Regression function and back-propagation through time training method for wind turbine neural network pitch-controller

Reliable wind turbines operation for production of electrical energy requires a modern control system. In order to obtain a reliable operation, this paper present an effective control system using artificial Recurrent Neural Network (RNN) trained by Sequential Response Surface (statistical method) learning method in a pitch-control, variable speed wind turbines. The main objective is to ensure stability and optimal operation of a variable speed, fixed pitch turbine at all operating points. The turbine was modeled in order to test the designed controller by simulation. Method consideration were based on above rated and below rated operations so as to extract maximum energy. This is achieved by keeping the rotor power coefficient at the maximum level all the time regardless of the wind speed.