Recursive wind speed forecasting based on Hammerstein Auto-Regressive model

Wind speed forecasting approach using secondary decomposition algorithm and Elman neural networks

“…In the predictor, the simultaneous perturbation stochastic approximation algorithm was performed to train the built MFNN. Ait Maatallah et al [20] put forward an artificial intelligent wind speed forecasting model by adapting the HM and the AR approach, which suited for a short-term horizon. The results proved that this intelligent model outperformed both of the ARIMA model and the ANN model.…”

Swarm and evolutionary computing algorithms for system identification and filter design: A comprehensive review

“…A Hammerstein model, which is essentially a static nonlinear block followed by a dynamic linear section has been a popular nonlinear model employed for modelling various nonlinear plants (Narendra and Gallman, 1966;Eskinat et al, 1991;Zi-Qiang, 1994) and processes including distillation columns (Eskinat et al, 1991), heat exchangers (Eskinat et al, 1991), brushless motors (Jing et al, 2013), loading process in diesel engines (Ayoubi, 1998), spark ignition engine torque (Togun et al, 2012), commercial quadrotor helicopters (Souza et al, 2012), wind speed forecasting (Maatallah et al, 2015), giant magnetostrictive actuators (GMAs) (Guo et al, 2015), superheated steam pressure in boiler (Zhao et al, 2014), DC/DC boost converter (Alonge et al, 2015), thermal process , loudspeaker precompensation (Defraene et al, 2014), speed regulation in induction motor and inverter (Congli et al, 2015) and ultrasonic motor (Jingzhuo et al, 2014). The Hammerstein model is popular in control problems because of its ability to effectively model actuators which contribute to a dominant nonlinearity and other nonlinearities 20 are non significant.…”

A Two-step Forecasting Solution and Upscaling Technique for Small Size Wind Farms Located in Hilly Areas of Romania

“…For instance, in the case of short-term prediction, stochastic methods (persistence, autoregressive models [12] and generalized equivalent Markov model [13]) are recommended. Furthermore, other research [14] has used the Kalman filter integrated with support vector regression (SVR) to obtain a 10% prediction improvement, comparing the obtained data with Artificial Neural Network (ANN) and autoregressive (AR).…”