Genetics-based control of a MIMO boiler-turbine plant

A genetic algorithm is used to develop an optimal controller for a nonlinear, multi-input/multi-output boiler-turbine plant. The algorithm is used to train a control system for the plant over a wide operating range in an effort to obtain better performance. The results of the genetic algorithm's controller are compared to the performance of an optimal controller designed from the linearized plant model at a nominal operating point. Because the genetic algorithm is well-suited to solving traditionally difficult optimization problems it is found that the algorithm is capable of developing the controller based on input/output information only. This controller achieves a performance comparable to the standard linear quadratic regulator.<>