Loading...

An Efficient Simulation-Optimization Model for Coastal Groundwater Management

Ketabchi, Hamed | 2014

1232 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 46384 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Ataie-Ashtiani, Behzad
  7. Abstract:
  8. Groundwater in coastal aquifers is one of the essential resources of freshwater in coastal regions, where are heavily populated or industrialized areas, and have critical ecosystems.The management of these resources is a priority to preserve groundwater from serious natural and human-induced threats such as salinization caused by seawater intrusion. The combined simulation-optimization techniques can be used for this purpose in determining optimal planning and management strategies. The implementation of such techniques for complex real-world and large -scale problems has major difficulties in terms of both computational efficiency (solution quality and required computational time) and flexibility in performance. Therefore, further explorations in terms of the applicabity and the efficiency of high-performance tools are required. In this research, we focus on the developing, evaluating, and implementing the available optimization methodologies, a correct and efficient model conceptualizing, implementing some efficient techniques such as parallel processing and surrogate models e.g., artificial neural networks to improve the efficiency of combined simulation-optimization models by reduction the overall computational time and improvement the solutions quality.In the recent decades, evolutionary algorithms with a varying degree of complexities have been focused for resolving optimization problems in the fields of groundwater resources management in coastal aquifers. Surveying the literature associated with the application of evolutionary algorithms in this field demonstrates that previous studies were mostly relied on a few particular algorithms application, mainly genetic algorithm and its variants, to a number of specific problems. These problems are often not representative cases of the variety of feasible processes may be occurred in coastal aquifers. In this study, eight evolutionary algorithms are considered and assessed on four separate benchmark problems with different degree of difficulty and variety. Based on the obtained results, continuous ant colony optimization and particle swarm optimization algorithms are recommended in terms of both solutions quality and required computational time criteria, for application in the coastal groundwater management problems of this study.We mainly focus on the groundwater system of small islands, one of important coastal regions, due to our real-case study, the groundwater system of Kish Island, an arid-region case in the Persian Gulf, Iran. Firstly, an analytical solution describing fresh groundwater lenses in circular islands, composed of two geological layers, is developed for the simplified case of steady-state and sharp-interface conditions. Using this solution, the sensitivity analyses are performed. The obtained results highlight the importance of the geologic conceptualization on fresh froundwater lens prediction and analysis. Together with the important controls of aquifer recharge and coastal slope, the layering properties (thickness and hydraulic conductivity) are fundamental and critical aspects that should be considered in groundwater system assessments on coastal islands.Then, the selected realcase island is focused and the required data to simulate the groundwater system are collected and analyzed. SUTRA, a density-dependent groundwater numerical model is applied for the 2D and 3D simulations while the numerical model parameters are calibrated using PEST, an automated parameter estimation code. A detailed conceptualization of the fresh groundwater lens model is completed to understand the sensitivity of the fresh groundwater lens to some particular aspects of the model prior to analysis of combined simulation-optimization scenarios. The island system is considered as a real-world setup to accomplish an integrated conceptual model comparison. The comparisons suggest that the two-layer properties of the Kish Island aquifer and the unsaturated zone have a significant influence on the fresh groundwater lens while the impacts of lateral-boundary irregularities are negligible. Also, fresh groundwater lens can be affected by sea-level rise, one of climate change impacts and even more so by subjected land-surface inundation and recharge rate variations. Three-dimensionality in the modeling procedure is a significant factor, especially in large-scale 3D systems of small islands, too.Based on the results obtained from the performance evaluations of evolutionary algorithms and the optimum conceptualization of groundwater system, two combined simulation-optimization of Kish island groundwater resources are developed including the 3D model constructed using SUTRA and the artificial neural network as a surrogate model.Shared memory parallelism technique is also implemented in both models. Two issues are of particular interest to the research are: (1) how the general idea of minimizing seawater intrusion can be effectively represented by objective functions within the framework of the simulation-optimization paradigm. Five different models have been compared, including, minimization of the root mean square change in concentrations over the planning period,minimization of the arithmetic mean, minimization of the trimmed arithmetic mean of concentration in the observation wells, maximization of pumped water with the protection of the observation wells from more seawater intrusion, and the last problem condition which the climate change impacts are added. These investigations show that, for example,a reduction of more than 20% for the optimized extraction rate from the proposed coastal aquifer is determined mainly due to the climate change impacts; and (2) how the implications of applying the methodology to a real-world small-island fresh groundwater lens can be performed. For instance, by application of parallel processing technique, a speedup ratio from up to 3.53 is achieved using 2- to 8-core processor. Also, the choice of robust algorithm (continuous ant colony optimization) in comparison with the poorest algorithm causes the value of speedup ratio of more than 2.5. Combined use of these techniques besides the replacing the numerical simulator with artificial neural network leads to a considerable increase for the value of speedup ratio up to 2,562. The obtained results show that integration of the latest innovative tools can provide the ability to solve complex real-world optimization problems in an effective way
  9. Keywords:
  10. Coastal Aquifer ; Sea Water Intrusion ; Kish Island ; Optimal Mnagement ; Efficient Simulation-Optimization Model

 Digital Object List

 Bookmark