A method, system, medium, device and product for predicting power load

Abstract

The present invention discloses a method, system, medium, equipment and product for predicting electric load, which belongs to the technical field of electric load prediction. The method includes: obtaining data features required for multiple electric load predictions and load labels corresponding to each feature; selecting the best data feature from all data features, the process includes: randomly selecting multiple data features to form multiple candidate feature subsets; calculating the fitness of each candidate feature subset; selecting the data feature in the candidate feature subset with the largest fitness as the best data feature; wherein the fitness of the candidate feature subset includes the average value of the correlation coefficient between each feature in the candidate feature subset and its corresponding load label, minus the average correlation coefficient between each feature in the candidate feature subset; using the best data feature, training the constructed electric load prediction model, and after the training is completed, obtaining the trained electric load prediction model. The training accuracy and efficiency of the electric load prediction model are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power load forecasting, and in particular to a power load forecasting method, system, medium, equipment and product. 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] Currently, the main method is to obtain relevant data for power load prediction, and use the relevant data to train and optimize the constructed power load prediction model. The trained power load prediction model is used to predict the power load. Due to the different correlations between the relevant data and the power load prediction results, the power load prediction model trained using all the relevant data does not have the optimal performance, and the model calculation volume is large, and the model prediction efficiency is low.

[0004] In the related technology, there is a method of calculating the correlation between each relevant data and its corresponding load forecast result, selecting the most relevant data from the relevant data for training and optimization of the power load forecasting model, thereby improving the prediction accuracy and prediction efficiency of the power load forecasting model.

[0005] However, when the relevant data are screened in the related technology, only the correlation between each relevant data and its corresponding load forecasting result is considered, and the redundancy between the relevant data is not considered, resulting in that the prediction accuracy and prediction efficiency of the finally trained power load forecasting model cannot reach the optimal level. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a power load forecasting method, system, medium, equipment and product, which improve the efficiency and accuracy of power load forecasting.

[0007] To achieve the above object, the present invention adopts the following technical solution:

[0008] Firstly, a power load forecasting method is proposed, comprising:

[0009] Obtain the data features required for various power load forecasts and the load labels corresponding to each feature;

[0010] Selecting the best data feature from all data features, the process includes: randomly selecting multiple data features to form multiple candidate feature subsets; calculating the fitness of each candidate feature subset; selecting the data feature in the candidate feature subset with the largest fitness as the best data feature; wherein the fitness of the candidate feature subset includes the average value of the correlation coefficient between each feature in the candidate feature subset and its corresponding load label, minus the average correlation coefficient between each feature in the candidate feature subset;

[0011] The constructed power load forecasting model is trained by using the optimal data features, and after the training is completed, a trained power load forecasting model is obtained;

[0012] The trained power load forecasting model is used to forecast the power load.

[0013] Furthermore, a genetic algorithm is used to select the best data feature from all data features, wherein an initial population of the genetic algorithm includes a plurality of candidate feature subsets formed by randomly selecting a plurality of data features.

[0014] Furthermore, the process of selecting the optimal data feature from all data features through genetic algorithm is as follows:

[0015] (1) Determine the initial population, and each candidate feature subset is used as an individual in the initial population;

[0016] (2) Use the fitness function to determine the fitness of each individual in the population;

[0017] (3) Select individuals with high fitness as parents;

[0018] (4) Perform crossover and mutation operations on the selected parent generation to form offspring;

[0019] (5) Merge the parent generation and the offspring generation to form a new generation population, and repeat (2)-(5) until the termination condition is met and the optimal individual is obtained;

[0020] (6) Decode the optimal individual to obtain the optimal data features.

[0021] Furthermore, the termination condition is that the number of iterations reaches a set maximum number of iterations, or the fitness reaches a preset fitness value.

