Abstract
One of the key technologies to take full advantage of wind power is to establish a wind turbine (WT) generator output estimation system with high accuracy. The static feed forward artificial neural network is widely used in previous WT generator output estimation technology. However, this method has many problems such as local minimization, a lack of dynamics, edge effect, and multi-correlation. To mitigate these shortcomings, in this paper, the principal component analysis (PCA) and the Elman artificial networks based on main component analysis are utilized to predict the WT-generator output. Firstly, the original multi-dimension variables of weather forecast values are processed, and the principal multivariate variables were pretreated by principal component analysis to obtain the simplified network structure with low dimensional variables. Then multi-hidden-layer and dynamic neuron estimation method are used to optimize the network path and new training strategy to get the final estimation result. The case study of Zhongbao windfarm short-term wind power prediction shows that the proposed method can effectively raise the estimation accuracy of windfarm output compared with traditional artificial neural networks, and is proved as an effective way to forecast short-term wind power.

This publication has 7 references indexed in Scilit: