Year: 2023 | Month: December | Volume: 10 | Issue: 12 | Pages: 42-56
DOI: https://doi.org/10.52403/ijrr.20231207
Multiplayer Perceptron and Simple Regression Linear Approaches to Predict Photovoltaic Active Power Plant: Case Study
APALOO BARA Komla Kpomonè1,2, APEKE Kodjo Séna3, PALANGA Eyouleki Tcheyi Gnadi1,2, BEDJA Koffi-Sa3
1Département de Génie Électrique, Ecole Polytechnique de Lomé (EPL), Université de Lomé, Togo
2Laboratoire de Recherche en Sciences de l’Ingénieur (LARSI), Université de Lomé, Togo
3Département de Génie Électrique, Institut de Formation Technique Supérieur, Togo
Corresponding Author: APALOO BARA Komla Kpomonè
ABSTRACT
This article deals with the forecast of electricity production from a 50 MW photovoltaic power plant in Blitta, a town in Togo. The objective is to use meteorological variables such as instantaneous irradiation (A), wind speed (B), ambient temperature (C) and module temperature (D) to predict the active power. Multilayer Perceptron architecture, Artificial Neural Networks and multiple linear regression are explored as methods in Python. A classification of variables is presented. Certain model performance evaluation criteria made it possible to observe the results of the models. 26989 data samples are used. The results give a strong correlation between the ambient temperature of the location and the temperature of the module, i.e. 87% and 40% between the wind speed and the instantaneous irradiation. Also, as results we have: MAE = 6.017; MSE = 67.392; RMSE = 8.209; RRMSE = 15.185% and R2 = 55.321 by multilayer perceptron and 60 neurons under the hidden layer then MAE = 6.93; MSE = 77.37; RMSE = 8.80; RRMSE = 51.42%, R2 = 69.97% obtained by linear regression. This shows that there is a strong correlation between the variables used but the high values of RRMSE will drive the need to use other algorithms.
Keywords: Artificial Neural Networks, Correlation, Photovoltaic Active Power Plant, Simple Regression Linear, Multi-layer Perceptron,
[PDF Full Text]