Power load short-term prediction method and system

Abstract

The invention relates to the technical field of power load prediction, in particular to a power load short-term prediction method and system. Acquiring power load data, meteorological data and social activity data influencing the power load, and performing preprocessing to form a data set; obtaining a daily average load value # imgabs0 # corresponding to the to-be-predicted day by using an artificial intelligence model; determining a power load reference curve of the same to-be-predicted day by using an artificial experience method; and calculating a daily average load value # imgabs1 # in the power load reference curve, taking a ratio of # imgabs2 # to # imgabs3 # as a curve adjustment proportion # imgabs4 #, correcting a load value corresponding to each time point in the power load reference curve, and taking the corrected load value as a prediction result. The AI algorithm and the artificial experience are combined, so that the load curve predicted by the artificial experience is corrected, and the trend of power load change can be better reflected.

Description

Technical Field

[0001] The present invention relates to the technical field of power load forecasting, and in particular to a method and system for short-term power load forecasting. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Short-term power load forecasting refers to a method of predicting changes in power system load in the next few hours or the next day. Traditional manual forecasting methods rely on rich experience and can comprehensively consider a variety of factors that affect load changes. As the volatility and uncertainty of power load increase during short-term forecasting, existing technologies have begun to use AI algorithms for short-term power load forecasting. AI algorithms have demonstrated strong capabilities in data processing and pattern recognition. AI algorithms are more process-oriented and consider more factors than manual forecasting, and the resulting forecast mean results are more credible, but they are slightly insufficient in the shape of the load curve, and it is still difficult to completely replace the unique experience accumulated by manual experience in complex system analysis and decision-making. Summary of the invention

[0004] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a method and system for short-term prediction of power load. Through the powerful data analysis capability of AI algorithm, the potential laws between power load and various influencing factors (such as weather, holidays, economic activities, etc.) are deeply excavated. At the same time, the prediction results of AI algorithm are verified and optimized by combining manual practical experience and professional knowledge. By combining the wisdom of human brain with the computing power of computer, the "computerization" of human brain wisdom and more accurate and reliable short-term load prediction are realized.

[0005] In order to achieve the above object, the present invention adopts the following technical solution: A first aspect of the present invention provides a method for short-term forecasting of electric load, comprising the following steps: Obtain power load data, meteorological data, and data on social activities that affect power load, and form a data set after preprocessing; The obtained data set uses the trained artificial intelligence model to obtain the daily average load value corresponding to the predicted day ; The obtained data set uses artificial experience method to determine the reference curve of power load on the same day to be predicted; According to the power load values ​​corresponding to all time points in the power load reference curve, calculate the daily average load value corresponding to the reference curve ,by and The ratio of the curve is used as the adjustment ratio , use the obtained curve to adjust the scale , correct the load value corresponding to each time point in the power load reference curve, and take the corrected load value as the prediction result.

[0006] Furthermore, the social activity data that affects the power load includes at least holiday information, production start and stop information, and electricity price adjustment information. The social activity data that affects the power load is used to determine the date type of the predicted day as at least one of a working day, a rest day, a compensatory holiday, a major holiday, and a day affected by economic activities.

[0007] Furthermore, preprocessing includes at least one of missing data filling, abnormal data detection and processing, multi-source data fusion and cleaning, and filtering and standardization processing.

[0008] Furthermore, the power load reference curve for the day to be predicted is determined by using an artificial experience method, including the following steps: Obtain power load data, meteorological data, and data on social activities that affect power load; Determine the daily accumulated precipitation on the day to be predicted based on meteorological data; Determine the date type of the day to be predicted based on social activity data; When the daily cumulative precipitation on the day to be predicted is not greater than the set value X, a day with the same date type as the day to be predicted and the greatest similarity in perceived temperature is selected from all the days in the same month of the previous year as the day to be predicted and the first 30 days of the year of the day to be predicted, and the power load curve of the selected day is obtained as the first artificial experience output curve; When the daily cumulative precipitation of the day to be predicted is greater than the set value X, the day with the same date type as the day to be predicted and the greatest precipitation similarity is selected from the days with daily cumulative precipitation of all months in the current year of the day to be predicted and the previous year that exceeds the set value X, and the power load curve of the selected day is obtained as the second artificial experience output curve; The obtained first artificial experience output curve or the second artificial experience output curve is the reference curve of the power load on the day to be predicted.

