Study On Inversion Model For Accidental Water Pollution Based On Gst-mq Collocation Method

In recent years, water pollution accidents occur more frequently in China and urban water-supply security is seriously threatened. It is therefore necessary to establish water-pollution accident emergency warning system. In this paper, the identification of pollution source and simulation of pollutant distribution lead to an inverse problem of boundary-condition control. Furthermore, the global space-time multiquadric (GST-MQ) radial basis function is built from the conventional radial basis collocation method (RBCM) to solve the inversion model for accidental water pollution. The conclusions are as follows.The least square method is introduced into the conventional RBCM to construct the least-square-based radial basis collocation method (LS-RBCM), which can effectively solve the boundary-condition identification problem for 2-dimensional steady diffusion equation using the boundary (internal) observations. The accuracy of the concentration (flux) solution from the LS-RBCM method is improved by 55% (90%) for the problem using the boundary observations at the error level e > 0.005, and the accuracy of the concentration solution from the LS-RBCM method is improved by 50% for the problem using the internal observations at the error level e > 0.05.The GST-MQ radial basis function is designed by incorporating time dimension into the inverse MQ function to solve the boundary-condition identification problem for 1-dimensional unsteady diffusion equation using the boundary (internal) observations. Comparing with the MFS method, the accuracy of the concentration (flux) solution from GST-MQ method is improved by 80.7% (82.5%) for the problem using the boundary observations at the error level e = 0.001, and the accuracy of the concentration (flux) solution from GST-MQ method is improved by 51.2% (39.0%) for the problem using the internal observations at the error level e = 0.1.The inversion model based on GST-MQ collocation method is able to solve the pollution source identification problem for 1-dimensional unsteady advection- dispersion equation using the concentration data at observation site. The ES error plot poses a V-shaped distribution as the values of shape parameter c0 and scaling factor w grow; as for other factors including time stepΔt,collocation point density N,observation number M 1 and observation sited , the optimal choices of their values can result in good quality solutions.The inversion model based on GST-MQ collocation method has been successfully applied to 1-dimensional surface water, 2-dimensional surface water and groundwater pollution accident. The results show that, the GST-MQ method can reasonably estimate the constant, intermittent and transient source release history and the concentration distribution in 1-dimensional surface water pollution accident. The error EC of estimated concentration distribution C ( x )at time t = 30,60,90,120 is (0.0536, 0.0133, 0.0177, 0.0185),(0.0550, 0.0437, 0.0138, 0.0278) and (0.0978, 0.0927, 0.0916, 0.1109), respectively. For the space-and-time dependent pollutant source identification problem in 2-dimensional surface water pollution accident, the E Serror of estimated source release s ( y , t )is 0.0617. Moreover, decreasing the collocation density N and adoption of domain decomposition method (DDM) can both save the storage requirement and ensure the accuracy of the solution. For the same problem in 2-dimensional groundwater pollution accident, the EC error of the estimated concentration distribution C ( x , y ) at time t = 0.05,0.10,0.15,0.20 is 0.2287, 0.1092, 0.0837 and 0.0801. The error in the estimated source release can not be magnified in the error in the estimated concentration distribution.