LSTM Based Short-term Electricity Consumption Forecast with Daily Load Profile Sequences

Abstract

For energy-related services and researches, not only the energy load data in the past but also the future are essential. In this paper, a short-term electricity consumption prediction method is proposed. The method utilizes Long-Short-Term-Memory (LSTM) network which takes a sequence of past consumption profiles to perform a month-ahead electricity consumption prediction as a sequence. For performance analysis, an experiment with a real dataset is done, and the experimental result validates that the proposed method performs well with the prediction accuracy of about 82.5%. The test accuracy can be improved with a longer period of training time and deliberate hyperparameter setting.