Enhancing PV Solar Output Prediction and Optimization using Artificial Neural Networks in a Remote Village in the Noun Division, West Region of Cameroon

Electricity generation and accessibility in rural areas are major concerns globally. Photovoltaic (PV) systems are among the most accessible, economical, and safe sources of electricity. However, climate change and global warming contribute to significant changes that affect the performance and reliability of PV systems. This study aims to develop an artificial neural network (ANN) model to predict the solar PV output at a mini off-grid solar park located in a rural village in the Noun Division, West Region of Cameroon. A Multilayer Feedforward Neural Network (MLFNN) architecture is used. The model achieved outstanding predictive performance across all the datasets. In the training phase, the MLFNN attained a Mean Squared Error (MSE) of 3.2630 × 10−5, Root Mean Square Error (RMSE) of 0.0057, and Mean Absolute Error (MAE) of 8.9440 × 10−4. Corresponding values for the validation phase were even lower, with an MSE 2.7374 × 10−5, RMSE of 0.0052, and MAE of 8.7649 × 10−4. During testing, the model maintained high performance with an MSE of 8.594 × 10−6, RMSE of 0.0029, and MAE of 6.6972 × 10−4. Furthermore, the model exhibited excellent coefficient of determination R2 values of 0.9995, 0.9996, and 0.9998 for the training, validation, and test sets, respectively. The correlation coefficients (R) were similarly high, reaching 0.99973 (training), 0.99978 (validation), and 0.99992 (test), confirming a strong agreement between the predicted and actual outputs across all phases. This study provides a highly accurate model for predicting solar energy generation, aiding in the efficient design, installation, and management of off-grid PV systems in rural areas.