Power consumption probability prediction method based on neural network

Abstract

The invention discloses an electric power consumption probability prediction method based on a neural network, and the method comprises the following steps: collecting historical data of electric power consumption, dividing the historical data into a training set and a test set, and carrying out the normalization processing of all variables; constructing a neural network model based on a convolutional architecture and a self-attention mechanism; training a neural network model by using the processed training set data, and selecting a model with the best prediction precision as a trained neural network model by using the test set; recent data of power consumption are selected and preprocessed, the preprocessed recent data are input into the model, and an output value of the model is subjected to inverse normalization processing to obtain a probability prediction result. Compared with a traditional power load prediction method, the method has the advantages that modeling of power consumption data of different users in a power grid is achieved at the same time by means of the constructed neural network model, short-term and long-term modes in a time sequence can be captured, high-precision prediction of the time sequence is achieved, and a point prediction result and a probability prediction result are output.

Description

technical field [0001] The invention relates to the field of power consumption probability prediction, and more particularly, to a power consumption probability prediction method based on a neural network. Background technique [0002] In recent years, with the rapid development of deep learning, many methods based on neural network for time series prediction have appeared one after another: DeepAR model based on long short-term memory network (LSTM), temporal convolution based on convolutional neural network (CNN) Network (Temporal Convolution Networks, TCN), Transformer model based on attention mechanism (Attention). These existing deep learning-based methods have achieved good forecasting accuracy on time series forecasting tasks. [0003] The above deep learning-based time series prediction architecture also has some problems. The recurrent neural network (RNN) has the problem of gradient disappearance and gradient explosion, and LSTM cannot model too long sequences. F...

AI Technical Summary

Problems solved by technology

[0006] In order to overcome the defects of the above-mentioned existing power consumption prediction technology, such as the weak ability to capture time series features, the inability to process the power consumption data of multiple users at the same time, and the inability to quantify the prediction results, the present invention provides a neural network-based Electric Power Consumption Probability Prediction Method

Images