[0022] Furthermore, the data features required for various power load forecasting are obtained, including week characteristics, month characteristics, time characteristics, historical load characteristics, temperature characteristics, historical temperature characteristics, first-order derivatives and second-order derivatives of historical loads, and first-order derivatives and second-order derivatives of historical temperatures.

[0023] Furthermore, the power load forecasting model takes the features in the optimal data features as input and the power load forecasting results as output, and is constructed using LSSVM, ELM, BP or LSTM.

[0024] In the second aspect, a power load forecasting system is proposed, comprising:

[0025] A data acquisition unit, used to acquire data features required for various power load forecasts and load labels corresponding to each feature;

[0026] The feature screening unit is used to select the best data feature from all data features, and the process includes: randomly selecting multiple data features to form multiple candidate feature subsets; calculating the fitness of each candidate feature subset; selecting the data feature in the candidate feature subset with the largest fitness as the best data feature; wherein the fitness of the candidate feature subset includes the average value of the correlation coefficient between each feature in the candidate feature subset and its corresponding load label, minus the average correlation coefficient between each feature in the candidate feature subset;

[0027] The model training unit is used to train the constructed power load prediction model using the optimal data features, and after the training is completed, a trained power load prediction model is obtained; and the power load is predicted using the trained power load prediction model.

[0028] In a third aspect, a computer device is provided, the device comprising:

[0029] a processor adapted to execute a computer program;

[0030] A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by the processor, an electric power load forecasting method proposed in the first aspect is implemented.

[0031] In a fourth aspect, a computer-readable storage medium is proposed, wherein the computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor and executing a power load forecasting method proposed in the first aspect.

[0032] In a fifth aspect, a computer program product is proposed, which includes a computer program. When the computer program is executed by a processor, it implements the power load forecasting method proposed in the first aspect.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The present invention proposes a method, system, medium, equipment and product for power load prediction. When determining the optimal data feature, the method randomly selects multiple data features to form multiple candidate feature subsets; calculates the fitness of each candidate feature subset; selects the data feature in the candidate feature subset with the largest fitness as the optimal data feature; wherein the fitness of the candidate feature subset includes the average value of the correlation coefficient between each feature in the candidate feature subset and its corresponding load label, minus the average correlation coefficient between each feature in the candidate feature subset, that is, when selecting the optimal data feature through fitness selection, not only the correlation between each data feature and the corresponding load label is considered, but also the correlation between each data feature is considered, so that the feature set with high correlation with the load label and low redundancy between features can be determined as the optimal data feature, and the optimal data feature can not only effectively predict the target value, but also reduce the negative impact of redundant features on model performance; when the optimal data feature is used to train the power load prediction model, the training accuracy and efficiency of the power load prediction model are improved, and ultimately the accuracy and efficiency of power load prediction are improved.

[0035] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings in the specification, which constitute a part of the present application, are used to provide further understanding of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute improper limitations on the present application.

[0037] Figure 1 A flowchart of a power load forecasting method disclosed in an embodiment;

[0038] Figure 2 The present invention is a flow chart of selecting the optimal data features for the genetic algorithm disclosed in the embodiment. DETAILED DESCRIPTION

[0039] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0040] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. 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 application belongs.

[0041] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0042] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0043] Example 1

[0044] In the field of big data and machine learning, feature selection is one of the key steps to improve model performance and efficiency. As the data dimension increases, feature selection becomes more and more important. Although there are many existing feature selection methods, they still have some limitations.

[0045] When faced with complex data on power loads and related influencing factors, traditional feature selection methods are unable to fully explore the entire feature space. In the optimization process, they can only find local optimal solutions and cannot perform global searches to find the optimal feature set. This is mainly because these methods have limited traversal and optimization mechanisms in the search space and lack a global perspective. For example, feature selection based on statistics or information theory metrics will result in the inability to fully consider the interactions and interdependencies between features when evaluating and selecting features, resulting in only suboptimal solutions.

