A Search Space Reduction Method for Optimizing Sequential Control by Hypothetically Achievable Bound Estimation of the Objective Function

Metaheuristic methods such as genetic algorithm, simulated annealing, and artificial bee colony algorithm methods take much time to obtain an optimal solution, particularly when a large scale simulator is employed for estimating the state of the environment. In this paper, a search space reduction method for accelerating the optimization of sequential control systems is proposed. The proposed method estimates a hypothetical achievable bound of the objective function and uses it as the prior knowledge to reduce the search space. The hypothetical achievable bound is estimated using the fact that large scale plants consisting of multiple components are in many cases controlled in a sequential manner. The size of the search space reduction obtained by the proposed method is evaluated by an example problem that minimizes the start-up time of a thermal power plant. As a result, the size of the search space is reduced by 65%. The proposed method does not lose the optimality of the optimization method to be accelerated. In addition, this method is also applicable to optimization problems other than sequential control if the hypothetical achievable bound of the objective function is estimable without measuring the state of the environment or using the simulator.