Neural-based generation control for highly varying and uncertain loads

The design of a neural fuzzy controller for the nonconforming electric load problem in automatic generation control (AGC) is presented. This new controller utilizes the predictive capabilities of neural networks, and the uncertainty compensation by fuzzy logic to formulate an intelligent AGC system. Area control error (ACE) and its integral (ACE) are used as input variables for this fuzzy controller, and the dispatcher's operating experiences are extracted to form a fuzzy control rule base. In order to reduce unnecessary movement of generating units, a combination of triangular and trapezoidal fuzzy membership functions are used for the input variable ACE. Performance of the neural fuzzy controller in a two-area tie-line model with actual toad data from a collaborating utility is demonstrated and compared with the present AGC system through simulations. Results show that the proposed neural fuzzy controller matches the demands of highly varying loads, and largely reduces unnecessary control movements of the generating units without detriment to the ACE or the frequency deviation.