Power load prediction method based on deep learning
Through deep learning of power load prediction method, the improved Autoformer model and sliding window method are used to combine environmental characteristics to predict power load, which solves the problem of low prediction accuracy in the existing technology and improves the operating efficiency and stability of the power system.
Patent Information
- Application Number
- CN202411155455.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-22
- Publication Date
- 2024-11-22
AI Technical Summary
The existing power load prediction methods cannot accurately predict based on the actual situation of the surrounding environment, such as temperature, humidity, holidays and other characteristics, resulting in low prediction accuracy.
Using a deep learning-based power load prediction method, the collection and preprocessing of historical power load data is used, and the improved Autoformer model is used for training, combining the sliding window method to capture the long-term dependencies in the time series, and the sequence decomposition block extraction of long-term trends and Periodic characteristics, predict future power loads, and add a prediction alarm module to the prediction results.
Accurate prediction of future power loads is achieved, and the operation efficiency and stability of the power system are improved.
Smart Images

Figure CN119010002A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power system load forecasting, and in particular to a power load forecasting method based on deep learning. Background Art
[0002] As the power system continues to develop, it is becoming increasingly important to the economic development of society. With the continuous advancement of power grid technology and the increase in economic and social demand for electricity, power energy services have now covered many fields. In this context, the normal operation of the power grid system is very important, and power demand forecasting is of great significance to the operation of the power grid system. The complexity and variability of the power grid system determine that the power load forecasting needs to have strong adaptability and high accuracy. The current mainstream power load forecasting method is based on deep learning methods, which selects and optimizes models and adjusts parameters to obtain better power load forecasting results. In the existing technology, the load forecasting function of power data can be realized through the load forecasting system.
[0003] There are the following problems with the existing technology: The existing prediction methods are unable to predict future power loads based on the actual conditions of the surrounding environment, such as temperature, humidity, holidays and other characteristics, and the prediction accuracy is low. Therefore, a power load prediction method based on deep learning is proposed. Summary of the invention
[0004] The purpose of the present invention is to provide a power load forecasting method based on deep learning to solve the problems raised in the above background technology.
[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: The method for predicting power load based on deep learning includes the following steps: Step S1, data preprocessing: Step S101: Collect a data set consisting of historical power load data, and perform data cleaning to remove outliers and noise. Each data set sample includes five indicators: whether it is a holiday, external environment temperature value, humidity value, power load value of the day, and average power load value of the next three days; Step S102: normalizing the data set to unify the value range of the data set to improve the training effect of the model; Step S2, model training: Step S201: input the preprocessed data set into the Autoformer model for training. The Autoformer model consists of two parts: an encoder and a decoder. The encoder is responsible for extracting features from the input sequence, and the decoder predicts future loads based on these features. Step S202: During the training process, a sliding window method is used to generate training samples so that the model can capture long-term dependencies in the time series; Step S3, load forecasting: Step S301: input the latest power load data into the trained model to predict the power load situation in the future; Step S302: Denormalize the prediction result to obtain the actual load value.
[0006] As a further improvement of the present technical solution: in the step S201: when the Autoformer model is trained using the data set, the Autoformer model uses four pre-processed indicators, namely, whether it is a holiday, the external environment temperature value, the humidity value, and the power load value of the day, as input, and uses the average power load value of the next three days as output.
[0007] As a further improvement of the present technical solution: in the step S201: the preprocessed data is divided into a training set and a validation set, the training set is input into the Autoformer model for training, and the validation set is used to evaluate the performance accuracy of the Autoformer model.
[0008] As a further improvement of the technical solution: whether the data set is a holiday, the input value is 1, and if not, the input value is 0.
[0009] As a further improvement of the present technical solution: for the need to predict the power load situation in the future, firstly measure whether the day is a holiday, the external environment temperature value, humidity value, and the power load value of the day, standardize them and input them into the Autoformer model to obtain the prediction result of the average power load value in the next three days.
[0010] As a further improvement of the technical solution: an improved sequence decomposition block is introduced into the encoder, so that the Autoformer model can better extract long-term trends and periodic features in the time series.
[0011] As a further improvement of this technical solution: multiple training samples are generated through a sliding window method, which enhances the model's ability to learn long-term dependencies in time series.
[0012] As a further improvement of the technical solution: it also includes a prediction alarm module, which is used to predict that when the average power load value data in the next three days exceeds the expected load, timely alarm and send out a prompt signal.
[0013] Due to the adoption of the above technical solution, the present invention has the following technical advances compared with the prior art: The present invention provides a power load forecasting method based on deep learning, which adopts an improved Autoformer model. Through the analysis and modeling of historical load data, accurate prediction of future power load is achieved, effectively improving the operating efficiency and stability of the power system. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 It is a schematic diagram of the prediction method flow of the present invention; Figure 2 The Autoformer model structure diagram of the present invention; Figure 3 A schematic diagram of training samples generated for the sliding window of the present invention. DETAILED DESCRIPTION
[0015] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention. The present invention is described in more detail by way of example with reference to the accompanying drawings in the following paragraphs. It should be noted that the accompanying drawings are all in a very simplified form and are not in precise proportions, and are only used to facilitate and clearly assist in explaining the purpose of the embodiments of the present invention.
