Combined Real and Reactive Power Economic Dispatch using Multi-Objective Reinforced Learning with Optimized Losses

Abstract

Most of the economic dispatch (ED) works so far deal with real power dispatch only. With the integration of renewable energy into the grid, reactive power dispatch cannot be ignored any longer due to its importance in providing security and reliability in power system planning, operation and control. This paper deals with the formulation of combined real and reactive economic dispatch (CRRED) subject to equality, inequality and stochastic constraints. An effective algorithm that uses a hybrid of distributed slack bus (DSB) formulated using combined participation factors (PF) and multi objective reinforcement learning (MORL) is proposed in this paper. The IEEE 14 Bus was used to validate the effectiveness of the proposed CRRED formulation and Hybrid method .The numerical results obtained show that combining real and reactive power results in a 0.95% decrease in the overall generation cost as compared to a case in which only real power is considered. Further, when the losses are distributed in the entire network using the DSB, then the overall generation cost is reduced by 29.6% due to the reduced losses in DSB model.