Univariant Time Series forecasting of Agriculture load by using LSTM and GRU RNNs

Abstract

In the Energy sector, the Agriculture sector is one of the highest energy consuming sectors. In the Agriculture sector due to the lack of complete metering infrastructure at consumer end, there always remains uncertainty in the metering of actual power consumption at the consumer end, which leads to information asymmetry between the generation and demand-side. This unbalance can risk the grid stability. Along with that, there always remains a non-linear and seasonal behaviour in Agriculture load which also affects the grid stability. To make a balance between generation and demand, forecasting of Agriculture load becomes essential. For Time Series forecasting many conventional models are used such as AR (Auto Regressive) model, MV (Moving Average) model and ARIMA (Auto Regressive integrated moving average) model, but in recent few years, the development and excellent performance of deep learning models like ANN, RNN, LSTM, and GRU have become most feasible for more accurate and precise Time series forecasting. In this paper for Agriculture load forecasting, Long short term Memory (LSTM) RNN and Gated Recurrent Unit (GRU) deep learning models are used for hourly short term Agriculture load forecasting for one month.