Smart Building Energy Optimization Using AI-Based Forecasting and Management Techniques

Effective functioning of smart structures depends on good management of energy assets in order to reach reduced running costs together with best use of available energy. Building sector energy usage represents 40% of the global total and scientists expect this number to exceed 50%. Proper energy prediction systems help organizations to distribute resources more effectively and prevent unnecessary energy usage. This research develops a deep learning model through the grouping of CNN and RNN which achieves hourly energy consumption forecasts. The proposal utilizes CNN to yield spatial energy characteristics but LSTM and GRU serve as RNN variants to detect temporal dependencies in the energy information. By surpassing traditional models including LR and ARIMA, the suggested CNN-GRU designs presented outstanding results. The CNN-GRU model received validation by testing on CUBEMS dataset that contains actual energy measurements from a Bangkok Thai academic building. The combination of previous energy usage records with time-related data attributes produces improved prediction outcomes according to experimental findings. Among the multiple forecasting methods that the study evaluated MAE, RMSE provided the evaluation metrics. Short-term energy forecasting in smart buildings found its most effective solution through the CNN-GRU model which delivered a 98% accuracy rate.