Research On Spatial-temporal Correlation Of Wireless Sensor Data

WSN (Wireless Sensor Network) is a monitoring network composed of many kinds of small and intelligent sensors. Nowadays, it is widely used in medical, military, scientific research and other fields. Although WSN has brought us convenience, we also have to pay attention to the limitations of sensors at the same time. First, sensors, powered by batteries, are usually distributed in the special environment. Thus, the power is limited and it's difficult to replace the battery. It's hot topic to study how to use its power and save its energy as much as possible at the same time. Secondly, there exists a certain degree of correlation between the data acquired by WSN and the deployment of nodes which leads sensors to collect redundant data. It will waste processing resources and increase the energy consumption as well. As a result the study in redundant data plays a significant role in WSN.Focusing on the problems appears in WSN, this paper uses the characteristics of spatial-temporal correlation and studies the spatial-temporal correlation based on data prediction. Improving the existing algorithm, the present dissertation contains the following major findings:Firstly, this paper analyzes the temporal correlation of the data collected by sensors. We reduce the data transmitted to the sink nodes and save the energy consumption by modeling the time series and predicting the collected data. In this paper, traditional temporal correlation algorithm is analyzed, as well as the characteristics of grey model and AR model (Autoregressive Model) in time series. In order to exploit the advantages of both grey model and AR model, this paper proposes a modeling method named G-AR (Grey and AR model) which combines grey model and AR model. The predicted data is defined as a combination of the two methods which will efficiently avoid the abnormal and unstable situation caused by single method. Simulation results show that G-AR is better than gray model, AR model and traditional model in data accuracy and data transmission rate. Thus, G-AR is used in data prediction to reduce data transmission and ensure data accuracy at the same time.Secondly, in view of the spatial correlation, this dissertation analyses the disadvantages of EAST (Efficient Data Collection Aware of Spatio-Temporal Correlation) algorithm proposed in recent years and proposes the improved method named IM-EAST (Improved EAST). Residual energy and communication distance are considered comprehensively when choosing a head during clustering. We consider the optimal hop number and the optimal transmission distance when choosing a relay node during relaying data to the sink. So we can reduce the energy consumption as much as possible. Our method is able to solve the deficiencies of strategy in choosing heads and relay nodes in EAST algorithm and give more fully thought of various kinds of factors. The experimental result shows that IM-EAST consumes less energy than EAST. Its lifetime of network is longer than EAST.Finally, we give a summary of this paper and analyze both the advantages and disadvantages. At the same time, the direction for further study is given in the end.