A statistical and predictive modeling study to analyze impact of seasons and covid-19 factors on household electricity consumption

Abstract

Load is dynamic in nature and changing from aggregated load to disaggregated loads. Hence, need to analyze individual household’s energy consumption pattern. Many factors are contributing to household electricity consumption (HEC). The most influencing factor is the end user’s behavioral aspect. The calendar and seasonal factors are directly affecting user’s behavior activities. This paper consists of two aim, first aim is to validate the performance of traditional predictive models and second aim is to identify the best-fitted predictive model from five predictive models namely: Random Forest, Linear Regression, Support Vector Machine, Neural Network (NN) and Adaptive Boosting. The orange tool is used to simulate the predictive models. The JASP tool is used for statistical analysis of the dataset. From the predictive modeling study, the NN model is the most fitted model. The values of the performance matrix parameter like MSE, RMSE and MAE of the NN model is observed to be 0.558, 0.747 and 0.562 respectively. This study gives insights to researchers and utility companies about traditional predictive models that can predict the HEC under anomaly situations like Covid-19. This study also helps the researchers in using Orange and JASP tool to perform the statistical and predictive modeling. © 2021 Published by peer-reviewed open access scientific journal.