[0009] Furthermore, the power load reference curve of the day to be predicted reflects the power load data every 15 minutes and has power load values ​​corresponding to 96 time points.

[0010] Furthermore, the similarity is maximized, specifically: the root mean square error is minimized.

[0011] Furthermore, according to the power load values ​​corresponding to all time points in the power load reference curve, the daily average load value corresponding to the reference curve is calculated. , as shown below: ; Where, N is the number of load points, which is 96 points; is the load at the i-th point at the corresponding moment.

[0012] Further, and The ratio of the curve is used as the adjustment ratio , use the obtained curve to adjust the scale , the load value corresponding to each time point in the corrected power load reference curve is used as the prediction result, as shown in the following formula: , ; In the formula, is the daily average load value corresponding to the reference curve, is the daily average load value corresponding to the day to be predicted, is the load at the moment corresponding to the i-th point in the reference curve.

[0013] Furthermore, the artificial intelligence model uses historical power load data, historical meteorological data, and data on social activities that affect power load as training sets, and obtains the daily average load value corresponding to the predicted day through optimization and training. .

[0014] A second aspect of the present invention provides a short-term power load forecasting system, comprising: The data acquisition preprocessing module is configured to: acquire power load data, meteorological data, and data of social activities that affect power load, and form a data set through preprocessing; The AI ​​prediction module is configured to: use the trained AI model to obtain the daily average load value corresponding to the predicted day. ; The artificial experience prediction module is configured to: determine the power load reference curve of the same to-be-predicted day using the artificial experience method with the obtained data set; The fusion processing module is configured to: calculate the daily average load value corresponding to the reference curve according to the power load values ​​corresponding to all time points in the power load reference curve ,by and The ratio of the curve is used as the adjustment ratio , use the obtained curve to adjust the scale , correct the load value corresponding to each time point in the power load reference curve, and take the corrected load value as the prediction result.

[0015] Compared with the prior art, one or more of the above technical solutions have the following beneficial effects: Considering that the average result predicted by AI (artificial intelligence algorithm) is more credible, but the result is slightly insufficient in the curve shape, and the curve shape can reflect the changing trend of power load, so that the prediction results obtained by AI algorithm have the characteristics of uneven distribution. This characteristic is mainly manifested in: the predicted load in different periods deviates greatly from the actual load locally, and the local deviation is small. In this case, the overall deviation of the predicted load is large and the accuracy is relatively low. On the one hand, the artificial experience method looks for a reference curve that is closer to the day to be predicted, such as the same holiday, the same working day, or the same date affected by economic activities, or the similarity of precipitation and the similarity of body temperature. On the other hand, it finds the influence of the load characteristics of the time period on the accuracy of the load of the whole network to realize the prediction of the load, which has more advantages in the overall load change trend. Therefore, according to the ratio of the average power load predicted by AI to the average load of the artificial experience reference curve, the reference curve is adjusted for amplitude fluctuation, and the shape of the day to be predicted and the reference curve is regarded as consistent. After adjusting the curve shape, the predicted load can become relatively more uniform (the 96-point predicted load is basically coincident or parallel to the actual load). When the accuracy of the predicted average load value is close to the actual load average, the adjusted predicted load curve will almost coincide with the actual load (if there is no special situation such as extreme weather or social events on the predicted day), and the accuracy will be greatly improved; even if there are some deviations in the average, the two curves are relatively parallel, and there will be no extreme deviations at any time of the day, and the accuracy will still be improved, which means that the load curve obtained by manual experience can better reflect the changing trend of power load after correction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0017] Figure 1 It is a schematic diagram of a short-term power load forecasting process provided by one or more embodiments of the present invention. DETAILED DESCRIPTION

[0018] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0019] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0020] The following example provides a method for short-term power load forecasting. Through the powerful data analysis capabilities of the AI ​​algorithm, the potential laws between power load and various influencing factors (such as weather, holidays, economic activities, etc.) are deeply explored. At the same time, the prediction results of manual experience are verified and optimized by combining manual practical experience and professional knowledge. By combining the wisdom of the human brain with the computing power of the computer, the "computerization" of the human brain wisdom and more accurate and reliable short-term load forecasting are achieved.

