A Novel Multi-Level Perceptron-Based Controller Design Used in Wind Turbine Output Power Stability

Abstract

Controllers are very used devices in industry. Basically, they can be divided into two main classes, classical and modern controllers. Classical controllers are controllers created in the Proportional Integral Derivative (PID) basic structure. Modern controllers are controllers that have fuzzy logic, neural network and heuristic optimization methods in their inside. In addition, controllers with a hybrid, which are a mixture of classic and modern controllers, can be considered as a third class. In this study, one of the modern control methods, neural network-based, online-trained multi level perceptron (MLP) controller is proposed. In this study, along with online training as new, the error of the controlled system was used to train the MLP network. The application of the developed controller has been used to limit the energy output of the turbine at wind speeds above the rated wind speed by controlling the blade pitch angles of a wind turbine. The simulations were conducted in the Matlab/Simulink environment. The results show that the new MLP controller can successfully control the outputs of the wind turbine.