Hysteretic neural network and its application in the prediction of the wind speed series

In order to improve the information processing capabilities of the traditional neural network, and improve the forecast accuracy of the wind speed series, a new hysteretic neural network based on the hysteretic neurons is proposed. The hysteretic neuron is constructed by adding a hysteretic operator into the activation function. The hysteretic characteristic can make the response of the neuron is related to not only the current input information but also the history input information. In this way, the amount of information used in the prediction process is increased, and the prediction ability of the neural network can be improved. In order to avoid the information redundancy or loss, the structure and the training samples are determined according to the autocorrelation function and the partial autocorrelation function of the wind speed series. Simulation results show that the hysteretic neural network can complete the prediction of the wind speed series, and the prediction performance is superior to that of the same type neural network.