[0021] Embodiment 1: The short-term forecasting method of electric load includes the following steps: Obtain power load data, meteorological data, and data on social activities that affect power load, and form a data set after preprocessing; The obtained data set uses the trained artificial intelligence model to obtain the daily average load value corresponding to the predicted day ; The obtained data set uses artificial experience method to determine the reference curve of power load on the same day to be predicted; According to the power load values ​​corresponding to all time points in the power load reference curve, calculate the daily average load value corresponding to the reference curve ,by and The ratio of the curve is used as the adjustment ratio , use the obtained curve to adjust the scale , correct the load value corresponding to each time point in the power load reference curve, and take the corrected load value as the prediction result.

[0022] like Figure 1 As shown, the short-term power load prediction method in this embodiment includes a part for obtaining the predicted load value through an AI algorithm and a part for obtaining the load curve shape trend through artificial experience. By combining the two parts, the predicted value obtained by the AI ​​algorithm is corrected using artificial experience, and the corrected predicted value is used as the final output prediction result.

[0023] The process of obtaining the predicted load value by AI algorithm (artificial intelligence algorithm) includes the following steps: Data preparation: Acquire data to form data sets, including but not limited to meteorological data, load data, production start and stop information, and economic activities, and divide them into load data sets and meteorological data sets according to data types; Preprocess the data, including missing data filling, abnormal data detection and processing, multi-source data fusion and cleaning, as well as data filtering and standardization; Use pre-processed load data, meteorological data, and production start and stop, production start and stop, and economic activity data as input data for AI algorithms; When the AI ​​algorithm obtains the predicted load value, the input data is segmented and constructed into a training set and a test set. By constructing a load prediction model and optimizing and adjusting the model, the predicted output of the load series is obtained.

[0024] During the period when the load forecasting model obtains the predicted load value, the factors affecting the load forecast are analyzed by performing correlation analysis on the load data set and the meteorological data set, and the characteristic factors are extracted. The extracted characteristic factors are numerically processed, and the required influencing factors are selected according to the factor size of the load and different influencing factors. The same analysis can also be performed on time factors such as seasons, date types, major holidays, and policy factors such as time-of-use electricity prices and production start and stop to obtain the corresponding influencing factors.

[0025] In terms of data set segmentation, a random segmentation method is used to divide the data set into training set, test set, and validation set in proportion for model training, and cross-validation technology is used to evaluate the model performance to obtain preliminary load forecasting results.

[0026] Implement hyperparameter tuning of the model, continuously optimize the model and feature set based on feedback information, continuously improve the model's prediction capability by increasing the number of training samples and making them closer to the characteristics of the date to be predicted and the algorithm's self-learning mechanism, and achieve initial optimization of power load prediction using artificial intelligence methods.

[0027] like Figure 1 As shown, the part of predicting the load curve shape trend by manual experience includes the following steps: Obtain basic data. The resolution of power load data and meteorological data in the basic data is 15 minutes. The social activity data in the basic data includes information such as major holidays, production start and stop, electricity price policy adjustments, and economic activities; Determine the daily accumulated precipitation on the day to be predicted based on meteorological data; When the daily cumulative precipitation of the day to be predicted is not greater than the set value, the date type of the day to be predicted is determined according to the social activity data, and the day with the same date type as the day to be predicted and the greatest similarity in body temperature is selected from all the dates of the same month of the previous year of the day to be predicted and the first 30 days of the year of the day to be predicted, and the power load curve of the selected day is obtained as the first artificial experience output curve; When the daily cumulative precipitation of the day to be predicted is greater than the set value, the date type of the day to be predicted is determined according to the social activity data, and the day with the same date type as the day to be predicted and the greatest precipitation similarity is selected from the dates whose daily cumulative precipitation of all months in the current year and the previous year exceeds the same set value, and the power load curve of the selected day is obtained as the second artificial experience output curve; The obtained first artificial experience output curve or the second artificial experience output curve is recorded as a reference curve. The reference curve reflects the power load data every 15 minutes to form a 96-point power load reference curve.

