further includes step seven, merging the expanded year data set with the historical power grid load and electricity data to form an expanded total data set, and then performing quality assessment and verification on the expanded total data set.
[0018] A system for load data expansion based on improved expansion, comprising:
[0019] A historical power grid load power data unit, used to obtain historical power grid load power data, wherein the historical power grid load power data at least includes monthly power grid power records of multiple historical years;
[0020] A week attribute classification unit, used for marking the week type of the extended day according to a predetermined week attribute classification rule;
[0021] A similar domain determination unit is used to identify similar domains in the historical power grid load power data, wherein the dates of the similar domains include the M days before and after the expansion date and the same date as the expansion date, and the dates of the similar domains have similar meteorological conditions and social and economic activities as the expansion date;
[0022] A week domain determination unit, used to filter out similar week domains matching the week type of the expanded day from the similar domains;
[0023] A randomly selected data unit is used to randomly select data points from the load and power data of a similar week domain to form the load and power data of the extended day. The working week extended by the data points must be consistent with the load variation law of the actual working week of the extended day data.
[0024] The extended data unit is used to receive the load and power data of the extended day sent by the randomly selected data unit and generate extended data for the extended days throughout the year, and combine to form a new extended year data set.
[0025] Furthermore, the week types predetermined in the week attribute classification unit include Monday to Friday, Saturday, and Sunday.
[0026] Furthermore, the similar week domain in the week domain determination unit is the load and electricity load on the date with the same week type as the expanded day in the similar domain.
[0027] Furthermore, data units are randomly selected to match the load curve shape of the data points with the expected load pattern of the expansion day.
[0028] Furthermore, it also includes a quality assessment and verification unit, which is used to combine the expanded year data set with the historical power grid load and power data to form an expanded total data set, and then perform quality assessment and verification on the expanded total data set.
[0029] In the above technical solution, the load data expansion method based on the improved expansion domain provided by the present invention has the following beneficial effects:
[0030] The data expansion method of the present invention is based on the original wind power data expansion method, and innovatively combines the periodicity and time series characteristics of historical load electricity. By identifying the load pattern matching between the predicted day and the historical similar day, and the consistency of the electricity change trend between the predicted month and the historical non-natural month, the effective expansion of historical data is achieved. It effectively solves the problem of insufficient data samples in the field of medium- and long-term power grid load forecasting, provides strong technical support for improving the accuracy of load forecasting and promoting the efficient operation of smart grids, and has important practical value and innovative significance. It not only significantly enhances the breadth and depth of the power grid load electricity forecasting data set, but also effectively improves the stability and accuracy of the forecasting model, providing strong technical support for the fields of power system scheduling, energy management, etc. Compared with the prior art, the present invention shows significant advantages in processing highly dynamic and environmentally dependent data, and provides strong methodological support for the accurate prediction of future electricity demand. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.
[0032] Figure 1 A flow chart of a load data expansion method based on an improved expansion domain provided by the present invention;
[0033] Figure 2 A schematic diagram of a load data expansion method based on an improved expansion domain provided by the present invention;
[0034] Figure 3 This is a schematic diagram of the expanded total data provided by the present invention. DETAILED DESCRIPTION
[0035] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.
[0036] A load data expansion method based on improved expansion domain comprises the following steps:
[0037] Obtain historical power grid load and power data, which at least includes monthly power grid power records for multiple historical years. Specifically, systematically collect and analyze monthly power grid power data for multiple years, and process bad data to provide a data basis for subsequent data expansion.
[0038] The week type of the extended day is marked according to the predetermined week attribute classification rules.
[0039] Among them, the preferred scheduled week types include Monday to Friday, Saturday, and Sunday. Monday to Friday are marked as 1, Saturday is marked as 2, and Sunday is marked as 3.
[0040] Identify similar domains in historical power grid load data. The dates of similar domains include the M days before and after the expansion date and the same date as the expansion date. The dates of similar domains have similar meteorological conditions and social and economic activities as the expansion date. M is a predetermined number of days. Figure 2 The expansion method schematic diagram shown in the figure, during expansion, data expansion is performed in sequence according to the unit length of day, and the similar domain length is selected according to the local load characteristics. The similar domain refers to the M days before and after the expansion date corresponds to the same historical data date.
[0041] Filter out the load and power data of similar week domains that match the week type of the expansion day in the similar domain. Figure 2 As shown in the expansion method diagram, because the load has a certain weekly pattern, it is necessary to finely distinguish the load pattern differences between weekdays and weekends, so each week is divided into three week types: Monday to Friday, Saturday, and Sunday, and the dates of the same week type as the expansion day in the similar domain are divided into similar week domains. The similar week domain is the load and electricity load of the date of the same week type as the expansion day in the similar domain.
[0042] Data points are randomly selected from the load and power data of similar week domains to form the load and power data of the expanded day. The work week expanded by the data points must maintain consistency with the load change law of the actual work week of the expanded day data. Because the load and power changes of the power grid from Monday to Sunday in a work week in the actual load and power data have certain rules, special considerations need to be taken during data expansion to ensure that the work week expanded by the selected data points maintains consistency with the load change law of the actual data work week.
