Multilayer perceptron neural network (MLPNN) controller for automatic generation control of multiarea thermal system

This paper presents a control scheme, for automatic generation control (AGC) of both two and three unequal area thermal system, having reheat turbines and generation rate constraints, based on multilayer perceptron neural network (MLPNN) whose weights are updated using reinforcement learning. The weights of MLPNN controller are adjusted dynamically onlinely using backpropagation updation technique with area control error (ACE) as the error and input being the summation of deviation in frequency in the respective area and deviation of tie line power flow. The performance of MLPNN controller considering a wide range of system loading conditions, changes in magnitude of step load perturbation and for simultaneous step load perturbations is compared with the conventional integral controller. Investigation clearly reveals the superior performance of MLPNN controller over the conventional integral controller.