Adaptive Neuro-Fuzzy Inference System-based Prediction of Heating and Cooling Loads in Residential Buildings

Accurate forecasting of energy consumption during the early design phases of buildings is crucial for optimizing energy performance, minimizing consumption, and reducing emissions. This study presents the development of an Adaptive Neuro-Fuzzy Inference System (ANFIS) model for estimating the heating and cooling energy loads of typical
Algerian multifamily residential buildings. Using dynamic simulations in EnergyPlus, calibrated with real climatic data from Biskra (2003-2017), a dataset of 1200 cases was generated based on six key building envelope variables identified via sensitivity analysis. Two separate ANFIS models were trained and validated using 80/20 data splits and Gaussian membership functions. Results demonstrate high accuracy with R² values of 0.9 for cooling and 0.88 for heating loads. The proposed ANFIS models enable fast, early-stage evaluation of design alternatives without the need for complex simulations. These findings support architects and decision-makers in creating more energy-efficient building designs under hot and dry climate conditions typical of Algeria.