Efficiency Assessment of ANN, ANFIS, and PSO-ANFIS for Predicting University Residence Energy Usage

To shape future energy strategies effectively, it is crucial to comprehend the dynamics of energy generation and utilization. Given the significance of accurate prediction, this investigation undertakes a comparative analysis of the predictive capabilities of artificial neural network (ANN), standalone adaptive neuro-fuzzy inference system (ANFIS), and its hybrid counterpart integrated with particle swarm optimization (PSO). The focus lies on forecasting energy consumption in student residences based on climatic variables, with the University of Johannesburg's student housing serving as a specific case study. The input variables encompass ambient wind speed, wind direction, temperature, relative humidity, and atmospheric pressure, while the target variable is the corresponding energy consumption for student accommodation. Performance evaluation metrics such as root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE), were employed to assess the efficacy of the developed models. Results obtained showed that the hybrid PSO-ANFIS outscored standalone ANN and ANFIS models with the lowest values of the RMSE, MAPE, and MAE, respectively. The developed model can aid in optimizing energy usage and support the design and dimensioning of alternative energy systems for campus housing.