A study on optimal energy management with power generation and consumption forecasting for ZEB provided by GWHP and energy storage

The rapid proliferation of renewable energy technologies, particularly photovoltaic (PV) systems, combined with the declining costs of energy storage, has significantly increased interest in advanced energy management strategies that prioritize the on-site use of excess electricity. Demand response (DR) programs and Model Predictive Control (MPC) have emerged as practical approaches to balance electricity demand while minimizing objectives such as economic costs and greenhouse gas emissions. A critical requirement for these strategies is the ability to forecast both energy generation and consumption accurately. This study focused on a medium-sized office building located in a cold climate, with the objective of reducing electricity expenses. Through optimization calculations, the most efficient operational schedules for various energy systems— including PV systems, electrical batteries, thermal storage tanks, HVAC units, and heat pump water heaters— were identified. PV generation was forecasted utilizing Long Short-Term Memory (LSTM) networks. Concurrently, forecasts of electricity demand and HVAC thermal loads were predicted with Transformer models enhanced by incremental learning to augment accuracy and adaptability. These forecasts were integrated with device performance characteristics, derived through piecewise linear regression, and operational constraints to formulate optimal schedules for electrical batteries, HVAC units, and heat pump water heaters using Mixed-Integer Linear Programming (MILP). A comparative analysis was conducted against a traditional Rule-Based Control (RBC) strategy, which demonstrated that the proposed optimization-based approach effectively shifted load peaks in response to excess generation and fluctuating electricity prices. This facilitated efficient energy charging and discharging, resulting in a reduction of electricity procurement costs and overall expenditure compared to the RBC methodology. The most significant benefits were observed during the summer season when PV generation was abundant.