WSN Data Fusion Research Based On Prediction Model

With the development of science and technology, wireless sensor network(WSN) as a new technology, has the widespread application in the material tracking, intelligent agriculture, health monitoring, supply chain management and environmental perception. A large number of nodes form a self organizing network, which integrates sensor technology, wireless communication technology, embedded technology, is an interdisciplinary frontier research and application field. Data fusion is a key technology to extend the life cycle of wireless sensor networks. When sink node collecting data, use data fusion algorithm to reduce the amount of data, reduce the problem and delay energy etc.This thesis first introduces the WSN technology components, for different types of WSN data fusion algorithms are outlined. Then, the data fusion algorithm based on prediction model is analyzed in detail, and describes the process of data fusion algorithm. Secondly, in-depth study of the gene expression programming(GEP) method for the prediction of the application of WSN, the algorithm of GEP was introduced in detail. Thirdly, in-depth study of the application of chaotic prediction algorithm in WSN, introduced the chaos identification method, the fusion algorithm of chaotic prediction algorithms and data process is discussed in detail. Finally, study nonlinear autoregressive(NARX) neural network application in time series prediction, introduced the NARX network data fusion step.Results of data simulation show that, GEP algorithm does not rely on a priori knowledge, not only can correctly predict the trend of data, also can guarantee the prediction accuracy, the prediction accuracy is higher than that of ARMA algorithm. Chaotic local forecasting method can predict nonlinear deterministic data, reduce the data communication, cost lower energy consumption, enhances the whole wireless network life cycle. NARX neural network can predict all the characteristics of the data, the calculation is simple, without the need for phase space reconstruction, saving storage space, with a wide range of applications.