[0046] Secondly, when processing high-dimensional, nonlinear data, traditional methods often find it difficult to fully explore the complex relationships between features, resulting in limited prediction performance. There may be complex nonlinear relationships and redundancy between features in high-dimensional data, and traditional linear methods such as principal component analysis (PCA) and linear discriminant analysis (LDA) are difficult to effectively capture this complexity. Therefore, the development of feature selection methods that can achieve global optimization and process high-dimensional, nonlinear data has become an urgent need to improve the performance and efficiency of machine learning models. In order to solve the above problems, this embodiment proposes a method for power load forecasting. By introducing a genetic algorithm, a method with global search capabilities, combined with a dual evaluation strategy of a fitness function, it can effectively overcome the limitations of traditional methods and provide a better feature selection solution.

[0047] Specifically, this embodiment proposes a method for predicting power load, such as Figure 1 As shown, including:

[0048] Obtain the data features required for various power load forecasts and the load labels corresponding to each feature;

[0049] Selecting the best data feature from all data features, the process includes: randomly selecting multiple data features to form multiple candidate feature subsets; calculating the fitness of each candidate feature subset; selecting the data feature in the candidate feature subset with the largest fitness as the best data feature; wherein the fitness of the candidate feature subset includes the average value of the correlation coefficient between each feature in the candidate feature subset and its corresponding load label, minus the average correlation coefficient between each feature in the candidate feature subset;

[0050] The constructed power load forecasting model is trained by using the optimal data features, and after the training is completed, a trained power load forecasting model is obtained;

[0051] The trained power load forecasting model is used to forecast the power load.

[0052] The data features required for various power load forecasting obtained in this embodiment include week features, month features, time features, historical load features, temperature features, historical temperature features, first-order derivatives and second-order derivatives of historical loads, and first-order derivatives and second-order derivatives of historical temperatures.

[0053] Specifically, the data features required for various power load forecasts include week characteristics, month characteristics, time characteristics, load characteristics at the same time from 1 day to 15 days ago, temperature characteristics, temperature characteristics at the same time from 1 day to 3 days ago, the first and second derivatives of the load at the same time of the previous day, and the first and second derivatives of the temperature, a total of 26 data characteristics.

[0054] This embodiment uses a genetic algorithm to select the optimal data feature from all data features, wherein the initial population of the genetic algorithm includes a plurality of candidate feature subsets formed by randomly selecting a plurality of data features.

[0055] like Figure 2 As shown in the figure, the process of selecting the optimal data feature from all data features through genetic algorithm is:

[0056] (1) Determine the initial population, and each candidate feature subset is regarded as an individual in the initial population.

[0057] Specifically, the initial population includes multiple individuals, each individual is a candidate feature subset, and each candidate feature subset is represented by a binary string, wherein 1 indicates that the candidate feature subset selects the data feature, and 0 indicates that the candidate feature subset does not select the data feature.

[0058] (2) Use the fitness function to determine the fitness of each individual in the population.

[0059] The fitness function plays a vital role in the genetic algorithm. It evaluates the quality of each individual (feature subset) and is the basis for the evolution and optimization of the genetic algorithm. Therefore, defining an effective fitness function is a key step to ensure that the genetic algorithm can find the global optimal feature subset. In order to comprehensively evaluate the quality of the candidate feature subset, the fitness function proposed in this embodiment simultaneously calculates the correlation between the candidate feature subset and the corresponding load label and the correlation between the features in the candidate feature subset. By calculating the correlation coefficient between the candidate feature subset and the corresponding load label, the relevance of the candidate feature subset to the target is evaluated; by calculating the average correlation coefficient between the features in the candidate feature subset, the redundancy of the candidate feature subset is evaluated; by simultaneously evaluating the relevance of the candidate feature subset to the target and the redundancy of the candidate feature subset, the optimal data features with high correlation and low redundancy are selected.