[0028] In this embodiment, by acquiring power load data with a resolution of 15 minutes, the acquired load data is cleaned and preprocessed to construct a load data set.

[0029] In this embodiment, meteorological data influencing the load, such as temperature, humidity, wind speed, precipitation, air pressure, irradiance and weather type, are acquired with a resolution of 15 minutes, and the acquired meteorological data are cleaned and preprocessed to construct a meteorological data set.

[0030] In this embodiment, a social activity data set is constructed by acquiring data such as major holidays, production starts and stops, electricity price policy adjustments, and economic activities.

[0031] According to the corresponding calculation principles, the perceived temperature and daily accumulated precipitation involved in load forecasting are calculated respectively; The calculation principle of body temperature is as follows: ; In the formula, Indicates temperature; Indicates humidity; Indicates the perceived temperature.

[0032] The calculation principle of daily cumulative precipitation is: obtain the measured meteorological data of a single meteorological station, and calculate the daily cumulative precipitation according to the 96-point cumulative measured precipitation / N (N is the number of measured meteorological stations). The daily cumulative precipitation of the day to be predicted is obtained by prediction based on the already obtained meteorological data. The method of predicting precipitation is a common method in the meteorological field, and this embodiment will not be repeated.

[0033] When multiple meteorological stations provide measured meteorological data, the average precipitation data of multiple meteorological stations at each time is determined first, and then the daily cumulative precipitation is calculated based on the 96-point cumulative precipitation / N.

[0034] According to the obtained daily cumulative precipitation, determine the selection method of the reference curve for the day to be predicted; The principle of determining the consistency of date types is: the date type of the day to be predicted and the date type of the reference curve are the same as working days, rest days, adjusted holidays, major holidays and economic activity impact days; When the accumulated precipitation on the day is not more than 10 mm, the similarity of the perceived temperature between the predicted day and the reference curve selection interval date is calculated based on the method of perceived similarity. The date with the maximum similarity (minimum root mean square error) and the same date type as the predicted day is determined as the perceived similarity reference curve of the predicted day based on the principle of consistent date types. The interval for selecting the similar reference curve of somatosensory sensation is determined as follows: all dates of the same month in the previous year and 30 days before the first day of the year to be predicted.

[0035] In this embodiment, 10 mm is selected based on manual experience.

[0036] When the accumulated precipitation on the day is greater than 10mm (moderate rain and above), the similarity of precipitation between the day to be predicted and the reference curve selection interval is calculated based on the precipitation similarity method. The date with the maximum similarity (minimum root mean square error) and the same date type as the day to be predicted is determined as the precipitation similarity reference curve for the day to be predicted based on the principle of consistency of date types. The interval for selecting the precipitation similarity reference curve is determined as: the date before the forecast date and the date when the daily cumulative precipitation in all months of the same season in the year before the forecast date and the previous year exceeds 10 mm.

[0037] Obtain 96 points of power load data of the somatosensory similarity reference curve or precipitation similarity reference curve of the day to be predicted, and record it as the reference curve; the obtained reference curve reflects the predicted load every 15 minutes on the prediction day, and all time in a day is divided into 96 data points, namely, the 96-point power load of the reference curve.

[0038] Combining manual experience and AI technology to determine the short-term forecast results of power load, including the following steps: The reference curve obtained by artificial experience is used as the predicted reference curve, recorded as the load shape curve, and the daily average load of the load shape curve is calculated. , ; In the formula, is the daily average load for calculating the shape curve; N is the number of points of the whole network load, which is 96 points; is the total network load at the i-th point at the corresponding moment.

[0039] The daily average load value is predicted by artificial intelligence algorithms (such as CNN-LSTM), which is recorded as the predicted load mean. , and calculate the curve adjustment ratio for: That is, the ratio between the daily average load calculated using the artificial experience reference curve and the daily average load predicted by the AI ​​algorithm is used as the curve adjustment ratio.

[0040] Based on the 96-point full network load of the reference curve and the curve adjustment ratio, the adjusted forecast load is calculated. for: ; That is, using the obtained adjustment ratio Correct the predicted load value of each point in the artificial experience reference curve.

[0041] In this embodiment, according to the obtained power load data, the time periods of a day are divided into: morning peak period (00:15-9:00), noon peak period (9:15-17:00) and evening peak period (17:15-24:00).

