The escalating impact of global warming underscores the urgent need to optimize energy efficiency, particularly in the building sector. This study introduces a reliable predictive model for air conditioner consumption, focusing on an overheated laboratory room undergoing diverse tests. Utilizing Matlab-Simscape, our model incorporates room and envelope characteristics, external temperatures, and internal heat sources. The control system regulates the air conditioner to maintain desired temperatures. The thermal model, validated through extensive data-driven and experimental comparisons, exhibits commendable accuracy. Neural network predictions for the temperature of each wall contribute to the model, enhancing its capability to forecast air conditioner consumption. By providing valuable insights into energy usage patterns in overheated environments, this study informs the sizing of photovoltaic systems. This ensures they are appropriately designed to meet energy demands while promoting sustainability.