(146e) Quantifying Uncertainty in Methane Hydrate Reservoir Simulations Using Monte Carlo Method

Methane gas hydrate deposits occurring in permafrost regions and the deep oceans are known to store a large volume of methane gas. The extraction of methane stored in these deposits can become an efficient replacement for the existing conventional energy resources. On the other hand, extraction of methane, which is a potent greenhouse gas, can reduce the chances of global climate change due to the spontaneous dissociation of gas hydrates. The short term production tests in the last decade clearly indicate the possibility of methane production from the gas hydrate deposits. However, the long term productions still have some geological and technical issues. Reservoir simulation is an extremely useful tool to forecast the production potential and to reduce the capital cost during field test, which require knowledge of geological and operational factors that control the methane production. However, these parameters, such as, porosity, saturation of hydrate, permeability, and initial temperature-pressure conditions are the major source of uncertainty in the reservoir model due to sparse knowledge about the reservoirs. This study aims to quantify uncertainty in simulation predictions for uncertainty in reservoir parameters using Monte Carlo method. We use class-2, confined, oceanic reservoirs.

An In-house multicomponent, multiphase, thermal, 3-D finite volume simulator is used that considers three components - water, methane and hydrate in four phases - gas, aqueous, hydrate and ice. The simulator has been validated against several other simulators in the code comparison study conducted by US DOE and history matched with hydrate formation and dissociation experiments. Energy and mass balance equations are solved in space and time domain to compute the production of gas in a reservoir. A confined aquifer layer below the hydrate zone is considered in this work where depressurization method provides a satisfactory results for gas production.

We consider a Gaussian distribution of porosity, initial hydrate saturation and initial temperature with 7.5% standard deviation from the mean. The results are obtained from 1000 deterministic simulations from each set of variable as a function of time. Total uncertainty in the gas production with ±2σ bound calculated from Monte Carlo simulations. We see that gas production is most sensitive to porosity; however, initial temperature does not affect the production significantly. The mean production profile shows more dispersion and lower cumulative gas production than the deterministic simulations at mean input porosity and the initial hydrate saturation.