Model-based predictive control of tunnel ventilation

Two predictive methods on ventilation control in tunnel are developed to offer a pleasant running environment for the drivers in the tunnel and reduce power consumption. The theoretical basis is described and its use is illustrated for Shanghai Xiang Yin Under-river Tunnel. One method is based on Time-series prediction and another is on ARX modeling. Both are based on online modeling identified by 24 hours data of CO concentration and visibility index before current time. The data preprocessing including filtering and data differencing is discussed. The sum of squares (SS) for the residuals, a function minimization algorithm, is used to give the parameter of the model. For time series analysis stationarity requirement has to be fulfilled before model fitting. Differencing data is adopted to transform time series to stationarity Cross Correlation Function calculation between the residuals and input of obtained model shows their validation. For ARX modeling a Determinant Ratio method is used for model's order estimation, A set of candidate model structures is used for comparison of their properties.