Research On Spatial Event Locating Methods And Simulation Based On Sensor Monitoring Networks

With the development of sensor technology,the application of low cost and multi-type sensors is more and more common.Compared with traditional monitoring network with a small number of high cost sensors,the modern monitoring network has more sensors,whose costs are lower and data quality is decreased.How to extract more useful information from this network data has become the primary problem of monitoring.Based on sensor monitoring network,this paper carries out the spatial event localization methods and simulation research,the contents are as follows.(1)A spatial event detection method based on multi-level PCA is proposed.Using centralized method to make a unified decision by the measurement data.After detecting whether a fault occurs,estimate the abnormal size,and provide the situation awareness according to statistic changes.(2)An event locating method based on sequential unscented Kalman filter is proposed.With the event concentration diffusion model,the dynamic sequence equation of system is constructed,and the static single point source is located by sequential unscented Kalman filter.It can reduce the conservatism of the linear least squares method in robustness effectively.(3)A nonlinear PLS modeling method and a new process monitoring algorithm based on ELM are proposed.The former aims to use ELM to establish PLS inner model,and to solve nonlinear fitting problem.The latter uses ELM quality prediction value to induce decomposition of PLS principal component subspace,aiming to improve the fault detection capability.(4)A method of combination of ELM and Kalman filter method to locate spatial event is proposed,aim to solve spatial event locating problem without concentration diffusion model.Use ELM to fit the system measurement equation,and combine sequential unscented Kalman filter to locate spatial event.Simulation verifies the effectiveness of this method.