Enhancing grid stability and efficiency in buildings through forecasting and intelligent energy management of distributed energy resources

Modern buildings face rising energy demands due to factors such as population growth, urbanization, and increased reliance on technology. Therefore, a Diagonally Masked Fusion Network (DMFN) is proposed, which involves predicting energy demand or generation based on intricate weather conditions, historical data, and known future inputs. The Multi-Head Diagonally Masked Attention mechanism enhances accuracy by considering both seasonal and historical holiday patterns, optimizing forecasts for diverse conditions. Diagonal masking in the attention mechanism ensures that attention remains focused on past information, preventing the model from relying on future data during predictions. The forecasting output provides real-time insights into future energy demand or generation patterns based on anticipated weather conditions, historical data, and known future inputs. This information allows energy managers to proactively understand how energy requirements might fluctuate in the coming hours or days. Based on the collected data and forecasts, Alpha-Guided Dwarf Mongoose (AGDM) Optimization is applied to determine the optimal operation schedule for DERs, considering grid constraints, cost minimization, and energy efficiency. The simulation results show that this approach ensures that DERs are utilized efficiently, matching energy supply and demand, and minimizing the operational cost.