This paper proposes optimal task decomposition agents (OTDAS) of the multiagent power plant control system. The goal, decision-making and operation of the OTDAS are introduced. The OTDAS optimally decompose the task of power plant control system through an optimization agent and a decomposition agent. The optimization agent makes decision using Genetic Algorithm (GA). The results of the OTDAS show that GA is an effective tool for the agent in making decision.