Abstract
In previous studies, a 0-1 mixed integer nonlinear programming optimization model (0-1MINLPOM) could only identify the location and release intensity for groundwater contamination sources (GCSs), and the location of each GCS was regarded as a 0-1 integer variable, selected from several locations determined in advance. However, in actual situations, the locations usually cannot be accurately isolated to a few GCSs and the number of GCSs is often unknown, so 0-1MINLPOM was improved in this study. Based on the estimation that there is a maximum of three GCSs in the study area, an improved 0-1 MINLPOM was established to simultaneously identify the number of GCSs (treated as 0-1 integer variable), the location (treated as integer variable) and release history of GCS (treated as continuous variables). The simulation model was constructed as an equality constraint embedded improved 0-1 MINLPOM. In the improved 0-1 MINLPOM solution process, repeatedly calling the simulation model would have incurred a massive computational load and taken a long time. Thus, a surrogate model based on kriging and extreme learning machine (ELM) was established respectively for the simulation model to avoid this shortcoming. The results show that the accuracy of the kriging surrogate model (Krig-SM) was higher compared with the ELM surrogate model (ELM-SM). The improved 0-1 MINLPOM could identify the number, location, and release history of GCSs simultaneously. The accuracy of identifying the number of GCSs was 100%, and the accuracies of identifying the locations and release history were above 91.67% and 90.14%, respectively.
Optimization of well field management to mitigate groundwater contamination using a simulation model and evolutionary algorithm
