Application of artificial intelligence deep learning in numerical simulation of seawater intrusion.

Seawater intrusion not only causes fresh water shortages in coastal areas, but also has a negative impact on regional economic and social development. Global climate change will affect precipitation, sea level, and many other factors, which will in turn affect the simulation and prediction results for seawater intrusion. By combining groundwater numerical simulation technology, an atmospheric circulation model, artificial intelligence methods, and simulation optimization methods, this study coupled a numerical simulation model of seawater intrusion with an optimization model to optimize the groundwater exploitation scheme in the study area under the condition of climate change. As a result, a groundwater exploitation scheme was obtained for a typical study area, which provided a scientific basis and a reference for the rational development of effective groundwater resource solutions. The results of this study can be described as follows. (1) By introducing the theory and method of deep learning from artificial intelligence, the problem of complex nonlinear mapping between the inputs and outputs of a three-dimensional variable-density seawater intrusion numerical simulation model under the condition of limited number of training samples is effectively solved, and the approximation accuracy of the surrogate model with respect to the simulation model is improved. (2) By solving the optimization model, a reasonable groundwater exploitation scheme was obtained, which provided a scientific basis for the rational development and efficient use of groundwater resources in the study area.

Optimal design of groundwater pollution monitoring network based on the SVR surrogate model under uncertainty.

The simulation-optimization method is widely used in the design of the groundwater pollution monitoring network (GPMN). The uncertainty of the simulation model will significantly affect the design results of GPMN. When the Monte Carlo method is used to consider the influence of model uncertainty on the optimization results, the simulation model needs to be invoked many times, which will cause a huge amount of calculation. To reduce the calculation load, the study proposed to use the support vector regression (SVR) method to construct the surrogate model to couple the simulation model and the optimization model in the optimal design of GPMN. The optimization goal is to maximize the accuracy of the spatial description of pollution plume in each monitoring period. The study also considered the dynamic changes in the migration and morphological of pollution plumes in the optimization of GPMN. Finally, the West Shechang coal gangue pile in Fushun of China was used as a case study to verify the effectiveness of the above method. The results demonstrate that the SVR surrogate model can fit the input-output relationship of the simulation model to a high degree with less computation. The optimized monitoring network can reveal essential and comprehensive information about pollution plumes. The study provides a stable and reliable method for the design of GPMN.

Multiobjective optimization of the groundwater exploitation layout in coastal areas based on multiple surrogate models.

Seawater intrusion is a common problem in coastal areas. The rational distribution of groundwater exploitation can minimize the scope of seawater intrusion and maximize groundwater exploitation. In this study, an optimization method for the groundwater exploitation layout in coastal areas was proposed. Based on the numerical simulation model of variable-density groundwater, a multiobjective groundwater management model was constructed with the objectives of maximizing groundwater exploitation and minimizing seawater intrusion. The optimization model was solved by nondominated sorted genetic algorithm-II (NSGA-II). To improve the computational efficiency of the optimization model, the surrogate models of the groundwater simulation model were built by using three different methods: kriging, support vector regression (SVR), and kernel extreme learning machines (KELM). Finally, the above methods were tested in Longkou City of China. The results show that the use of surrogate models can greatly reduce the computing time for solving seawater intrusion management problems. The surrogate model of the variable-density groundwater simulation model based on the SVR method has the best performance. The groundwater exploitation layout optimized by the above method is reasonable and can reflect the actual hydrogeological conditions in the study area. This study provides a reliable way to optimize the groundwater exploitation layout in coastal areas.

Modeling and uncertainty analysis of seawater intrusion based on surrogate models.

When using a simulation model to study seawater intrusion (SI), uncertainty in the parameters directly affects the results. The impact of the rise in sea levels due to global warming on SI cannot be ignored. In this paper, the Monte Carlo method is used to analyze the uncertainty in modeling SI. To reduce the computational cost of the repeated invocation of the simulation model as well as time, a surrogate model is established using a radial basis function (RBF)-based neural network method. To enhance the accuracy of the substitution model, input samples are sampled using the Latin hypercube sampling (LHS) method. The results of uncertainty analysis had a high reference value and show the following: (1) The surrogate model created using the RBF method can significantly reduce computational cost and save at least 95% of the time needed for the repeated invocation of the simulation model while maintaining high accuracy. (2) Uncertainty in the parameters and the magnitude of the rise in sea levels have a significant impact on SI. The results of prediction were thus highly uncertain. In practice, it is necessary to quantify uncertainty to provide more intuitive predictions.

The application of 0-1 mixed integer nonlinear programming optimization model based on a surrogate model to identify the groundwater pollution source.

The optimization model is presently used for the identification of pollution sources and it is based on non-linear programming optimization. The decision variables in this model are continuous, resulting in a weak recognition of integer variables including pollution source location. In addition, as the number of pollution sources increase, so the calculated load increases exponentially and accuracy decreases. Compared with previous studies, this study makes a series of improvements by adopting a 0-1 mixed integer nonlinear programming optimization model to enable the simultaneous identification of both location (integer variable) and the release intensity (continuous variable) of the pollution source. One of the constraints in the optimization model is a simulation component which requires thousands of calls during the calculation process and therefore requires considerable computational load. To avoid this problem, the Kriging surrogate model is established in this study to reduce computational load, while at the same time ensuring the accuracy of the simulation results. The identification result is solved using a genetic algorithm (GA) and represents the real location of the pollution source, while release intensities are close to actual ones with small relative errors. The Kriging surrogate model is based on a 0-1 mixed integer nonlinear programming optimization model and can simultaneously identify both the location and the release intensity of the pollution source with a high degree of accuracy and by using short computational times.

Application of ensemble surrogates and adaptive sequential sampling to optimal groundwater remediation design at DNAPLs-contaminated sites.

In this study, we aimed to develop an optimal groundwater remediation design for sites contaminated by dense non-aqueous phase liquids by using an ensemble of surrogates and adaptive sequential sampling. Compared with previous approaches, our proposed method has the following advantages: (1) a surrogate surfactant-enhanced aquifer remediation simulation model is constructed using a Gaussian process; (2) the accuracy of the surrogate model is improved by constructing ensemble surrogates using five different surrogate modelling techniques, i.e., polynomial response surface, radial basis function, Kriging, support vector regression, and Gaussian process; (3) we conducted comparisons and analyses based on 31 surrogate models derived from different combinations of the five surrogate modelling techniques; and (4) the reliability of the optimal solution was improved by implementing adaptive sequential sampling. The two proposed methods were applied to a hypothetical perchloroethylene-contaminated site in order to demonstrate their performance. The results showed that the best surrogate model integrated all five of the surrogate modelling methods, with an R2 value of 0.9913 and a root mean squared error of 0.0159, thereby demonstrating the advantage of using ensemble surrogates. In addition, the reliability of the optimization model solution was improved by adaptive sequential sampling, which avoided false solutions.