Light non-aqueous phase liquids simulation using artificial intelligence models: Esmaeilabad aquifer case study

Abstract

One of the most important pollutants in the aquifers adjacent to oil facilities is LNAPL (Light non-aqueous phase liquids). LNAPL recovery from the aquifer is one of the fastest and most convenient methods for aquifer cleanup. Identifying LNAPL thickness and fluctuations is very important to determine the LNAPL recovering method and maximizing the recovery. The feasibility of two artificial intelligence models including gene expression programming (GEP), adaptive neuro-fuzzy inference system (ANFIS) and classical multivariate linear regression (MLR) techniques are investigated in this study for LNAPL level forecasting. Discharge rate of LNAPL and groundwater level fluctuations were used as input attributes for the developed GEP, ANFIS and MLR models. Based on the comparison of three methods, it was found that the GEP could be successfully utilized in forecasting LNAPL level fluctuations in recovery process. Also, the GEP models can identify the relationship between dependent and independent variables and provide an equation. The identified equation based on GEP can be useful for planning the recovery method and algorithms. The results indicate that there is a high degree of agreement between the predicted values of the GEP based equation and the actual values.