Large power consumer electricity demand prediction system based on multi-task large model

Abstract

The invention discloses a multi-task large model-based power consumption demand prediction system for large power consumers, which comprises a power consumption prediction system, the power consumption prediction system comprises a data acquisition unit, and the data acquisition unit is in butt joint with a stable user unit, and relates to the technical field of power consumption demand prediction. According to the multi-task large model-based large power consumer electricity demand prediction system, new and old consumers of the same category are automatically compared, measured and calculated, so that short-term and long-term prediction of electricity consumption of the new consumers can predict whether the electricity consumption of the new consumers is reasonable or not through equal-proportion conversion and comparison with the old consumers of the same category; and the actual operation cycle of a new user is automatically detected through a transition unit, after the normal electricity utilization stable cycle of the new user is reached, the user attribute is automatically transferred, and more detailed and accurate multi-task electricity utilization prediction is carried out on the new user again according to the historical data in the stable cycle and according to the electricity utilization condition of the new user.

Description

Technical Field

[0001] The present invention relates to the technical field of electricity demand forecasting, and in particular to an electricity demand forecasting system for large power users based on a multi-task large model. Background Art

[0002] Power systems must balance power generation and consumption in real time to maintain stable grid operation. By predicting user electricity demand, power dispatchers can plan power generation in advance, ensuring that power supply matches demand and avoiding power shortages or waste caused by supply-demand imbalances.

[0003] Although the existing power demand forecasting system plays an important role in power planning and operation, it still has some shortcomings and limitations: it is inconvenient to reasonably predict the power consumption of new large users, which makes it inconvenient to reasonably judge whether there is any abnormality in the power consumption of new users. Therefore, a power demand forecasting system for large power users based on a multi-task large model is proposed to solve the existing problems. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a power demand forecasting system for large power users based on a multi-task large model, which solves the problem that it is inconvenient to reasonably predict the power consumption of new large users, resulting in inconvenience in reasonably judging whether there is any abnormality in the power consumption of new users.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a large power user electricity demand prediction system based on a multi-task large model, including an electricity demand prediction system, the electricity demand prediction system including a data acquisition unit, the data acquisition unit is connected to a stable user unit, the data acquisition unit is connected to an emerging user unit, the stable user unit and the emerging user unit are both connected to a data classification unit, the data classification unit is connected to a feature extraction unit, the feature extraction unit is connected to a multi-task prediction unit, the multi-task prediction unit is connected to a prediction output unit, the emerging user unit is connected to a transition unit, the transition unit is connected to a stable user unit, the emerging user unit is connected to a comparison unit, the comparison unit is connected to a reference unit, the comparison unit is connected to an abnormality analysis unit, the multi-task prediction unit is connected to the comparison unit, the multi-task prediction unit is connected to the reference unit, and the reference unit is connected to the prediction output unit;

[0006] The data collection unit is used to collect information about the industry category, enterprise scale, production process characteristics, historical electricity consumption data, relevant weather conditions, temperature, etc. of large users.

[0007] The present invention is further configured as follows: the stable user unit is used to add identification to the data of the long-term large electricity users in the data collection unit;

[0008] The emerging user unit is used to add identification to the data of newly settled large electricity users in the data collection unit.

[0009] The present invention is further configured as follows: the data classification unit is used to classify and organize the stable user unit and emerging user unit data into the same category, and the feature extraction unit is used to extract the peak and valley period features of the user's historical power consumption data.

[0010] The present invention is further configured such that: the multi-task forecasting unit forecasts the long-term monthly and annual electricity demand trends of large users, and forecasts the short-term daily and weekly electricity demand fluctuations, based on historical data and macroeconomic information, in combination with real-time meteorological data and short-term economic dynamics;

[0011] The prediction output unit generates a power demand prediction report for large power users;

[0012] The reference unit is used to extract and store the predicted values ​​in the multi-task prediction unit;

[0013] The abnormality analysis unit is used to issue an alarm for abnormal values ​​compared by the comparison unit.

[0014] The present invention is further configured as follows: the comparison unit includes an extraction module, a matching module and a comparison module, the extraction module is connected to the matching module, and the matching module is connected to the comparison module.

