Climatology
Ali Zarei; Asadollah Khoorani
Abstract
The study aimed to introduce the most effective model for estimating energy consumption through the modeling of residential electrical energy consumption. .Reanalysis climate data from ECMWF spanning the years 2003 to 2022 were acquired, along with the annual electrical energy consumption data of the ...
Read More
The study aimed to introduce the most effective model for estimating energy consumption through the modeling of residential electrical energy consumption. .Reanalysis climate data from ECMWF spanning the years 2003 to 2022 were acquired, along with the annual electrical energy consumption data of the residential sector across Iranian provinces. Pearson correlation coefficient was employed to analyze the relationships between variables, and the Mann-Kendall non-parametric test was utilized to scrutinize trends in these variables. Four regression and Artificial Intelligence (AI)-based models, namely Multiple Linear Regression (MLR), Artificial Neural Networks (ANN), Support Vector Machine (SVM), and Random Forest (RF), were employed to model electric energy consumption. The performance of these models was assessed using metrics such as Root Mean Squared Error (RMSE), Coefficient of Determination (R2), and Standard Deviation (SD), evaluated through a Taylor diagram. Provinces. In 22 provinces, a positive correlation was observed, whereas in 9 provinces, a negative correlation was identified. Analysis of the temporal changes indicates a consistent increase in minimum and maximum temperatures as well as electrical energy consumption across all provinces. However, it is noteworthy that the UTCI displayed a negative trend in several provinces. The stepwise regression model revealed that in 23 provinces, the sole influential variable is the minimum temperature. Notably, in the provinces of Isfahan, South Khorasan, and Kerman, both minimum temperature and the UTCI were identified as influential variables. Conversely, in Ardabil, Gilan, and Golestan provinces, only the maximum temperature featured in the regression equation. Modeling outcomes underscored the superior performance of the ANN model in comparison to the other three models. The ANN model exhibited the highest correlation coefficient at 0.79, coupled with the RMSE of 360. Following in ranking, the MLR, SVM, and RF models demonstrated progressively lower levels of performance