[0042] In this embodiment, the load characteristics are analyzed, including the load characteristics analysis at different time periods within a day and the load characteristics analysis at the same time period during different date types.

[0043] The load characteristics analysis at different time periods during the day is as follows: draw the load curves for all dates in the season of the day to be predicted, and observe the load variation patterns at different time periods of the day.

[0044] The load characteristics analysis of the same period during the day for different date types is as follows: draw the load curves of all dates in the season of the day to be predicted, and observe the changing rules of the daily loads of the considered date types.

[0045] The final forecast load of the forecast day is obtained through the analysis of load characteristics in different time periods based on artificial experience. , as shown below: .

[0046] The core idea of ​​the above method is that the average load of the day to be predicted predicted by the artificial intelligence algorithm is equivalent to the predicted electricity volume of the day to be predicted. This result comprehensively considers many factors (such as various types of weather, time span, historical load, etc.). This method is more process-oriented than manual prediction, takes more factors into consideration, and is less affected by human subjective assumptions. Therefore, the predicted average result of the artificial intelligence algorithm is more credible, but the result is slightly insufficient in the shape of the curve. This deficiency is mainly reflected in the fact that the actual load of 96 points in the entire network obtained with a resolution of 15 minutes has different amplitude deviations from the predicted load in different time periods, resulting in the curve being locally close to the actual load and locally deviating from the actual load at the same time, making the shape of the predicted load curve slightly less consistent with the shape of the actual load curve, and it is difficult to reflect the changing trend of the power load.

[0047] The key point of the artificial experience method is, on the one hand, to find a reference curve that is closer to the day to be predicted. The approximation can be the date type (for example, the same holiday, the same working day, or the same date affected by economic activities), the similarity of precipitation and the similarity of body temperature. On the other hand, find the impact of time-divided load characteristics on the accuracy of the entire network load.

[0048] According to the ratio of the power generation predicted by the artificial intelligence algorithm to the power generation (load mean) of the reference curve, the reference curve is adjusted for amplitude fluctuations, and the shapes of the day to be predicted and the reference curve are considered to be consistent. This is mainly due to the periodicity of the shape of the load of the entire network, which is affected by various factors such as time, season, and weather. The cluster analysis (K-Means) method is used to calculate the results of meteorological similarity, date type similarity, and time span, and the load curve with the highest meteorological similarity and date type similarity and the closest date to the day to be predicted is selected as the reference curve for the day to be predicted. Its curve shape will also be closer to the load shape of the day to be predicted, further increasing the accuracy of the predicted load curve shape.

[0049] In addition, since the daily load changes in different time periods are different, the load characteristics of different time periods are analyzed by combining manual experience summary, data mining and other means to further obtain the impact of date type and the load in the first 1 / 3 of the day on the load of the entire network.

[0050] In summary, the method of integrating artificial experience and AI technology can not only use artificial intelligence methods to obtain accurate electricity (load) for the day to be predicted, but also obtain the shape of the load for the day to be predicted based on artificial experience, and obtain a more accurate predicted load for the day to be predicted through the fusion process.

[0051] In this embodiment, the acquired power load data is the load data of the entire network.

[0052] When cleaning and preprocessing data, we often face the complexity of power load and meteorological data. These data features are highly volatile, have multiple noise points, and are nonlinear. Useful signals and noise signals are intertwined, resulting in uneven quality of load sample data, which directly affects the accuracy of load forecasting. In addition, affected by automation systems, data acquisition systems, etc., data missing and abnormal values ​​caused by data mutations often appear in load and meteorological sample data, which obscures the true change pattern of historical load and meteorology. To this end, this embodiment adopts a data preprocessing strategy based on "cross-period comparison and time series filling".

[0053] For missing data: If a single value is missing, interpolation is performed by taking the mean of the two values ​​before and after the current data; If 2-3 values ​​are missing consecutively, linear interpolation is used; When three or more consecutive values ​​are missing, the load data of the same date type and time last week of the current date will be used as a substitute.

