Acid Gas Re-Injection System Design Using Machine Learning

Abstract

An “energy evolution” is necessary to manifest an environmentally sustainable world while meeting global energy requirements, with natural gas being the most suitable transition fuel. Covering the ever-increasing demand requires exploiting lower value sour gas accumulations, which involves an acid gas treatment issue due to the greenhouse gas nature and toxicity of its constituents. Successful design of the process requires avoiding the formation of acid gas vapor which, in turn, requires time-consuming and complex phase behavior calculations to be repeated over the whole operating range. In this work, we propose classification models from the Machine Learning field, able to rapidly identify the problematic vapor/liquid encounters, as a tool to accelerate phase behavior calculations. To set up this model, a big number of acid gas instances are generated by perturbing pressure, temperature, and acid gas composition and offline solving the stability problem. The generated data are introduced to various classification models, selected based on their ability to provide rapid answers when trained. Results show that by integrating the resulting trained model into the gas reinjection process simulator, the simulation process is substantially accelerated, indicating that the proposed methodology can be readily utilized in all kinds of acid gas flow simulations.

Applications of Machine Learning in Subsurface Reservoir Simulation—A Review—Part I

“…As a result, the recommended methodology was shown to be able to adapt to all types of acid gas flow simulations. Recently, Anastasiadou et al [39] moved similarly by trying to solve the phase stability problem, this time for an acid gas reinjection system where the required phase behavior calculations are more complex and time-consuming since they need to be repeated for an even broader compositional space to cover for the acid components (H 2 S and CO 2 ) and the hydrocarbon contaminants that are being reinjected into the reservoir. The authors proposed three classification ML approaches, ANNs, decision trees (DTs) and SVMs, to solve the phase stability problem, which is crucial in acid gas reinjection designs, at a fraction of the time needed by conventional iterative methods.…”

“…In this kind of classification problem, the input consists of fluid composition, pressure and temperature values for the classifier to reach a solution (stable or unstable mixture) based on the phase boundaries of the p-T phase diagrams. That way, the trained classifier can replace the traditional iterative stability algorithm and substantially accelerate flow simulations with its direct non-iterative predictions [39].…”

“…SVMs and ANNs [39,53,55,57,58] Phase stability/split [10,54,56,59,65] Phase split [22] Phase split Deep learning ANNs [60,64] Phase stability/split [61][62][63] Black oil PVT properties Supervised Z-factor prediction ANNs [67,[70][71][72][73][74] SVMs and LSSVM with optimization methods [75,76,78,83] Kernel Ridge Regression [80] Saturation pressure ANNs [77,85,92,95,96] SVMs [90,91] ANFIS [93] ANNs coupled with optimization methods [97,98] SVMs and LSSVM with optimization methods [99,100,102] SVMs and DTs [101] For the ML strategies concerning HM, two approaches have dominated the research area. The first entails indirect HM which defines the difference between the real and the predicted production data as the model's output.…”

Enhancement of Machine-Learning-Based Flash Calculations near Criticality Using a Resampling Approach

“…The ANN was trained using an extensive dataset obtained from the simulation of various gas recycling schemes, covering any possible compositional changes that might occur inside a reservoir to account for the large compositional variability in the gas reinjection process. Later, Anastasiadou et al [46] progressed similarly by trying to solve the phase stability problem for an even more complex acid gas reinjection system. The authors proposed three classification approaches, ANNs, decision trees (DTs) and SVMs, to solve the phase stability problem, using a large ensemble of training data.…”

Prospects of an Acid Gas Re-Injection Process into a Mature Reservoir

“…Additionally, the dissolved acid gas is produced and separated at the surface, thus allowing for its recycling at a continuously increasing quantity. High enough reservoir pressure is a prerequisite to ensure single-phase mixing, known as miscibility [18][19][20].…”