Economic Emission Dispatch with Stochastic Renewable Power and Voltage Source Converters via Reinforcement Learning Multi-Objective Differential Evolution

Multi-objective economic emission dispatch (MOEED) is a key fundamental problem for the optimal operation of power systems. With the increasing scale of grid integration of renewable power sources such as wind and solar power, the VSC-based multi-port systems can improve the system operation flexibility to increase the consumption of renewable power. In this context, the structure of a multi-port system is more compact, making the corresponding MOEED more complicated. Considering the uncertainty of wind and solar power, a MOEED model for VSC-based multi-port systems is established in this study. The overestimation and underestimation situations of renewable power are modeled. To solve the MOEED model, we presented an enhanced multi-objective differential evolution, namely RLMOQILDE. Moreover, mating pool-based quadratic interpolation, reinforcement learning, and constraint processing technology are combined to boost its performance. Finally, the feasibility of the MOEED model and the effectiveness of RLMOQILDE are verified on two three-port systems constructed by expanding a modified IEEE 30-bus system.