Well-Placement Optimization with Gradient-Base Methods

Oil production optimization is one of the main concerns of the reservoir management. In this regard, optimal location of wells can lead to a significant increase in the oil production. The purpose of this research is to investigate a new extension of the Automatic well-placement optimization technique maximizing the net present value (NPV) of the oil production. Since finding the optimal location of wells may require expensive and time consuming iterations through the reservoir simulator, we propose a deterministic (gradient-based) algorithm to address this issue. Our approach is consisted of searching the neighbourhood of the initial well which is called the pseudo-wells in order to find the optimal location of wells. Since these pseudo-wells inject or produce at a very low rate, they have a minor effect on the overall flow throughout the reservoir. In this work, we first calculate the gradient of NPV with respect to the flow rate in the pseudo-wells using an adjoint-based method. This helps us to find improving directions on the basis of which the optimal locations of wells can be determined. This searching method continues until no further improvement in the NPV is achieved. The main contribution to our work compares to last papers of Well-Placement, is that instead of using a fixed step size, a variable step size strategy is used which reduces the total convergence time. The method is applied to three waterflooding cases.