Learning Approach for Energy Consumption Forecasting in Residential Microgrid

Abstract

Residential energy consumption plays an important role in the social and economic development of the country. Highly accurate forecasting can aid in decision making and forecast for future residential electricity demand for smooth management of power system operations. However, residential load characteristics are influenced by human behavior, seasonal variation, and other social factors. Thus the share of uncertainty in the load will be at a significant level. Therefore, obtaining highly accurate load forecasts is a challenging task for the power system operator. In this research article, the authors propose a recurrent neural network based LSTM, GRU, Bi-LSTM, and Bi-GRU based learning approach for short-term residential energy consumption forecasting. Simulation results on a real 30 minute time interval energy consumption data set for 9 months of a residential prosumer microgrid located in central-Norway. The numerical results are show that the Bi-GRU model is achieving higher performance than others on the given load data set.