@article {10.3844/ajassp.2024.15.27,
article_type = {journal},
title = {Multiple Linear Regression to Predict Electrical Energy Consumption Based on Meteorological Data: Application to Some Sites Supplied by the CEB in Togo},
author = {Kpomonè, Apaloo Bara Komla and Gnadi, Palanga Eyouleki Tcheyi and Yao, Bokovi and Dosseh, Kuevidjen and Komla, Nomenyo},
volume = {21},
year = {2024},
month = {Jun},
pages = {15-27},
doi = {10.3844/ajassp.2024.15.27},
url = {https://thescipub.com/abstract/ajassp.2024.15.27},
abstract = {The prediction model developed in this article is based on the use of meteorological variables to estimate the consumption of electrical energy at the substations of the Electric Community of Benin. The objective is to predict this consumption in order to adapt production to it. The posts (Lomé Aflao, Légbasito, and Lomé port) are the targets that were used in the study. The input variables are Relative Humidity (H), Direct Normal Irradiance (I), Precipitation (P), Temperature (T), and wind speed (V). The data collection period extends from 2019 to 2021. Multiple linear regression is used as the algorithm. Mean Absolute Error (MAE), root Mean Square Error (MSE), root mean square error (RMSE), and linear correlation coefficient (R2) were used to evaluate the performance of each model. A statistical characterization of each variable is carried out. It shows a good distribution of temperature, relative humidity, and wind speed values. This is not the case for direct normal irradiance, precipitation, and diffuse radiation. These latter at times have zero and extreme values at the same time. Furthermore, the modeling results show that the worst model is IPV giving MAE = 16.066; MSE = 385.847; RMSE = 19.643, and R2 =21.021%, and is not good for consumption forecasting. On the other hand, the best model is obtained by the HIPTV configuration thus giving MAE = 13.214; MSE = 282.199; RMSE = 16.798, and R2 = 77.284% showing that the parameters considered are necessary for its prediction. The correlation coefficient R2 exceeds 50%, the results of this study show that from meteorological data, it is possible to predict the power to be consumed in the area considered. However, as it is not very close to 1, the exploration of other algorithms is necessary to resume this study.},
journal = {American Journal of Applied Sciences},
publisher = {Science Publications}
}