[0015] The present invention is further configured such that: the extraction module is configured to receive, as a reference for comparison, long-term and short-term prediction values ​​calculated by the multi-task prediction unit based on actual daily power consumption of emerging user units, and receive, as a reference for comparison, long-term and short-term prediction standard values ​​calculated by the multi-task prediction unit based on historical power consumption of stable user units;

[0016] The matching module matches the predicted standard value and the predicted standard value of the same category in equal proportion according to the industry category, enterprise scale and production process characteristics of the large user;

[0017] The comparison module performs proportional conversion on the predicted standard value and predicted value of the same category in the matching module. If the predicted standard value meets the predicted standard value within the reference range, the normal state can be judged, and then the normal actual power consumption data is fed back to the emerging user unit. Otherwise, the abnormal state is sent to the abnormal analysis unit, and the abnormal state power consumption data is synchronously stopped from being fed back to the emerging user unit.

[0018] The present invention is further configured as follows: the transition unit includes a log generation module, a detection module and a transfer module, the log generation module is connected to the detection module, and the detection module is connected to the transfer module.

[0019] The present invention is further configured as follows: the log generation module is used to receive normal real-time power consumption data transmitted by emerging user units and generate a log;

[0020] The detection module detects the log generation cycle, and when the log cycle does not reach the preset cycle, it is continuously stored in the log generation module;

[0021] After the log cycle reaches the preset cycle, the data is transferred to the stable user unit through the transfer module, and the emerging users in the stable state are converted into stable users.

[0022] Beneficial effects

[0023] The present invention provides a large-scale electricity demand forecasting system for power users based on a multi-task large-scale model. Compared with existing technologies, it has the following advantages:

[0024] (1) The power demand forecasting system for large power users based on a multi-task large model automatically compares and measures new and old users of the same category, so that the short-term and long-term forecasts of the new user's power consumption can be predicted by proportional conversion and comparison with the old users of the same category, and whether the new user's power consumption is reasonable can be predicted. The actual operation cycle of the new user is automatically detected through the transition unit. After the normal power consumption stable cycle is reached, the user attributes are automatically transferred, and the historical data within the stable cycle is re-calculated to be more detailed, accurate and suitable for its own power consumption, so as to perform multi-task power consumption forecasting.

[0025] (2) The power demand forecasting system for large power users based on a multi-task large model can alarm the abnormal prediction values ​​of new users by adopting an abnormal analysis unit, and stop generating logs of abnormal power consumption information, thereby avoiding the problem of abnormal historical power consumption data being entered and affecting subsequent data forecasts. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a system principle block diagram of the present invention;

[0027] Figure 2 This is a system principle block diagram of the comparison unit of the present invention;

[0028] Figure 3 This is a system principle block diagram of the transition unit of the present invention.

[0029] In the figure: 1. Electricity consumption prediction system; 2. Data acquisition unit; 3. Stable user unit; 4. Emerging user unit; 5. Data classification unit; 6. Feature extraction unit; 7. Multi-task prediction unit; 8. Prediction output unit; 9. Transition unit; 901. Log generation module; 902. Detection module; 903. Transfer module; 10. Comparison unit; 101. Extraction module; 102. Matching module; 103. Comparison module; 11. Reference unit; 12. Abnormal analysis unit. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0031] See also Figure 1-3 The present invention provides a technical solution: a large-scale power user electricity demand forecasting system based on a multi-task large model, comprising an electricity forecasting system 1 consisting of an electricity forecasting system 1, a data acquisition unit 2, a stable user unit 3, an emerging user unit 4, a data classification unit 5, a feature extraction unit 6, a multi-task prediction unit 7, a prediction output unit 8, a transition unit 9, a comparison unit 10, a reference unit 11 and an abnormality analysis unit 12, wherein the data acquisition unit 2 is used to collect the industry category, enterprise scale, production process characteristics, historical electricity consumption data, consistent weather, temperature, etc. of the large user.