[0043] Preferably, the load curve shape of the data point needs to match the expected load law of the expansion day. When selecting data points, it is not only necessary to check whether the work week expanded by the data point maintains consistency with the load change law of the actual data work week of the expansion day, but also to verify whether the load curve shape of the data point and the expected load law of the expansion day meet the predetermined threshold.
[0044] The random sampling technique is used to select data points from similar week domains to ensure that the expanded data increases the diversity of the data set while maintaining the law of workload variation during the work week.
[0045] Until the extended data is generated for the extended days of the whole year, a new extended year data set is formed. The entire extension process is performed for 365 forecast days throughout the year to form a high-quality extended year data set.
[0046] Preferably, after the expanded year data set is finally merged with the historical power grid load data to form the expanded total data set, the expanded total data set is quality assessed and verified. Strict quality assessment and verification steps are performed to ensure seamless connection between the expanded data set and the original data in terms of quality and consistency, providing a solid foundation for model training.
[0047] The present invention proposes a load data expansion method based on an improved expansion domain, which is particularly suitable for accurate prediction of medium- and long-term power grid load and electricity. The method aims to efficiently utilize historical data resources, expand the amount of existing data, and enhance the diversity and representativeness of the model training set, thereby improving the accuracy and generalization ability of the prediction model. The method effectively solves the problem of insufficient sample data in medium- and long-term load forecasting.
[0048] In order to make the technical solution and advantages of the present invention clearer, the effect of the data expansion method is analyzed by taking the full-caliber electricity consumption in Dalian from 2014 to 2019 as an example.
[0049] The historical year is 100 years.
[0050] Step 1: Deep mining of historical data.
[0051] First, we systematically collect and analyze monthly power consumption data of the power grid over multiple years, and process the bad data to provide a data basis for subsequent data expansion.
[0052] Step 2: Expand data preprocessing.
[0053] Determine the length of the extended data. Here, the extended data is 5 extended days, and mark the extended data with the week type.
[0054] Step 3: Define similar domains. Figure 2 The expansion method schematic diagram shown in the figure, during expansion, data expansion is performed in sequence according to the unit length of day, and the similar domain length is selected according to the local load characteristics.
[0055] S4: Similar week domain definition, dividing the dates of the same week type as the extended day in the similar domain into the similar week domain.
[0056] S5: Expand data synthesis and quality assurance.
[0057] Random sampling technology is used to select data points from similar week domains to ensure that the expanded data increases the diversity of the data set while maintaining the load variation law of the working week. The entire expansion process is performed for 365 expansion days throughout the year to form a high-quality expanded year data set. Finally, through strict quality assessment and verification steps, the seamless connection between the expanded data set and the original data in terms of quality and consistency is ensured, providing a solid foundation for model training. After data expansion, MATLAB software is used to predict load data, and the prediction method used is LSTM neural network. The prediction results are shown in Table 1.
[0058] Table 1 shows the comparison of load power data prediction errors before and after the use of the data augmentation method from 2017 to 2019. Among them, 4.74%, 4.20%, and 1.71% represent the average prediction errors of 2017, 2018, and 2019 before the use of the data augmentation method, respectively, while 4.17%, 3.71%, and 1.39% represent the average prediction errors of the corresponding years after the use of the data augmentation method. Data analysis shows that the prediction error is relatively reduced by 13.0% by using the data augmentation method, indicating that this method helps to improve the accuracy of load power forecasting.
[0059] Table 1 Comparison of forecast results from 2017 to 2019
[0060]
[0061] A system for load data expansion based on improved expansion domain, comprising:
[0062] The historical power grid load and electricity data unit is used to obtain historical power grid load and electricity data. The historical power grid load and electricity data at least includes monthly power grid electricity records of multiple historical years.
[0063] Among them, the data for historical years are at least two years.
[0064] The week attribute classification unit is used to mark the week type of the extended day according to a predetermined week attribute classification rule.
[0065] Among them, the week types predetermined in the preferred week attribute classification unit include Monday to Friday, Saturday, and Sunday. Since the load power of the power grid changes from Monday to Sunday in a working week in the actual load power data has a certain regularity, special consideration is required during data expansion to ensure that the working week expanded from the selected data point maintains consistency with the load change regularity of the actual data working week.
[0066] The similar domain determination unit is used to identify similar domains in historical power grid load and power data. The dates of the similar domains include the M days before and after the expansion date and the same date as the expansion date. The dates of the similar domains have similar meteorological conditions and social and economic activities as the expansion date.
[0067] Here, M is a number of days that can be interactively input into the system.
[0068] The week domain determination unit is used to filter out the load and power data of similar week domains that match the week type of the expansion day in the similar domain, so as to further improve the consistency of the data of the expansion day and the historical data.
