UC San Diego

Optimization tools require prohibitive computational time to model energy systems with annual hourly input data, and input is typically reduced into representative periods to increase solver speed. Data reduction in microgrid optimizations impacts the objective function accuracy. Methods preserving demand data fluctuations through reduction into representative days show improved accuracy with increases in number of representative periods. This work presents a method of data reduction that aggregates annual hourly demand data into typical weekdays and weekends, while explicitly preserving demand peaks in distinct representative profiles. The proposed method is tested in an energy system optimization using historical 15-minute resolution annual demand data from a gymnasium in La Jolla, California, and the system is optimized in terms of total annual costs. Results are in good agreement with a full-resolution optimization of the energy system, establishing the validity of the proposed technique. Additionally, a comparison of method performance demonstrates a significant improvement in accuracy with the inclusion of peak demand profiles.