Research On Clustered Data Aggregation With Time Series Prediction Model Based On NARX Neural Network

Wireless sensor networks(WSNs)are composed of densely distributed sensor nodes which perceive environmental information periodically and upload the collected data to achieve the purpose of environmental monitoring.The data collected by the adjacent nodes has spatial correlation and the data collected by the same node has time dependence in the continuous period in WSNs.These two kinds of correlation make the sensed data of nodes with great redundancy.Considering the characteristics of sensor nodes which are limited energy without being added,limited storage space,and limited network communication bandwidth,the redundant data sensed need be aggregated to be transferred.The advantage of this method is to effectively reduce transmission capacity of the node information,save node energy and prolong the network life cycle,and increase the utilization rate of bandwidth.Thesis presents a clustering algorithm based on neural of clustered network data aggregation(N-CDAA),in order to solve the problem that the accuracy of the existing data aggregation algorithms based on prediction data is poor.The algorithm takes the spatial and temporal correlation of node collecting data into full consideration in WSNs.By introducing two clustering norms which are index the residual energy of nodes and clusters within the average minimum reachable power,thesis use vector quantization(Vector Quantization VQ)algorithm to select the representative node as the cluster head,in order to balance the node energy consumption.At the same time,the NARX neural network is used to improve the prediction accuracy of nodes in the cluster,and further reducing the amount of data communication.The simulation results show that the N-CDAA algorithm has a higher degree of data fusion,can effectively reduce the amount of data transmitted by the node,save the energy of the nodes in the network and prolong the network life cycle.In order to solve the problem that how to judge the abnormal data in the process of data aggregation,thesis uses an anomaly detection algorithm based on singular value decomposition of Confidence Interval(SVD-CI).In hierarchical WSNs,the Singular value decomposition is used to descend dimension to eliminate redundant data for the node data with spatial correlation.According to the characteristics of time correlation,the confidence interval is constructed by using time series to detect abnormal data.The simulation results show that the SVD-CI algorithm has higher detection rate and lower false alarm rate when detecting anomaly data,and it saves the energy of cluster head node to a certain extent.