Many-objective virtual power plants resource scheduling based on evolutionary multifactorial optimization

A virtual power plant (VPP) is a novel power system that aggregates diverse resources, including distributed generation, user load, and energy storage. However, with multiple uncertainties affecting the market price, carbon emissions from thermal power units, and output from energy storage, the improvement of VPP scheduling optimization methods becomes an urgent problem. In this paper, a multiple uncertainties with VPP for power resource scheduling problem is proposed. It is applied to VPP integrating distributed energy, thermal power units, energy storage, and participating in electricity trading. Then, employing random, fuzzy, and interval representation methods, a many-objective optimization model for VPP was constructed. This model accounts for system operating costs, power plant carbon emissions, the balance of control resource types, and the decay rate of energy storage efficiency. Based on the functional characteristics, the model was decomposed and reformulated into an uncertainty-based many-objective multi-tasking optimization model. Particularly, we designed a meta-learning many-objective multi-tasking evolutionary algorithm based on an inner-outer layer co-evolution mechanism to achieve automated feedback correction of initial algorithm parameters. Finally, the experimental results on actual data from a specific region show that the model is able to reduce the impact of uncertainty on the system objectives effectively, and the proposed algorithm also shows a significant performance improvement.