Combining Sparse Data with Reaction Kinetics Using Fuzzy Logic to Predict Reservoir Souring

Abstract

\n Hydrogen Sulphide (H2S) is a colourless, flammable and highly toxic gas with a strong odour of rotten eggs that is found in many reservoir fluids and aquifers in the world. This gas is commonly a result of reservoir souring – a process which increases the H2S concentration. Increasing amounts of this gas pose serious health, safety and environmental concerns. This can result in significant costs associated with replacement of downhole and surface equipment and increased processing costs, but more lethally a potential loss of life.\n Many reservoirs particularly those undergoing waterflooding face increasing levels of hydrogen sulphide (H2S) production with time. H2S is a highly toxic gas that can be fatal even at low concentrations. Being able to predict the risk potential of a particular reservoir to increasing H2S production with time would be highly valuable. The objective is to determine apriori whether a reservoir would likely see dangerously high levels of H2S being produced during the lifetime of the reservoir, and if so, be a catalyst in supporting further investigation and mitigation of H2S early in the reservoir development.\n There is very little published field data with regards to reservoir souring, hence a purely data driven model would not be possible to create. However, we do have a good understanding of the reaction kinetics that goes into the biological process that generates H2S. To this end the best modelling paradigm that can assimilate sparse data with first principles dynamics is fuzzy logic. A fuzzy logic model has been built around the reaction kinetics and then conditioned to the published field data. The model created matches the published field data fairly well. It is now a ready tool that can be used by engineers to make a quick assessment of their reservoirs before going into full blown expensive sampling and laboratory analysis.\n The novel aspect of this paper is being able to use fuzzy logic to combine the first principles chemistry together with sparse data to produce a model that can be used practically. Fuzzy Logic has been out of the news of late as machine learning and neural networks are the current hot potatoes, however it is often overlooked that fuzzy logic can still be used in low dimensional cases where only sparse data is available.