[0060] The fitness function of this embodiment takes into account the redundancy between features in the candidate feature subset while evaluating the correlation between the candidate feature subset and the target variable. The fitness function is:

[0061]

[0062] In the formula, Fitness is the fitness, n represents the number of features in the candidate feature subset X, and x i and x j represents the i-th feature and the j-th feature in the candidate feature subset X, corr(Y,x i ) represents the target value Y and feature x i The correlation coefficient between i ,x j ) represents the feature x i and feature x j The correlation coefficient between k and B k Respectively represent the values ​​of two variables A and B at the kth observation value, and They represent the means of variables A and B respectively, and m represents the number of observations.

[0063] (3) Select individuals with high fitness as parents.

[0064] This embodiment selects the two individuals with the highest fitness as parents.

[0065] (4) Perform crossover and mutation operations on the selected parent generation to form the offspring generation.

[0066] In this embodiment, the gene fragments of two selected parents are exchanged at random crossover points to perform a crossover operation; the mutation rate is set, that is, each gene of each individual has a certain probability of mutation, and the corresponding gene is reversed according to the set mutation rate, thereby generating new individual offspring.

[0067] (5) Merge the parent generation and the offspring generation to form a new generation population. Repeat (2)-(5) until the termination condition is met and the optimal individual is obtained.

[0068] In this embodiment, the parent generation and the offspring generation are merged to form a new generation of candidate feature subsets, which replaces the old population and repeats the above process until the termination condition is met, and the individual with a high fitness value is output as the optimal individual, and the candidate feature subset corresponding to the individual is the optimal candidate feature subset.

[0069] This embodiment defines the termination condition as the number of iterations reaching a set maximum number of iterations, or the fitness reaching a preset fitness value.

[0070] (6) Decode the optimal individual to obtain the optimal data features.

[0071] After obtaining the optimal data features, this embodiment uses the optimal data features to train the constructed power load prediction model. After the training is completed, a trained power load prediction model is obtained to predict the power load.

[0072] Among them, the power load forecasting model takes the features in the optimal data features as input and the power load forecasting results as output. It is constructed using LSSVM (least squares support vector machine), ELM (extreme learning machine), BP (back propagation neural network) or LSTM (long short-term memory network).

[0073] This embodiment selects the whole social load data of a province in 2022 as sample one and the industrial load data as sample two for load forecasting. The data are sampled every 15 minutes to verify the disclosed method of this embodiment.

[0074] Obtain the data features required for various power load forecasts and form an initial data set, including the weekly feature W k , monthly characteristics M, time characteristics t, historical load characteristics L at the same time from the previous day to the previous 15 days t,d-1 , ..., L t,d-15 , temperature characteristics T, temperature characteristics T at the same time from 1 day to 3 days ago t,d-1 , T t,d-2 , T t,d-3 , the first-order derivative of the load at the same time the previous day L t ' ,d-1 and the second-order derivative L t ″,d-1 and the first derivative of temperature T t ' ,d and the second-order derivative T t ″ ,d , a total of 26 data features.

[0075] In order to select the optimal data features for load forecasting, for all data features, first, an initial population containing multiple candidate feature subsets is generated, where each candidate feature subset is represented by a binary string, 1 means the data feature is selected, and 0 means the data feature is not selected; then, the correlation coefficient between each selected candidate feature subset and the load data of the day is calculated. At the same time, the average correlation coefficient between the features within the subset is calculated, and then the fitness of each candidate feature subset is calculated; then, the population is sorted according to the fitness, and the elite selection strategy is adopted, that is, the individuals with the top 20% fitness are selected as parents; two individuals with higher fitness are selected as parents, and their gene fragments are exchanged at random intersections for crossover operation; the mutation rate is set to 0.1, that is, each gene of each individual has a 10% probability of mutation, and the corresponding gene is reversed during mutation, thereby generating new individual offspring; the parent and offspring are merged to form a new generation of candidate feature subsets, replacing the old population and repeating the above process until the termination condition is met, and the optimal data feature with a high fitness value is output. The final selection results are shown in Table 1.

