Prediction of Electricity Consumption using LSTM and AR models for Uttarakhand

This work proposed two machine learning models, LSTM and Autoregressive models, for predicting electricity consumption in Uttarakhand for the 16 months in terms of forecast validation. Planning of electricity generation & its system development, plays an important role in any state’s infrastructure and economic growth.The two models used are Long Short-Term Memory (LSTM) and Autoregressive (AR) on electricity consumption dataset for 27 months i.e., from Jan 2020 to April 2022 of Uttarakhand. The comparison of both type of models is done based on evaluation metrics. Using LSTM model, the values of MAE, MSE, are 6.278720315297445, 56.14943373418841. The value of RMSE & MAPE are as 7.493292582983025 & 21.523579813749183 respectively, whereas forecast errors i.e., MAE, MSE, RMSE and MAPE for Autoregressive model are 8.177309617471362, 83.8039388215467, 9.154449127148323 and 19.97477081154592, respectively. Therefore, results showed LSTM has lower forecast error and performed better than AR model for the prediction of required data for different purpose.