Optimal reactive power dispatch based on a multitask-assisted constrained multimodal multi-objective evolutionary algorithm

In modern power systems, multi-objective optimal reactive power dispatch is crucial for reducing power loss while maintaining bus voltage stability. However, gaps in multimodal feature mining and multi-objective decision-making demand further progress. This study proposes a constrained multimodal multi-objective evolutionary algorithm with a multitask-assisted strategy, assigning global exploration and local exploitation to separate evolutionary tasks and adopting information transfer for inter-task coordination. A minimum Manhattan distance based decision-making scheme is developed to select trade-off solutions and provide valuable alternatives. Experimental validation on 17 benchmarks confirms the algorithm's superior multimodal optimization performance. Case studies on IEEE 30/57/118-bus systems verify its effectiveness in optimal reactive power dispatch. For the case with wind power and load uncertainties, it achieves an expected power loss of 2.1916 MW and a voltage deviation of 0.3133 p.u. Experimental results show the algorithm can efficiently optimize power loss and aggregate voltage deviation while satisfying all constraints.