As the global warming issue is gathering much more attention than ever before, the major countries in the world have set goals of achieving carbon neutrality in decades. The movement has been promoting the development of renewable energy utilization and energy-saving technologies. In ZEBs (Zero Energy Buildings), in addition to achieving a net zero annual primary energy balance, maximizing energy self-sufficiency is also important. In this case, the introduction of energy storage systems such as batteries and thermal storage is essential, and accurate solar radiation forecasting is indispensable for controlling such energy storage systems. Therefore, this study examined the impact of solar radiation forecasting methods on energy self-sufficiency and purchased electricity costs in a building equipped with PV and battery. The forecasting methods used were an open-loop type that updates predictions every hour using measured values and a hybrid model combining CNN (Convolutional Neural Network) and LSTM. The results obtained with the cluster-hybrid approach showed the highest predication accuracy. Mixed-Integer Linear Programming (MILP) algorithm uses constraints to find the minimum or maximum value of the objective function, including some variables with integer values to reduce calculation time, which is suitable for the situation in this research for optimization calculation. Renewable thermal energy is fully utilized by applying optimal operation schedules to energy storage facilities and heating and cooling appliances to adjust flexible electrical and thermal loads. The operation schedule is calculated based on prediction data and advanced optimization algorithms. As a result, the cluster-hybrid approach provided a better overall performance.
Hongzhi et al. (Tue,) studied this question.