Multi-objective optimal tuning of power plant controls using genetic algorithms

Often, power plant controllers have to deal with multiple and demanding performance objectives. This task becomes even more stringent if the objectives are to be met for multiple operating points. Many new control techniques have been shown to deal with difficult control situations with relative ease. However, their design process is usually not straightforward, in addition to the extra requirements they may impose during the implementation phase. This article shows that simple or classical controllers can also resolve difficult performance objectives provided they are carefully tuned. A genetic algorithm (GA) is used here to tune two different controllers. The first is a Proportional and Derivative (PD) controller and the second is an optimal Linear Quadratic Regulator (LQR). Both controllers are tested with a 160 MW steam power plant. Despite the conflicting design objectives, the GA-tuned controllers were capable of offering reasonable performance while alleviating the efforts of the tuning process. Quantitative simulation results in terms of output performance is given for 15 different operating points