Short term wind speed forecasting using three combination neural network based on divide and conquer

Wind power is one of the most accessible renewable energy. Wind speed forecasting with high accuracy, will be effective for the development of this power. This paper presents an appropriate solution for Wind speed forecasting problem, using three hybrid neural networks based on divide and conquer. The three networks are boosting by filtering (BF), mixture of expert (ME) and boosted mixture of experts (BME) respectively. In these networks, the problem spaces are divided between the base classifiers and then, with a determined approach are combined. Tests based on actual wind data of Mahshahr show that the BME method can predict the wind speed with higher accuracy compared to other methods. In boosted mixture of experts at first, the problem space divided by boosting structure and then obtained weight from this structure, considered as the initial weight of the mixture. For main classifier of all structure, we used multilayer perceptron neural network (MLP).Also, both error criterion and performance have been used for assessing the results.