A backward automatic differentiation framework for reservoir simulation

Abstract

In numerical reservoir simulations, Newton's method is a concise, robust and, perhaps the most commonly used method to solve nonlinear partial differential equations (PDEs). However, as reservoir simulators incorporate more and more physical and chemical phenomena, writing codes that compute gradients for reservoir simulation equations can become quite complicated. This paper presents an automatic differentiation (AD) framework that is specially designed for simplifying coding and simultaneously maintaining computational efficiency. First a parse tree for a mathematical expression is built and evaluated with the backward mode AD, and then the derivatives with respect to the expression's arguments are transformed to derivatives with respect to the PDE's independent variables. The first stage can be realized either by runtime polymorphism to gain higher flexibility or by compile-time polymorphism to gain faster execution speed; the second stage is realized by linear combinations of sparse vectors, which can be accelerated by recording the target column indices. The AD framework has been implemented in an in-house reservoir simulator. Individual tests on some complex mathematical expressions were carried out to compare the speed of the manual implementation, the runtime polymorphic implementation and the compile-time polymorphic implementation of the differentiation. Then the performance of the three was analyzed in complete simulations. These cases

Review of Field Development Optimization of Waterflooding, EOR, and Well Placement Focusing on History Matching and Optimization Algorithms

“…When compared to most stochastic methods fewer simulation runs are needed for this adjoint well placement method, as only a forward and a reverse or adjoint simulation are required. In addition to adjoint optimization, automatic differentiation poses a potential for improvements in reservoir history matching and optimization [93,94]. Automatic differentiation is generally more accurate than numerical methods and can more easily compute higher deviates and partial derivatives with respect to many inputs in gradient based optimization algorithms.…”

MRST-AD – an Open-Source Framework for Rapid Prototyping and Evaluation of Reservoir Simulation Problems

“…This way, the Jacobi matrices needed in the nonlinear Newton-type iterations can be constructed from the derivatives that are implicitly computed from when evaluating the residual equations. In (Li and Zhang 2014), the authors discuss how to use the alternative backward-mode differentiation to improve computational efficiency.…”

A multi-continuum multiple flow mechanism simulator for unconventional oil and gas recovery

“…UNCONG introduces a backward automatic differentiation algorithm to compute analytical derivatives of complex expressions (Li and Zhang, 2014). For expressions with static symbolic forms, which are the usual cases in reservoir simulators, the algorithm has an execution speed that is one-third of the hand crafted code, hence having little influence on the total time cost of the simulation.…”

“…8. For complicated derivatives, the automatic differentiation library integrated with UNCONG (Li and Zhang, 2014) can help to simplify the code. The calculated porosities and permeabilities, as well as their derivatives with respect to independent variables of the fluid equations, should be stored in the form shown in Fig.…”