[0032] Among them, in order to facilitate the classification, feature extraction and long-term and short-term electricity consumption forecasting of user data, the stable user unit 3 is used to identify and add data of long-term large electricity users in the data collection unit 2, and the emerging user unit 4 is used to identify and add data of newly settled large electricity users in the data collection unit 2. The data classification unit 5 is used to classify and organize the data of the stable user unit 3 and the emerging user unit 4 into the same category. The feature extraction unit 6 is used to extract the peak and valley period characteristics of the user's historical electricity consumption data. The multi-task prediction unit 7 is based on historical data and macroeconomic information, and combines real-time meteorological data and short-term economic dynamics to predict the long-term electricity demand trends of large users in terms of monthly and annual consumption, as well as predict the short-term fluctuations in electricity demand for a single day and a single week. The prediction output unit 8 generates a large electricity user electricity demand forecast report. The reference unit 11 is used to extract and store the predicted values ​​in the multi-task prediction unit 7.

[0033] As a preferred solution, in order to facilitate the prediction of the power consumption of emerging users and to make reasonable predictions and judgments on their actual power consumption, the comparison unit 10 includes an extraction module 101, a matching module 102 and a comparison module 103. The extraction module 101 is connected to the matching module 102, and the matching module 102 is connected to the comparison module 103. The extraction module 101 is used to receive the long-term and short-term prediction values ​​calculated by the multi-task prediction unit 7 based on the actual single-day power consumption of the emerging user unit 4 as a reference to be compared, and to receive the long-term and short-term prediction standard values ​​calculated by the multi-task prediction unit 7 based on the historical power consumption of the stable user unit 3 as a reference to be compared. The matching module 102 matches the predicted standard value and the predicted standard value of the same category in equal proportion according to the industry category, enterprise scale and production process characteristics of the large user. The comparison module 103 converts the predicted standard value and the predicted value of the same category in the matching module 102 in equal proportion. If the predicted standard value meets the predicted standard value within the reference range, the normal state can be judged, and then the normal actual electricity consumption data is fed back to the emerging user unit 4. Otherwise, the abnormal state is sent to the abnormal analysis unit 12, and the abnormal state electricity consumption data is synchronously stopped from being fed back to the emerging user unit 4. The abnormal analysis unit 12 is used to alarm the abnormal value compared by the comparison unit 10.

[0034] As a preferred solution, in order to facilitate the transformation of emerging users who use electricity reasonably into stable users, the transition unit 9 includes a log generation module 901, a detection module 902 and a transfer module 903. The log generation module 901 is connected to the detection module 902, and the detection module 902 is connected to the transfer module 903. The log generation module 901 is used to receive normal real-time electricity consumption data transmitted by the emerging user unit 4 and generate a log. The detection module 902 detects the log generation cycle. When the log cycle does not reach the preset cycle, it is continuously stored in the log generation module 901. After the log cycle reaches the preset cycle, the data is transferred to the inside of the stable user unit 3 through the transfer module 903, thereby transforming the emerging user in a stable state into a stable user.

[0035] During use, the stable user unit 3 predicts long-term and short-term electricity consumption data for stable users with historical electricity consumption data through the multi-task prediction unit 7, the prediction output unit 8 and the transition unit 9. The emerging user unit 4 predicts long-term and short-term electricity consumption data for emerging users through the multi-task prediction unit 7, the prediction output unit 8 and the transition unit 9, based on the actual electricity consumption data of a single day. The reference unit 11 extracts the predicted electricity consumption data of stable users as a standard. In the comparison unit 10, the predicted electricity consumption data is matched with the predicted value calculated by the emerging user unit 4 through the multi-task prediction unit 7, and the same category is matched and proportionally converted. If the predicted electricity consumption value of the emerging user meets the range of the predicted value of the stable user, the electricity consumption can be judged to be reasonable, and the reasonable daily electricity consumption data of the emerging user is stored by the transition unit 9. After the stability reaches a preset period, it is uniformly included in the stable users. If the predicted electricity consumption value of the emerging user does not meet the range of the predicted value of the stable user, the abnormal analysis unit 12 can be used to judge the electricity consumption abnormality, and the transmission of the actual unreasonable electricity consumption data of the emerging user to the transition unit 9 is terminated, and an alarm prompt is issued and the cause of the abnormality is analyzed.

