FMU PEM

Appearance

Model file formats ​

This section targets developers and rock physics specialists responsible for (re)calibrating model parameter files.

Model files are used in three contexts: (1) the pressure sensitivity model, (2) the T-Matrix EXP variant, and (3) the T-Matrix PETEC variant. All model files are in pickle format. Examples can be found in the test data sets for fmu-pem in the project GitHub repository under test/data/sim2seis/model.

Pressure model files ​

Velocity sensitivity to effective pressure is implemented in the module rock-physics-open . As the design supports generic model wrappers, the information is split between two .pkl files:

  • Metadata file (model descriptor): keys model_type, pointer to core parameter file (nn_mod), input feature list, label variable and units, optional preprocessing artefact references.
  • Parameter file: numerical calibration coefficients (a_factor, b_factor), valid pressure range (model_max_pressure), provenance (description). Briefly define a_factor and b_factor (exponential law coefficients for pressure sensitivity of velocity).
python
import os, pickle
os.chdir(r'./tests/data/sim2seis/model')

print("Metadata file:")
with open('carbonate_pressure_model_vs_exp.pkl', 'rb') as fin:
    pres_vs_dict = pickle.load(fin)
for key, value in pres_vs_dict.items():
    print(f'{key}: {value}')

print("Parameter file:")
with open('vs_exp_model.pkl', 'rb') as fin:
    vs_exp_model_dict = pickle.load(fin)
for key, value in vs_exp_model_dict.items():
    print(f'{key}: {value}')

Output:

text
Metadata file:
model_type: Exponential
nn_mod: vs_exp_model.pkl
scaler: None
ohe: None
label_var: VSX
label_units: m/s
feature_var: ['VSX', 'PEFF_in_situ', 'PEFF_depleted']

Parameter file:
a_factor: 0.09164501
b_factor: 8.23144402
model_max_pressure: 40.0
description: plug-based calibration data

T-Matrix EXP model ​

The EXP type T-Matrix optimisation contains parameters for mineral properties of carbonate and shale, as well as geometry, connectivity and anisotropic parameters. Note that the mineral properties are given as a fraction of the upper bound in the optimisation process (values near 1.0 indicate convergence at the imposed cap). Calibration is performed in rock-physics-open.

Parameter definition:

  • f_ani: anisotropy (crack alignment / directional weighting) factor.
  • f_con: connectivity (effective fluid communication / crack network) factor.
  • alpha_opt: array of pore aspect ratios (e.g. alpha_opt[0] equant / stiff, alpha_opt[1] crack-like).
  • v_opt: optimised porosity (or volume fraction parameter; clarify if it is total porosity vs a specific pore system).
  • k_carb, mu_carb, rho_carb, etc.: normalised carbonate bulk modulus, shear modulus, density; similarly for shale (k_sh, mu_sh, rho_sh).
  • opt_vec: concatenated optimisation vector in parameter order (state vector snapshot).
python
import os, pickle
os.chdir(r'./tests/data/sim2seis/model')

with open('t_mat_params_exp.pkl', 'rb') as fin:
    exp_params = pickle.load(fin)
for key, value in exp_params.items():
    print(f'{key}: {value}')

Output:

text
well_name: unknown
opt_ver: exp
f_ani: 0.1931401051312247
f_con: 0.0006487365851106763
alpha_opt: [0.50024826 0.07042638]
v_opt: 0.5016505733981023
k_carb: 0.6029058965366181
mu_carb: 0.7274039302389138
rho_carb: 0.9999862334411195
k_sh: 0.9999999999999999
mu_sh: 0.9999999999999999
rho_sh: 0.9090909090909092
opt_vec: [1.93140105e-01 6.48736585e-04 5.00248264e-01 7.04263774e-02
 5.01650573e-01 6.02905897e-01 7.27403930e-01 9.99986233e-01
 1.00000000e+00 1.00000000e+00 9.09090909e-01]

T-Matrix PETEC model ​

The PETEC model type of T-Matrix contains fewer parameters than the EXP model, as in an appraisal or production phase, more detailed information about minerals and fluids is available, and can be applied directly in the modelling. Optimisation code is located in rock-physics-open.

python
import os, pickle
os.chdir(r'./tests/data/sim2seis/model')

with open('t_mat_params_petec.pkl', 'rb') as fin:
    exp_params = pickle.load(fin)
for key, value in exp_params.items():
    print(f'{key}: {value}')

Output:

text
well_name: unknown
opt_ver: min
f_ani: 0.11140201779907752
f_con: 0.9999999998567579
alpha_opt: [0.5        0.06980489]
v_opt: 0.5000000000000977
opt_vec: [0.11140202 1.         0.5        0.06980489 0.5       ]