Enhancing Load Frequency Control in a Four-Area Power System Network with an Optimal ANN Controller

This research describes an optimal artificial neural network (ANN) controller for load frequency control (LFC) of a four-area interconnected power system with non-linearity. Automatic load frequency control is the main area of concern in the operation of an interconnected power system. A feed forward neural network with multi-layers and Bayesian regularization backpropagation (BRB) training function is used. This controller is designed on the basis of optimal control theory to overcome the problem of load frequency control as load changes in the power system. The system comprises of transfer function models of two thermal units, one nuclear unit and one hydro unit. To accommodate for non-linearity, the model integrates the generation rate constraint (GRC) of distinct units. The typical system parameters obtained from IEEE press power engineering series and EPRI books. The network training is based on the data collected from the optimal controller for different perturbations or step load changes. As the inputs are applied to the network, the outputs are compared to the target values, and the supervised learning rule is used. The robustness, effectiveness, and performance of the proposed optimal ANN controller for a step load change and random load change in the system is simulated through using MATLAB-Simulink. The time response characteristics are compared with that obtained from the proportional, integral and derivative (PID) controller and non-linear autoregressive-moving average (NARMA-L2) controller. The results show that the algorithm developed for proposed controller has a superiority in accuracy as compared to other two controllers. The proposed controller performs satisfactorily under random step load changes and thus desirable dynamic control of the system is achieved.