Short Term Load Forecasting based on Human Behaviour using Efficient Learning Techniques

Abstract

Electricity load forecasting has become an important issue due to variations in operational efficiency of power systems. Electricity consumption is an important economic index and plays a significant role in drawing up an energy development policy for each country. Accurate load prediction is a challenging task due to non-linear character of time series or varying weather conditions. The accuracy of load forecasting is important for utility companies as well as the consumers. For this reason, it may be necessary to keep on adjusting based on seasons and other factors that may affect the way consumers use the power. In addition, the forecast should rely on accurate data and best forecasting practices. Load forecasting is usually made by constructing models on relative information, such as climate and previous load demand data. In this project we are going to implement a model that predicts the maximum load for next hours of a day on the basis of weather conditions, previous available data and human behavior. Multivariate techniques and time-series analysis have been proposed to deal with electricity consumption forecasting, but a large amount of historical data is required to obtain accurate predictions. LSTM (Long Short Term Memory) based RNN (Recurrent Neural Networks) is able to exploit the long term dependencies in the electric load time series for more accurate forecasting. Experiments will be conducted using several techniques for accurate short term load forecasting and result will be compared to determine most efficient model for STLF.