[0054] Abnormal data correction: A tree-based unsupervised anomaly detection algorithm, the isolation forest algorithm, is used to construct a decision tree by randomly selecting features and values. The anomaly scores of all data points are calculated and the scatter plots are drawn to comprehensively select the data points with higher anomaly scores and in an isolated state to determine the outliers.

[0055] Identify outliers in the data set and use interpolation to correct missing data to achieve the purpose of correction, thereby maintaining data consistency and reliability.

[0056] The corrected data was verified by drawing a data box plot to compare the distribution characteristics before and after the data correction to ensure that all outliers had been correctly processed and that the overall distribution of the data had not changed significantly.

[0057] Data filtering: Apply Savitzky-Golay filtering to the data. Savitzky-Golay filtering smoothes the data by fitting a local polynomial to the data points, removing noise and unnecessary fluctuations in the data, making the data smoother and easier to analyze.

[0058] Standardization: Given the significant scale differences between data features, it is necessary to implement scale unification measures in the data preprocessing stage. This process uses the Min-Max Scaling method to adjust the scale of the data to the range of 0 to 1 to ensure that each feature is comparable and consistent in the subsequent analysis and comparison process. The processing method of the standardized value is: Normalized value = (original value - minimum value) / (maximum value - minimum value).

[0059] Numericalization of non-numeric data: When processing non-numeric data, if you encounter information such as date type, convert it into a label encoding format and use it as an input feature in the model training process. This conversion method specifically maps the day of the week in the date (from Monday to Sunday) to numbers 1 to 7, so that the model can recognize and effectively use these date features for training.

[0060] It is further necessary to explain that the Spearman correlation coefficient analysis technique is used to explore the key factors affecting the performance of various load forecasting models and extract characteristic factors accordingly. This process involves calculating the correlation coefficient between the load characteristics and each potential influencing factor, thereby identifying the factors with the strongest influence on load changes.

[0061] It is further noted that a customized convolutional neural network (CNN) architecture is used, which incorporates parallel attention branches to efficiently extract key significant features. On this basis, the time dependency in the action sequence is modeled by combining the capabilities of the long short-term memory network (LSTM). By introducing the attention mechanism into the CNN-LSTM model, the fusion of coarse-grained to fine-grained features is achieved, thereby comprehensively and deeply capturing the dynamic characteristics of time series data, and finally building a time series short-term load forecasting model based on the CNN-LSTM method of the attention mechanism.

[0062] It should be further explained that in the study of short-term load forecasting methods, the acquisition of manual experience is crucial. The specific ways to obtain manual experience are: mainly through interviews with experts in multiple locations, in-depth understanding of the characteristics of power markets in different regions and the laws of load changes; at the same time, questionnaires are designed to widely collect insights and predictions on load change trends from inside and outside the industry. In addition, data analysis and mining technology is used to extract potential laws and trends that are difficult for experts to intuitively capture from historical load data.

[0063] Embodiment 2: The short-term power load forecasting system includes: The data acquisition preprocessing module is configured to: acquire power load data, meteorological data, and data of social activities that affect power load, and form a data set through preprocessing; The AI ​​prediction module is configured to: use the trained AI model to obtain the daily average load value corresponding to the predicted day. ; The artificial experience prediction module is configured to: determine the power load reference curve of the same to-be-predicted day using the artificial experience method with the obtained data set; The fusion processing module is configured to: calculate the daily average load value corresponding to the reference curve according to the power load values ​​corresponding to all time points in the power load reference curve ,by and The ratio of the curve is used as the adjustment ratio , use the obtained curve to adjust the scale , correct the load value corresponding to each time point in the power load reference curve, and take the corrected load value as the prediction result.

[0064] Through the powerful data analysis capabilities of AI algorithms, we can deeply explore the potential rules between power load and various influencing factors (such as weather, holidays, economic activities, etc.), and at the same time combine manual practical experience and professional knowledge to verify and optimize the prediction results of manual experience. By combining the wisdom of the human brain with the computing power of the computer, we can achieve the "computerization" of human brain wisdom and more accurate and reliable short-term load forecasting.

