Complex-valued estimation of wind profile and wind power

This paper addresses the problem of wind profile and wind turbine power estimation. A complex-valued pipelined recurrent neural network (CPRNN) architecture is proposed. The network is trained by the complex-valued real-time recurrent learning (CRTRL) algorithm with a general ´fully´ complex activation function which makes it suitable for forecasting wind signal in its complex form (speed and direction). The subsequent complex-value based prediction of wind turbine power is shown to significantly differ from the one based on independent prediction of wind speed and wind direction with the latter mainly being more optimistic in predicting the turbine power output.