[0069] The similar week field in the week field determination unit is a date in the similar field that has the same week type as the expanded day.
[0070] The data unit is randomly selected to randomly select data points from the load and power data of similar week domains to form the load and power data of the extended day. The working week expanded by the data points must maintain consistency with the load variation law of the actual extended day data working week.
[0071] Among them, the data unit is randomly selected to match the load curve shape of the data point with the expected load law of the expansion day. Specifically, when selecting the data point, it is not only necessary to check whether the work week expanded by the data point maintains consistency with the load change law of the actual data work week of the expansion day, but also to verify whether the matching degree of the load curve shape of the data point and the expected load law of the expansion day meets the predetermined threshold.
[0072] The extended data unit is used to receive the load and power data of the extended day sent by the randomly selected data unit. The randomly selected data unit can send the data of the data point to the extended data unit to form the extended data of an extended day. Continue to use the historical power grid load and power data unit, the weekly attribute classification unit, the similar domain determination unit, the weekly domain determination unit and the randomly selected data unit to generate extended data for the extended day of the whole year, and combine to form a new extended year data set. The entire expansion process is executed for 365 forecast days throughout the year to form a high-quality extended year data set.
[0073] Preferably, the system further comprises a quality assessment and verification unit for merging the expanded year data set with the historical power grid load and electricity data to form an expanded total data set, and then performing quality assessment and verification on the expanded total data set.
[0074] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems or computer program products. Therefore, the application can adopt the form of complete hardware embodiment, complete software embodiment, or the embodiment in combination with software and hardware. Moreover, the application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiment of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.
[0075] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0076] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0077] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0078] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.
[0079] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.
[0080] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.
Claims
1. A load data expansion method based on improved expansion domain, characterized in that: include: Acquire historical power grid load power data, wherein the historical power grid load power data at least includes monthly power grid power records for multiple historical years; Mark the week type of the extended day according to the predetermined week attribute classification rules; Identify similar domains in the historical power grid load data, where the dates of the similar domains include the M days before and after the expansion date and the same date as the expansion date, and the dates of the similar domains have similar meteorological conditions and social and economic activities as the expansion date; Screening out similar weekday domains that match the week type of the expanded day from the similar domains; Randomly select data points from the load and power data of similar week domains to form the load and power data of the extended day. The working week extended by the data points must maintain consistency with the load variation law of the actual working week of the extended day data. Until the expanded data is generated to form the expanded days for the whole year, a new expanded year data set is formed.
2. A load data expansion method based on improved expansion domain according to claim 1, characterized in that: The week types are scheduled to include Monday to Friday, Saturday, and Sunday.
3. The load data expansion method based on improved expansion domain according to claim 1, characterized in that: The similar week domain is the load and electricity load on the date with the same week type as the expanded day in the similar domain.
4. The load data expansion method based on improved expansion domain according to claim 1, characterized in that: The load curve shape of the data points must match the expected load pattern on the expansion day.
5. The load data expansion method based on improved expansion domain according to claim 1, characterized in that: It also includes merging the expanded year data set with the historical power grid load and electricity data to form an expanded total data set, and then performing quality assessment and verification on the expanded total data set.
6. A system for load data expansion based on improved expansion domain, characterized in that: include: A historical power grid load power data unit, used to obtain historical power grid load power data, wherein the historical power grid load power data at least includes monthly power grid power records of multiple historical years; A week attribute classification unit, used to mark the week type of the extended day according to a predetermined week attribute classification rule; A similar domain determination unit is used to identify similar domains in the historical power grid load power data, wherein the dates of the similar domains include the M days before and after the expansion date and the same date as the expansion date, and the dates of the similar domains have similar meteorological conditions and social and economic activities as the expansion date; A week domain determination unit, used to filter out similar week domains that match the week type of the expanded day from the similar domains; A randomly selected data unit is used to randomly select data points from the load and power data of a similar week domain to form the load and power data of the extended day. The working week extended by the data points must be consistent with the load variation law of the actual working week of the extended day data. The extended data unit is used to receive the load and power data of the extended day sent by the randomly selected data unit and generate extended data for the extended days throughout the year, and combine to form a new extended year data set.
7. A system for load data expansion based on improved extended domain according to claim 6, characterized in that: The week types predetermined in the week attribute classification unit include Monday to Friday, Saturday, and Sunday.
8. A system for load data expansion based on improved extended domain according to claim 6, characterized in that: The similar week domain in the week domain determination unit is the load and electricity load on the date with the same week type as the expanded day in the similar domain.
9. A system for load data expansion based on improved extended domain according to claim 6, characterized in that: The data units are randomly selected to match the load curve shape of the data points with the expected load pattern of the expansion day.
10. The system for load data expansion based on improved extended domain according to claim 6, characterized in that: It also includes a quality assessment and verification unit, which is used to combine the expanded year data set with the historical power grid load and power data to form an expanded total data set, and then perform quality assessment and verification on the expanded total data set.