Method for seven-day prediction of 24-point power load values based on optimized LSTM network

Abstract

The invention provides a method for seven-day prediction of 24-point power load values based on an optimized LSTM network. The method comprises data acquisition, data analysis and data preprocessing.The method comprises the following concrete steps: dividing a data set into a training set and a verification set; based on the training set, conducting parameter optimization on a long-term and short-term memory neural network (LSTM) by using a particle swarm optimization algorithm (PSO); determining optimal values of three parameters, namely the number of hidden layer nodes, a learning rate andthe number of iterations, of the LSTM network suitable for 24-point power load prediction; and with the training set as input, predicting 24-point power load values in the future 7 days based on the optimized LSTM network, comparing power load data output by a model with the verification set, and determining the prediction accuracy of the model by further taking two indexes, namely a mean absoluteerror (MAE) and a root-mean-square error (RMSE) into consideration. By means of the prediction method, prediction accuracy and timeliness are improved, and the prediction effect is better than the prediction effect of an existing power load prediction method.

Description

technical field [0001] The invention relates to the fields of power system planning and scheduling, and in particular to a 7-day prediction method for 24-point power load values ​​based on an optimized LSTM network. Background technique [0002] Accompanied by all aspects of power equipment, power load forecasting has extremely high commercial and research value. Accurately predicting short-term power loads can enable power companies to adjust load equipment in time, reduce waste of resources, and improve performance and stability of power networks. The essence of power load forecasting is to find the implicit relationship between load data sets, use known discrete data to establish a fitting model, and speculate on the data value at a certain moment or in a certain period of time in the future. Short-term power load forecasting technology is generally used to predict the power load in the next day to one week. The accuracy of the forecast directly affects the economic cost...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the deficiencies in the prior art, the present invention proposes a 7-day prediction method for 24-point power load value based on an optimized LSTM network, which only uses time-series data of power load as input, reduces the influence proportion of other factors, and overcomes the existing In the short-term power load forecasting method, the defects of heterogeneous data sources and large differences in dimensions; and optimize the LSTM network through the particle swarm optimization algorithm PSO, determine the parameters of the LSTM network applicable to the 24-point power load forecasting of the present invention, and based on the optimization The LSTM network is used to predict the power load value at 24:00 on the 7th, which improves the accuracy and timeliness of the power load forecasting method, and solves the data source heterogeneity and large dimension difference existing in the short-term power load forecasting technology in the prior art The relationship between time series load data is complicated, the prediction is difficult, and the technical problems of insufficient prediction accuracy and timeliness

Images