Research on Configuration Optimization of Distributed Energy Systems under Demand Uncertainty

As energy and environmental challenges intensify, distributed energy systems (DES) are gaining attention for their high energy efficiency. In this study, Latin Hypercube sampling (LHS) technique was used to generate load uncertainty data, and two-stage stochastic programming (TSSP) model was used to optimize device capacity configuration. The research shows that the battery capacity and air conditioning capacity of the system increase after load uncertainty is taken into account, and the cooling load demand satisfaction rate increases significantly. Although the LHS approach resulted in a ${1 4 . 1 6 \%}$ and ${5 . 2 5 \%}$ increase in initial investment and operating costs, respectively, the number of hours that were not met with cooling load requirements was significantly reduced by ${9 2 . 0 9 \%}$, resulting in improved system stability and economy. Although the annual cost increased by $6.58 \%$, the LHS approach effectively coped with load fluctuations. Future research could explore deep learning and reinforcement learning techniques to improve the efficiency and accuracy of scene generation.