Identification of Source Information for Sudden Water Pollution Incidents in Rivers Based on Variable Fidelity Surrogate-dream Optimization

For sudden water pollution incidents in river, the ability to quickly identify the pollution source is of great importance for early accident warning and emergency control.Based on bayesian reasoning, a method for variable fidelity surrogate -- DREAM optimization coupling traceability is established in the posterior space of pollution sources. By this method, the variable fidelity surrogate model is composed of the high and low fidelity ones fused by the addition bridge function, with the high and low fidelity model established by using the gaussian process generated according to the Latin hypercube sampling and the high and low precision hydrodynamic -- water quality numerical calculation. To reduce the uncertainty of the surrogate model and improve the efficiency of source identification, this paper integrates the variable fidelity surrogate --DREAM optimization process; based on the minimize surrogate prediction criterion, new samples are inserted to improve the precision of surrogate model in the posterior space on the pollution sources , forming a more efficient computing model to determine source location and releasing time and released mass for sudden water pollutants in river. The proposed approach is tested by numerical and engineering examples. The result show that the new approach can effectively improve the calculation efficiency and satisfy the identification of source information for sudden water pollution incidents in rivers, which is used for the identification source information and emergency response in river .