Claims

1. A large-scale electricity user electricity demand forecasting system based on a multi-task large model, comprising an electricity demand forecasting system (1), characterized in that: The power consumption prediction system (1) includes a data acquisition unit (2), the data acquisition unit (2) is connected to a stable user unit (3), the data acquisition unit (2) is connected to an emerging user unit (4), the stable user unit (3) and the emerging user unit (4) are both connected to a data classification unit (5), the data classification unit (5) is connected to a feature extraction unit (6), the feature extraction unit (6) is connected to a multi-task prediction unit (7), the multi-task prediction unit (7) is connected to a prediction output unit (8), and the The emerging user unit (4) is connected to the transition unit (9), the transition unit (9) is connected to the stable user unit (3), the emerging user unit (4) is connected to the comparison unit (10), the comparison unit (10) is connected to the reference unit (11), the comparison unit (10) is connected to the abnormality analysis unit (12), the multi-task prediction unit (7) is connected to the comparison unit (10), the multi-task prediction unit (7) is connected to the reference unit (11), and the reference unit (11) is connected to the prediction output unit (8); The data collection unit (2) is used to collect the industry category, enterprise scale, production process characteristics, historical electricity consumption data, corresponding weather conditions, temperature, etc. of large users.

2. The power demand forecasting system for large power users based on a multi-task large model according to claim 1 is characterized by: The stable user unit (3) is used to add identification to the data of the long-term large electricity users in the data collection unit (2); The emerging user unit (4) is used to add identification to the data of newly settled large electricity users in the data collection unit (2).

3. The power demand forecasting system for large power users based on a multi-task large model according to claim 1 is characterized by: The data classification unit (5) is used to classify and sort the data of stable user units (3) and emerging user units (4) into the same category, and the feature extraction unit (6) is used to extract the peak and valley period features of the user's historical power consumption data.

4. The power demand forecasting system for large power users based on a multi-task large model according to claim 1 is characterized by: The multi-task prediction unit (7) predicts the long-term electricity demand trends of large users on a monthly and annual basis, and predicts the short-term electricity demand fluctuations on a daily and weekly basis, based on historical data and macroeconomic information, as well as real-time meteorological data and short-term economic dynamics; The prediction output unit (8) generates a power demand prediction report for large power users; The reference unit (11) is used to extract and store the predicted values ​​in the multi-task prediction unit (7); The abnormality analysis unit (12) is used to issue an alarm for abnormal values ​​compared by the comparison unit (10).

5. The power demand forecasting system for large power users based on a multi-task large model according to claim 1 is characterized by: The comparison unit (10) comprises an extraction module (101), a matching module (102) and a comparison module (103); the extraction module (101) is connected to the matching module (102), and the matching module (102) is connected to the comparison module (103).

6. The power demand forecasting system for large power users based on a multi-task large model according to claim 5, characterized in that: The extraction module (101) is used to receive the long-term and short-term prediction values ​​calculated by the multi-task prediction unit (7) based on the actual daily power consumption of the emerging user unit (4) as a reference for comparison, and to receive the long-term and short-term prediction standard values ​​calculated by the multi-task prediction unit (7) based on the historical power consumption of the stable user unit (3) as a reference for comparison; The matching module (102) matches the predicted standard value and the predicted standard value of the same category in equal proportion according to the industry category, enterprise scale and production process characteristics of the large user; The comparison module (103) performs proportional conversion on the predicted standard value and the predicted value of the same category in the matching module (102). If the predicted standard value meets the predicted standard value within the reference range, the normal state can be judged, and then the normal actual power consumption data is fed back to the emerging user unit (4). Otherwise, the abnormal state is sent to the abnormal analysis unit (12), and the abnormal state power consumption data is fed back to the emerging user unit (4) at the same time.

7. The power demand forecasting system for large power users based on a multi-task large model according to claim 1 is characterized by: The transition unit (9) includes a log generation module (901), a detection module (902) and a transfer module (903), wherein the log generation module (901) is connected to the detection module (902), and the detection module (902) is connected to the transfer module (903).

8. The power demand forecasting system for large power users based on a multi-task large model according to claim 7, characterized in that: The log generation module (901) is used to receive normal real-time power consumption data transmitted by the emerging user unit (4) and generate a log; The detection module (902) detects the log generation cycle, and when the log cycle does not reach the preset cycle, the log is continuously stored in the log generation module (901); After the log cycle reaches the preset cycle, the data is transferred to the interior of the stable user unit (3) through the transfer module (903), and the emerging user in the stable state is converted into a stable user.

Images