Long Term Electricity Consumption Forecast Based on DA-LSTM

Abstract

Electricity consumption is the barometer and weathervane of economic development. In this research, a deep learning long term electricity consumption prediction model based on data enhancement is proposed, and the long term power time series is investigated by using the deep learning method and data enhancement techniques. Firstly, the monthly power quantity is upsampled by interpolation method to generate data with finer granularity, and data points are extracted at equal intervals to form a data series with the same dimension as the original data. Secondly, the augmented data are used as inputs to the deep learning model, so as to allow the deep learning model to have a better generalization ability in the presence of more training data, thus attenuating the over fitting problem of the model. The deep learning model is adopted respectively. LSTM model, Bi-LSTM model, GRU model and MLP model were used. Finally, the model was verified to have a high prediction accuracy using the electricity consumption of urban residents in a province.