Study Of Tunnel Sensor Data Prediction Based On Time Series Analysis

In recent years, with the establishment of the road system coming into ahigh-speed development period，more and more highway tunnels are put into operationin mountainous regions. Tunnel monitoring system collects numerous tunnelenvironment data, while these data are idle as not being paid enough attention to. Inorder to predict the future trend of the pollutant within tunnels, this paper attempts tostudy the tunnel environmental data from two aspects: chaotic time series predictionand time series similarity search.Firstly, this paper introduces the basic concepts, classification and predictionmodel of time series forecasting methods briefly, then focuses on the concepts ofchaotic theory and chaotic identification method. The chaotic theory is also applied tothe analysis of the tunnel environmental data.Secondly, according to the chaos characteristic of tunnel environmental parametertime series, a chaotic time series prediction algorithm based on Empirical ModeDecomposition and Volterra model is proposed, which predicts smoke concentrationdata sets and public data sets, and is compared with the traditional time seriesforecasting algorithm. The experiment shows that the new method can effectivelypredict the future trend of the pollutant within tunnels.Finally, in allusion to the flaws of the present piecewise linear representationmethods, a piecewise linear representation method based on feature points ismeliorated. This method divides the data points into extreme points and ordinary points,which can partition the time series of tunnel smoke concentration stably and effectively.On this basis, a similarity search technology based on morphological features isimproved, finding the similar data segments from the existing database to predict thefuture trend of the environmental parameters within tunnels. Thus it can predict thechanging tendency of time series for excessive amount of tunnel smoke concentration.