[0076] Table 1 Optimal data features

[0077]

[0078] For sample one and sample two, all features and optimal data features are used as inputs of the prediction model, respectively, and the final prediction evaluation results are shown in Table 2. Compared with all features, the performance of feature extraction by the method disclosed in this embodiment on both sample one and sample two is significantly improved. Specifically, for sample one, compared with all features, the RMSE (root mean square error) of the method disclosed in this embodiment is reduced by 132.72MW, the MAE (mean absolute error) is reduced by 120.59MW, the MAPE (mean absolute percentage error) is reduced by 1.83%, and the R2 (coefficient of determination) is increased by 0.73; for sample two, compared with all features, the RMSE of the method disclosed in this embodiment is reduced by 53.85MW, the MAE is reduced by 47.39MW, the MAPE is reduced by 0.85%, and the R2 is increased by 0.20.

[0079] Table 2 Prediction evaluation results

[0080]

[0081] The example results show that when the method disclosed in this embodiment performs optimal data feature selection, it can adaptively search in the entire feature space through the fitness function, discover feature combinations of nonlinear relationships and complex interactions, comprehensively consider external influencing factors and historical load changes, and select the optimal feature set with global optimality, high correlation and low redundancy, and achieve more accurate predictions and better fitting effects in load forecasting.

[0082] The present embodiment discloses a method for predicting power load, which performs optimal data feature selection based on a genetic algorithm. The genetic algorithm can explore and optimize in a huge search space by simulating the process of natural selection and genetic variation, thus overcoming the limitation that traditional feature selection methods can only obtain local optimal solutions, thereby finding the globally optimal feature set.

[0083] At the same time, genetic algorithms are particularly suitable for processing high-dimensional and nonlinear data. Through their diverse individual representations and mutation operations, they can fully explore the complex relationships between features, so that they can effectively identify and select features that contribute significantly to prediction performance in these complex data environments.

[0084] This embodiment discloses a method for predicting power loads, in which the fitness function can select features that are closely related to the target value and filter out features that are less closely related to the target value by evaluating the correlation between the candidate feature subset and the target value, thereby ensuring that the selected features have the greatest contribution to predicting the target value. By calculating the correlation between the features in the candidate feature set and considering the redundancy between the features, it is possible to avoid selecting features with high correlation and reduce redundant information in the feature set. This dual evaluation strategy can more accurately measure the importance of features when selecting features, thereby selecting a feature set that significantly contributes to the model prediction performance.

[0085] Example 2

[0086] In this embodiment, a power load forecasting system is disclosed, comprising:

[0087] A data acquisition unit, used to acquire data features required for various power load forecasts and load labels corresponding to each feature;

[0088] The feature screening unit is used to select the best data feature from all data features, and the process includes: randomly selecting multiple data features to form multiple candidate feature subsets; calculating the fitness of each candidate feature subset; selecting the data feature in the candidate feature subset with the largest fitness as the best data feature; wherein the fitness of the candidate feature subset includes the average value of the correlation coefficient between each feature in the candidate feature subset and its corresponding load label, minus the average correlation coefficient between each feature in the candidate feature subset;

[0089] The model training unit is used to train the constructed power load prediction model using the optimal data features, and after the training is completed, a trained power load prediction model is obtained; and the power load is predicted using the trained power load prediction model.

[0090] The present invention also discloses a computer device, which includes:

[0091] a processor adapted to execute a computer program;

[0092] A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by the processor, a method for predicting electric load disclosed in Example 1 is implemented.

[0093] The present invention also discloses a computer-readable storage medium, which stores a computer program. The computer program is suitable for being loaded by a processor and executing a power load forecasting method disclosed in Example 1.

[0094] The present invention also discloses a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements a power load forecasting method disclosed in Example 1.

[0095] The method disclosed in Example 1 can be directly embodied as a hardware processor, or can be executed by a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.

