Wind energy potential prediction of Bamenda, Cameroon based on artificial neural network models

This study evaluates the performance of Artificial Neural Network (ANN) models for long-term monthly wind speed forecasting in Bamenda, Cameroon, a mid-altitude tropical region characterized by complex topography and challenging climatic conditions. Reliable long-term wind forecasting is essential for effective wind energy planning and sustainable renewable energy development, particularly in under-studied tropical regions. The research compares two ANN architectures, the Multilayer Perceptron (MLP) and the Nonlinear Autoregressive model with Exogenous Inputs (NARX), to determine which provides more accurate monthly wind speed predictions. Using 35 years of meteorological data (1979–2014), the study incorporates five input variables: relative humidity, precipitation, solar radiation, wind direction, and minimum and maximum temperatures. A rigorous sensitivity analysis was conducted to examine the effects of the activation functions, hidden layer neurons, input delays, and different training, validation, and testing data proportions on the model accuracy. The Levenberg–Marquardt training algorithm was applied, while model performance was assessed using statistical indicators such as the coefficient of determination (R²), Mean Squared Error, Root Mean Squared Error, and Mean Absolute Percentage Error. Contrary to expectations favouring more complex temporal models, the simpler MLP architecture achieved superior predictive performance, with an overall R² value of 0.85255 compared to 0.71689 for the NARX model. The findings demonstrate that MLP models are computationally efficient and highly reliable for long-term wind energy assessment in Bamenda, thereby supporting renewable energy planning, environmental management, urban development, and Cameroon’s efforts toward achieving the UN Sustainable Development Goals by 2030.