Wind speed forecasting using an echo state network with nonlinear output functions

Abstract

Wind turbines are among the most popular sources of renewable energy. The energy available from the wind varies widely because it is strongly related to wind speed and direction, which are constantly changing. Wind farms have difficulties with system scheduling and energy dispatching because the availability of wind power is not known in advance. Humidity, air temperature, average gust speed and direction, dew point, and wind speed and direction provided by the Nevada Department of Transportation roadway weather stations in the Reno, NV area are used as the inputs to the network for the case study. Nonlinear states in an echo state network are introduced to improve prediction accuracy over published methods. Simulation results of the nonlinear echo state network (N-ESN) show fewer computations, and improved prediction performance as compared to ESNs with fixed sizes and topologies.