[0016] like Figure 1-3 As shown, the present invention provides a method for predicting power load based on deep learning, which includes the following steps: Step S1, data preprocessing: Step S101: Collect a data set consisting of historical power load data, and perform data cleaning to remove abnormal values and noise. Each data set sample includes five indicators: whether it is a holiday, external environment temperature value, humidity value, power load value of the day, and average power load value of the next three days. In the data set, if it is a holiday, the input value is 1, and if not, the input value is 0; Step S102: normalizing the data set, taking the data set value as (-1, 1) to unify the value range of the data set, so as to improve the training effect of the model; Step S2, model training: Step S201: input the preprocessed data set into the Autoformer model for training. When the Autoformer model is trained using the data set, the Autoformer model uses four indicators, namely, whether it is a holiday, the external environment temperature value, the humidity value, and the power load value of the day, as input, and uses the average power load value of the next three days as output; the preprocessed data is divided into a training set and a validation set, the training set is input into the Autoformer model for training, and the validation set is used to evaluate the performance accuracy of the Autoformer model; like Figure 2 As shown in the figure, the Autoformer model consists of two parts: an encoder and a decoder. The encoder contains several self-attention mechanism layers and sequence decomposition blocks, which are used to extract long-term trends and periodic features in the input sequence. The encoder introduces an improved sequence decomposition block, which enables the Autoformer model to better extract long-term trends and periodic features in the time series. The decoder predicts future loads based on these features. Step S202: During the training process, a sliding window method (such as Figure 3 As shown in the figure, training samples are generated. Each sample contains historical load data of a fixed length, which enables the model to capture long-term dependencies in the time series. Multiple training samples are generated by the sliding window method, which enhances the model's ability to learn long-term dependencies in the time series. Step S3, load forecasting: Step S301: For the power load situation that needs to be predicted for a period of time in the future, firstly measure whether the day is a holiday, the external environment temperature value, humidity value, and the power load value of the day, and input them into the Autoformer model after standardization to obtain the prediction result of the average power load value in the next three days; Step S302: Denormalize the prediction result to obtain the actual load value.
[0017] The present invention also comprises a prediction alarm module, which is used for timely alarming and sending out prompt signals when predicting that the average load value data of the power in the next three days exceeds the expected load.
[0018] In summary, the present invention provides a power load forecasting method based on deep learning, which adopts an improved Autoformer model. By analyzing and modeling historical load data, accurate prediction of future power load is achieved, effectively improving the operating efficiency and stability of the power system.
[0019] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any ordinary technician in the industry can smoothly implement the present invention as shown in the drawings and described above. However, any equivalent changes, modifications and evolutions made by technicians familiar with the profession without departing from the scope of the technical solution of the present invention using the technical content disclosed above are all equivalent embodiments of the present invention. At the same time, any equivalent changes, modifications and evolutions made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the technical solution of the present invention.
Claims
1. A power load forecasting method based on deep learning, characterized by: The power load forecasting method based on deep learning includes the following steps: Step S1, data preprocessing: Step S101: Collect a data set consisting of historical power load data, and perform data cleaning to remove outliers and noise. Each data set sample includes five indicators: whether it is a holiday, external environment temperature value, humidity value, power load value of the day, and average power load value of the next three days; Step S102: normalizing the data set to unify the value range of the data set to improve the training effect of the model; Step S2, model training: Step S201: input the preprocessed data set into the Autoformer model for training. The Autoformer model consists of two parts: an encoder and a decoder. The encoder is responsible for extracting features from the input sequence, and the decoder predicts future loads based on these features. Step S202: During the training process, a sliding window method is used to generate training samples so that the model can capture long-term dependencies in the time series; Step S3, load forecasting: Step S301: input the latest power load data into the trained model to predict the power load situation in the future; Step S302: Denormalize the prediction result to obtain the actual load value.
2. The method for predicting power load based on deep learning according to claim 1, characterized in that: In step S201: when the Autoformer model is trained using the data set, the Autoformer model uses four pre-processed indicators, namely, whether it is a holiday, the external environment temperature value, the humidity value, and the power load value of the day, as input, and uses the average power load value of the next three days as output.
3. The method for predicting power load based on deep learning according to claim 2 is characterized in that: In the step S201: the preprocessed data is divided into a training set and a validation set, the training set is input into the Autoformer model for training, and the validation set is used to evaluate the performance accuracy of the Autoformer model.
4. The method for predicting power load based on deep learning according to claim 3 is characterized in that: Whether the data set is a holiday, the input value is 1 if it is, and 0 if it is not.
5. The method for predicting power load based on deep learning according to claim 4 is characterized in that: In step S301: for the power load situation that needs to be predicted for a period of time in the future, firstly measure whether the day is a holiday, the external environment temperature value, humidity value, and the power load value of the day, standardize them and input them into the Autoformer model to obtain the prediction result of the average power load value in the next three days.
6. The method for predicting power load based on deep learning according to claim 5, characterized in that: An improved sequence decomposition block is introduced into the encoder, so that the Autoformer model can better extract long-term trends and periodic features in time series.
7. The method for predicting power load based on deep learning according to claim 6, characterized in that: Multiple training samples are generated through the sliding window method, which enhances the model's ability to learn long-term dependencies in time series.
8. The method for predicting power load based on deep learning according to claim 7 is characterized in that: It also includes a prediction alarm module, which is used to predict that when the average power load value data in the next three days exceeds the expected load, it will issue a timely alarm and send out a prompt signal.