[0096] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0097] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for predicting power load, characterized in that: include: Obtain the data features required for various power load forecasts and the load labels corresponding to each feature; Selecting the best data feature from all data features, the process includes: randomly selecting multiple data features to form multiple candidate feature subsets; calculating the fitness of each candidate feature subset; selecting the data feature in the candidate feature subset with the largest fitness as the best data feature; wherein the fitness of the candidate feature subset includes the average value of the correlation coefficient between each feature in the candidate feature subset and its corresponding load label, minus the average correlation coefficient between each feature in the candidate feature subset; The constructed power load forecasting model is trained by using the optimal data features, and after the training is completed, a trained power load forecasting model is obtained; Use the trained power load forecasting model to forecast the power load; The data features required for the various power load forecasts obtained include week features, month features, time features, historical load features, temperature features, historical temperature features, first-order derivatives and second-order derivatives of historical loads, and first-order derivatives and second-order derivatives of historical temperatures; The fitness function not only evaluates the correlation between the candidate feature subset and the target variable, but also considers the redundancy between the features in the candidate feature subset. The fitness function is: In the formula, Fitness is the fitness, n represents the number of features in the candidate feature subset X, and x i and x j represents the i-th feature and the j-th feature in the candidate feature subset X, corr(Y,x i ) represents the target value Y and feature x i The correlation coefficient between i ,x j ) represents the feature x i and feature x j The correlation coefficient between k and B k Respectively represent the values ​​of two variables A and B at the kth observation value, and They represent the means of variables A and B respectively, and m represents the number of observations.

2. A method for predicting power load according to claim 1, characterized in that: A genetic algorithm is used to select the optimal data feature from all data features, wherein an initial population of the genetic algorithm includes a plurality of candidate feature subsets formed by randomly selecting a plurality of data features.

3. A method for predicting power load according to claim 2, characterized in that: The process of selecting the optimal data features from all data features through genetic algorithms is as follows: (1) Determine the initial population, and each candidate feature subset is used as an individual in the initial population; (2) Use the fitness function to determine the fitness of each individual in the population; (3) Select individuals with high fitness as parents; (4) Perform crossover and mutation operations on the selected parent generation to form offspring; (5) Merge the parent generation and the offspring generation to form a new generation population, and repeat (2)-(5) until the termination condition is met and the optimal individual is obtained; (6) Decode the optimal individual to obtain the optimal data features.

4. A method for predicting power load according to claim 3, characterized in that: The termination condition is that the number of iterations reaches the set maximum number of iterations, or the fitness reaches the preset fitness value.

5. A method for predicting power load according to claim 1, characterized in that: The power load forecasting model takes the features in the optimal data features as input and the power load forecasting results as output, and is constructed using LSSVM, ELM, BP or LSTM.

6. A power load forecasting system, using a power load forecasting method as claimed in any one of claims 1 to 5, characterized in that: include: A data acquisition unit, used to acquire data features required for various power load forecasts and load labels corresponding to each feature; The feature screening unit is used to select the best data feature from all data features, and the process includes: randomly selecting multiple data features to form multiple candidate feature subsets; calculating the fitness of each candidate feature subset; selecting the data feature in the candidate feature subset with the largest fitness as the best data feature; wherein the fitness of the candidate feature subset includes the average value of the correlation coefficient between each feature in the candidate feature subset and its corresponding load label, minus the average correlation coefficient between each feature in the candidate feature subset; The model training unit is used to train the constructed power load prediction model using the optimal data features, and after the training is completed, a trained power load prediction model is obtained; and the power load is predicted using the trained power load prediction model.

7. An electronic device, characterized in that: The device comprises: a processor adapted to execute a computer program; A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by the processor, an electric power load forecasting method according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor and executing a power load forecasting method according to any one of claims 1-5.

9. A computer program product, characterized in that The computer program product includes a computer program, and when the computer program is executed by a processor, it implements the power load forecasting method according to any one of claims 1 to 5.

Images