Virginia M. Johnson

Optimal field-scale groundwater remediation using neural networks and the genetic algorithm.

We present a new approach for field-scale nonlinear management of groundwater remediation. First, an artificial neural network (ANN) is trained to predict the outcome of a groundwater transport simulation. Then a genetic algorithm (GA) searches through possible pumping realizations, evaluating the fitness of each with a prediction from the trained ANN. Traditional approaches rely on optimization algorithms requiring sequential calls of the groundwater transport simulation. Our approach processes the transport simulations in parallel and ``recycles`` the knowledge base of these simulations, greatly reducing the computational and real-time burden, often the primary impediment to developing field-scale management models. We present results from a Superfund site suggesting that such management techniques can reduce cleanup costs by over a hundred million dollars.

Applying soft computing methods to improve the computational tractability of a subsurface simulation–optimization problem

Abstract Formal optimization strategies normally evaluate hundreds or even thousands of scenarios in the course of searching for the optimal solution to a given management question. This process is extremely time-consuming when numeric simulators of the subsurface are used to predict the efficacy of a scenario. One solution is to train artificial neural networks (ANNs) to stand in for the simulator during the course of searches directed by some optimization technique such as the genetic algorithm (GA) or simulated annealing (SA). The networks are trained from a representative sample of simulations, which forms a re-useable knowledge base of information for addressing many different management questions. These concepts were applied to a water flood project at BPs Pompano Field. The management problem was to locate the combination of 1–4 injection locations that would maximize Pompanos simple net profit over the next 7 years. Using a standard industry reservoir simulator, a knowledge base of 550 simulations sampling different combinations of 25 potential injection locations was created. The knowledge base was first queried to answer questions concerning optimal scenarios for maximizing simple net profit over 3 and 7 years. The answers indicated that a considerable increase in profits might be achieved by expanding from an approach to injection depending solely on converting existing producers to one involving the drilling of three to four new injectors, despite the increased capital expenses. Improved answers were obtained when the knowledge base was used as a source of examples for training and testing ANNs. ANNs were trained to predict peak injection volumes and volumes of produced oil and gas at 3 and 7 years after the commencement of injection. The rapid estimates of these quantities provided by the ANNs were fed into net profit calculations, which in turn were used by a GA to evaluate the effectiveness of different well-field scenarios. The expanded space of solutions explored by the GA contained new scenarios that exceeded the net profits of the best scenarios found by simply querying the knowledge base. To evaluate the impact of prediction errors on the quality of solutions, the best scenarios obtained in searches where ANNs predicted oil and gas production were compared with the best scenarios found when the reservoir simulator itself generated those predictions during the course of search. Despite the several thousand CPU hours required to complete the simulator-based searches, the resulting best scenarios failed to match the best scenarios uncovered by the ANN-based searches. Lastly, results obtained from ANN-based searches directed by the GA were compared with ANN-based searches employing an SA algorithm. The best scenarios generated by both search techniques were virtually identical.

Chapter 18 Applying soft computing methods to improve the computational tractability of a subsurface simulation-optimization problem

Abstract Formal optimization strategies normally evaluate hundreds or even thousands of scenarios in the course of searching for the optimal solution to a given management question. This process is extremely time-consuming when numeric simulators of the subsurface are used to predict the efficacy of a scenario. One solution is to train artificial neural networks (ANNs) to stand in for the simulator during the course of searches directed by some optimization technique such as the genetic algorithm (GA) or simulated annealing (SA). The networks are trained from a representative sample of simulations, which forms a re-useable knowledge base of information for addressing many different management questions. These concepts were applied to a water flood project at BPs Pompano Field. The management problem was to locate the combination of 1-4 injection locations which would maximize Pompanos simple net profit over the next seven years. Using a standard industry reservoir simulator, a knowledge base of 550 simulations sampling different combinations of 25 potential injection locations was created. The knowledge base was first queried to answer questions concerning optimal scenarios for maximizing simple net profit over three and seven years. The answers indicated that a considerable increase in profits might be achieved by expanding from an approach to injection depending solely on converting existing producers to one involving the drilling of three to four new injectors, despite the increased capital expenses. Improved answers were obtained when the knowledge base was used as a source of examples for training and testing ANNs. ANNs were trained to predict peak injection volumes and volumes of produced oil and gas at three and seven years after the commencement of injection. The rapid estimates of these quantities provided by the ANNs were fed into net profit calculations, which in turn were used by a GA to evaluate the effectiveness of different well-field scenarios. The expanded space of solutions explored by the GA contained new scenarios which exceeded the net profits of the best scenarios found by simply querying the knowledge base. To evaluate the impact of prediction errors on the quality of solutions, the best scenarios obtained in searches where ANNs predicted oil and gas production were compared with the best scenarios found when the reservoir simulator itself generated those predictions during the course of search. Despite the several thousand CPU hours required to complete the simulator-based searches, the resulting best scenarios failed to match the best scenarios uncovered by the ANN-based searches. Lastly, results obtained from ANN-based searches directed by the GA were compared with ANN-based searches employing an SA algorithm. The best scenarios generated by both search techniques were virtually identical.

Improving Gas Storage Development Planning Through Simulation-Optimization

This is the first of two papers describing the application of simulator-optimization methods to a natural gas storage field development planning problem. The results presented here illustrate the large gains in cost-effectiveness that can be made by employing the reservoir simulator as the foundation for a wide-ranging search for solutions to management problems. The current paper illustrates the application of these techniques given a deterministic view of the reservoir. A companion paper will illustrate adaptations needed to accommodate uncertainties regarding reservoir properties.

Chapter 4 Heuristic approaches to combinatorial optimization

Publisher Summary This chapter provides an intuitive understanding of the reasoning behind the choice of two heuristic search techniques, the genetic algorithm (GA) and simulated annealing (SA), to optimize the reservoir engineering problem. The existing production wells are identified by black circles. Now suppose that the optimization problem is to identify the subset of the five prospective injection locations which will maximize the productivity of the field, defined perhaps as the fields Net Present Value (NPV). The discretization of the decision variable is not done to simplify the problem. To the contrary, discretization actually complicates the optimization problem by making it difficult to apply the most efficient and powerful search techniques. Rather, the discretization is tolerated when it is the most natural representation of the real-world problem to be solved.

Optimal Groundwater Remediation Using Artificial Neural Networks

The significant cost and complexity of groundwater remediation and water resources management has encouraged integration of optimization techniques with groundwater flow and transport modeling to search for efficient groundwater management strategies. This integration of methodologies has often been referred to as simulationmanagement modeling or simulation-optimization groundwater management modeling. The values of the objective function and constraints of the optimization problem are calculated by the groundwater flow and transport models run as a submodel of the optimization driver. In the area of remediation, example objectives could be minimizing costs or maximizing contaminant mass removed; constraints might be avoiding dewatering or a total pumping volume limit. A general groundwater management model will use optimization techniques to search among almost infinite numbers of treatment or control strategies possibilities for ones that meet management goals while minimizing cost. The main advantage of applying these mathematical tools to decision-making problems is that they are less restricted by human imagination than case-by-case comparisons. As the number of competing engineering, economic, and environmental planning objectives and constraints increases, it becomes difficult for human planners to track complex interactions and select a manageable set of promising scenarios for examination. Using optimization techniques, the search can range over all possible combinations of variables, locating strategies whose effectiveness is not always obvious to planners.