Abstract
Renewable energy is currently driving innovation in the energy industry. However, the widespread distribution of renewable energy resources across various regions and the temporal variability in production pose challenges for the operation and management of virtual power plant (VPP). To address this issue, a Dynamic Virtual Power Plant (DVPP) has been proposed for the dynamic configuration of VPP. In this paper, we propose the MaxProfit Time Series Clustering for Virtual Power Plant (MTPSC-VPP) to effectively aggregate virtual power plant. We perform time series clustering based on density using the normalized Mean Absolute Error (nMAE) between actual measurements and predictions. Subsequently, we compare the proposed algorithm with non-time series and existing algorithms through experimental evaluations. The results indicate that MTPSC-VPP yields approximately 93%p higher prediction revenue compared to non-time series methods and about 15%p higher compared to existing algorithms. However, it exceeds the calculation criteria of up to 30% of reward on specific dates, suggesting directions for achieving better performance in future research.