[0065] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A short-term power load forecasting method, characterized in that: The following steps are involved: Obtain power load data, meteorological data, and data on social activities that affect power load, and form a data set after preprocessing; The obtained data set uses the trained artificial intelligence model to obtain the daily average load value corresponding to the predicted day ; The obtained data set uses artificial experience method to determine the reference curve of power load on the same day to be predicted; According to the power load values ​​corresponding to all time points in the power load reference curve, calculate the daily average load value corresponding to the reference curve ,by and The ratio of the curve is used as the adjustment ratio , use the obtained curve to adjust the scale , correct the load value corresponding to each time point in the power load reference curve, and take the corrected load value as the prediction result.

2. The method for short-term power load forecasting according to claim 1, characterized in that: The social activity data that affect the power load include at least holiday information, production start and stop information, and electricity price adjustment information. The social activity data that affect the power load are used to determine the date type of the predicted day as at least one of a working day, a rest day, a compensatory holiday, a major holiday, and a day affected by economic activities.

3. The method for short-term power load forecasting according to claim 1, characterized in that: Preprocessing includes at least one of missing data filling, abnormal data detection and processing, multi-source data fusion and cleaning, and filtering and standardization processing.

4. The method for short-term power load forecasting according to claim 1, characterized in that: Using artificial experience method, the power load reference curve of the forecast day is determined, including the following steps: Obtain power load data, meteorological data, and data on social activities that affect power load; Determine the daily accumulated precipitation on the day to be predicted based on meteorological data; Determine the date type of the day to be predicted based on social activity data; When the daily cumulative precipitation on the day to be predicted is not greater than the set value X, a day with the same date type as the day to be predicted and the greatest similarity in perceived temperature is selected from all the days in the same month of the previous year as the day to be predicted and the first 30 days of the year of the day to be predicted, and the power load curve of the selected day is obtained as the first artificial experience output curve; When the daily cumulative precipitation of the day to be predicted is greater than the set value X, the day with the same date type as the day to be predicted and the greatest precipitation similarity is selected from the days with daily cumulative precipitation of all months in the current year of the day to be predicted and the previous year that exceeds the set value X, and the power load curve of the selected day is obtained as the second artificial experience output curve; The obtained first artificial experience output curve or the second artificial experience output curve is the reference curve of the power load on the day to be predicted.

5. The method for short-term power load forecasting according to claim 1, characterized in that: The power load reference curve of the day to be predicted reflects the power load data every 15 minutes and has power load values ​​corresponding to 96 time points.

6. The method for short-term power load forecasting according to claim 1, characterized in that: The similarity is the largest, specifically: the root mean square error is the smallest.

7. The method for short-term power load forecasting according to claim 1, characterized in that: According to the power load values ​​corresponding to all time points in the power load reference curve, calculate the daily average load value corresponding to the reference curve , as shown below: ； Where N is the number of time points of the load; is the load at the i-th point at the corresponding moment.

8. The method for short-term power load forecasting according to claim 1, characterized in that: by and The ratio of the curve is used as the adjustment ratio , use the obtained curve to adjust the scale , the load value corresponding to each time point in the corrected power load reference curve is used as the prediction result, as shown in the following formula: ， ； In the formula, is the daily average load value corresponding to the reference curve, is the daily average load value corresponding to the day to be predicted, is the load at the moment corresponding to the i-th point in the reference curve.

9. The method for short-term power load forecasting according to claim 1, characterized in that: The artificial intelligence model uses historical power load data, historical meteorological data, and data on social activities that affect power load as training sets. Through optimization and training, it obtains the daily average load value corresponding to the day to be predicted. .

10. A short-term power load forecasting system implementing the method according to any one of claims 1 to 9, characterized in that: include: The data acquisition preprocessing module is configured to: acquire power load data, meteorological data, and data of social activities that affect power load, and form a data set through preprocessing; The AI ​​prediction module is configured to: use the trained AI model to obtain the daily average load value corresponding to the predicted day. ; The artificial experience prediction module is configured to: determine the power load reference curve of the same to-be-predicted day using the artificial experience method with the obtained data set; The fusion processing module is configured to: calculate the daily average load value corresponding to the reference curve according to the power load values ​​corresponding to all time points in the power load reference curve ,by and The ratio of the curve is used as the adjustment ratio , use the obtained curve to adjust the scale , correct the load value corresponding to each time point in the power load reference curve, and take the corrected load value as the prediction result.

Images