Nonlinear predictive controller design for load frequency control in power system using quasi Newton optimization approach

Power plants are highly nonlinear systems demand a powerful identification method for prediction of their future values or for control applications. In this paper, a generalized predictive controller (GPC) is developed by neural network for application of power plants load-frequency. In this case, the identified model is characterized by nonlinear model structure based on neural network. The control objectives are to maintain the frequency within a desired range in the presence of load disturbance and governor parameters uncertainty. Based on the nonlinear model predictive control (NMPC), a controller is designed, with particular emphasis on an efficient quasi-Newton algorithm. Quasi newton optimization method is considered for update the inverse Hessian matrix for minimization of NMPC criteria. The algorithm is compared with common PID controller for reference tracking and disturbances rejection.