Forecasting Monthly Electricity Consumption Using Machine Learning and Statistical-Based Models: A Case Study
DOI:
https://doi.org/10.68924/6hnqap90Keywords:
ARIMA, Forecasting, Electricity Consumption Demand, Kunduz Province, MLPAbstract
Forecasting electricity consumption demand plays a vital role in short-term load balancing and long-term planning for having secure power systems, new energy generation, and transmission infrastructures. Knowing the consumption demand is also essential in preparing the power system according to the anticipated load status. In this research paper, we conducted a time series forecasting analysis on the monthly electricity consumption data of Kunduz province gathered from April 2011 to July 2022 to predict the consumption demand in this province. For this purpose, we employed two well-known forecasting methods: Multilayer perceptron (MLP) from machine learning and ARIMA from statistics. The MAPE and MAE metrics were used to measure and compare the predictive performance of developed MLP and ARIMA models. The comparison result found that MLP outperforms ARIMA in model forecasting accuracy. We believe that the outcomes of this research work significantly help Kunduz’s regional electricity authorities in better energy management and allow them to make effective decisions regarding electrical energy cost and efficiency.