Identification of influential parameters through sensitivity analysis of the TOUGH + Hydrate model using LH-OAT sampling

The over-parameterisation, as a common issue in numerical models, may be unavoidable in the TOUGH + Hydrate (T + H) model. Therefore, a global sensitivity analysis of the influence of main or uneasily-measured 26 input parameters on five model outputs, i.e. cumulative gas production, gas production rate, hydrate saturation, gas saturation and aqueous saturation, under depressurization and brine-inhibitor injection was performed using the LH-OAT sampling method. The 26 parameters were classified as ‘Very important’, ‘Important’, ‘Slightly important’, and ‘Unimportant’ by analyzing their sensitivity indices at different sampling times and sampling positions in the modeling area, using the TSPAC segmentation package. The sensitivity of an input parameter to a model output was dependent on sampling time, sampling position and the identity of the model output in question. Cumulative gas production and gas production rate were governed by four permeability-related parameters (parameter n and residual aqueous saturation (in the relative permeability-saturation model), permeability reduction exponent and critical mobile phase saturation) and four inhibitor-related parameters (inhibitor density, the specific enthalpy of the dissolution of the inhibitor in water, reference inhibitor mole fraction in the aqueous phase and inhibitor-induced reference temperature depression). Hydrate saturation was governed by two hydrate dissociation-related parameters (temperature-dependent hydrate specific heat and hydrate density) and two parameters related to the flow of released fluids in the pore network (permeability and pore structure expansion). Gas saturation was governed by several types of input parameters: six parameters related to the flow of pore gas (parameter n and residual aqueous saturation (in the relative permeability-saturation model), critical mobile phase saturation, intrinsic permeability, permeability reduction exponent and pore compressibility), three inhibitor-related parameters (the specific enthalpy of the dissolution of the inhibitor in water, reference inhibitor mole fraction in the aqueous phase and inhibitor-induced reference temperature depression) and two hydrate dissociation-related parameters (temperature-dependent hydrate density polynomial and temperature-dependent hydrate specific heat polynomial). Finally, aqueous saturation was governed by five permeability-related parameters (parameter λ and residual aqueous saturation (in the saturation-capillary pressure function model), pore expansion, grain specific heat and the tortuosity factor model). Overall, the most influential parameter for all five model outputs was the parameter n in the relative permeability-saturation function model.