Application of simulation- optimization model to control groundwater level in constructional projects (Case study: Metro network of Tabriz)

Groundwater level control is an important issue in construction projects. In order to find the flow rate and appropriate locations for pumping the simulation models are often used. The problem with the simulation models is the number of simulation runs to find the best possible scenario. Even though it is not guaranteed to have the optimum solution. However, combination of optimization and simulation models can be used to predict the groundwater behavior and to find the optimal operation scenario. Two main approaches in groundwater optimization-simulation modeling are solving linear and nonlinear relationships between groundwater level and extraction. Using nonlinear relationships leads to complex formulation and the computation cost is usually high. The objective of this research was to use a simulation-optimization approach to find the optimum groundwater model operation in conjunction with the construction projects. The proposed approach was tested for Tabriz Aquifer in northwest of Iran. Tabriz City is a growing city and it is the capital of the East Azarbaijan Province. As part of the city’s development plan, Tabriz subway is under construction and main parts of the tunnels is below the groundwater table. The subway plan consists of four lines within the City limits and Line 1 is approximately 18 km. Due to the shallow thickness of the aquifer, the tunnel in Line 1 obstructs the groundwater flow in many areas. Some of the stations are also constructed below the groundwater level. The path of the Line 1 is located in the area which the groundwater level is already high and close to the ground surface. The basements of many houses in this area suffer from the groundwater seepage and the municipality of Tabriz is pumping continuously groundwater to decline the groundwater level. In order to formulate the relationship between groundwater extraction and water level, MODFLOW was used to simulate the groundwater behavior in Tabriz Aquifer. MODFLOW is one of the most commonly used groundwater models in the world and has been tested in many areas. Tabriz Aquifer is a two-layer aquifer with shallow unconfined aquifer on top and deep confined layer at the bottom separated by a clay aquitard. Groundwater level at monitoring wells was used to calibrate and validate the groundwater flow model using MODFLOW. Hydraulic conductivity, groundwater extraction, porosity, and infiltration rate at each cell were the calibration parameters. Each optimization problem consists of a set of decision variables, an objective function, and a set of constraints. Groundwater extraction by wells in each management cell and a binary variable which was 37 wells along the subway line were considered as the decision variables. The binary variable was used to determine if a well is required to be added or not. The objective function was the cost of operation of existing wells and increasing costs due to adding new wells. The objective function was subject to constrains including the upper and lower limits of the well discharge, upper and lower limits for groundwater level fluctuation, and maximum and minimum number of active wells in each management cell. GWM software was used for optimizing the groundwater level. GWM uses the linear programming and links the results of the MODFLOW to the optimization part. It is developed using the FORTRAN programming language and uses a response-matrix approach to solve linear, nonlinear, and mixed-binary linear groundwater management formulations. Several simulation runs were used to determine the upper and lower limits of the pumping rates considering the capacity of the aquifer, well and geological conditions. The results of these simulations were used to extract a linear relationship between the groundwater level and pumping rate in each management cell.
The objective of the optimization was to reduce the water level below the subway tunnel level, to facilitate the construction, operation and maintenance of the tunnels. The priority was given to the use of the existing wells by changing the operation schedule or increasing the well discharge rate and the second option was to drill new wells. The initial results showed that the existing wells alone cannot reduce the water level to desired depth. The results showed that by increasing the groundwater pumping rates of existing wells in six management cells and adding only 13 new wells out of 37 proposed wells, it is possible to withdraw the groundwater level below the subway tunnel within three years. To check the impact of the linear assumption, the results of the optimization model were used in a simulation model to predict the groundwater level. The model showed that, assuming linear behavior between groundwater level and pumping rate has little impact on the accuracy of the results.