Methods for predicting formation properties
A method for predicting a formation property of a formation, including the steps of: drilling a first well that penetrates the formation, obtaining well properties from the first well, determining a measured formation property from the first well, determining a correlation function that estimates a predicted formation property from the well properties, wherein the correlation function comprises coefficients, calibrating the correlation function by modifying the coefficients until the predicted formation property best matches the measured formation property, drilling a second well that penetrates the formation, obtaining well properties from the second well, determining the formation property from the second well by applying the correlation function with the modified coefficients to the well properties.
1. A method for predicting a formation property of a formation, comprising the steps of:
drilling a first well that penetrates the formation,
obtaining well properties from the first well,
determining a measured formation property from the first well,
determining a correlation function that estimates a predicted formation property from the well properties, wherein the correlation function comprises coefficients,
calibrating the correlation function by modifying the coefficients until the predicted formation property best matches the measured formation property,
drilling a second well that penetrates the formation,
obtaining well properties from the second well,
determining the formation property from the second well by applying the correlation function with the modified coefficients to the well properties,
using, by a wellbore stability model, the formation property to plan an optimal well trajectory of the second well,
using the formation property to plan a mud program that minimizes instabilities of the second well;
using the formation property to recommend an optimal lost circulation material (LCM) blend to mitigate mud loss in the second well, and
drilling the second well,
wherein the coefficients for k correlation functions are modified by the following steps:
determining a coefficient adjustment vector for k well property vectors,
outputting k predicted formation properties by the correlation function,
y
1
C
=
f
(
x
1
,
θ
)
y
2
C
=
f
(
x
2
,
θ
)
⋮
y
k
C
=
f
(
x
k
,
θ
)
(
3
)
determining k measured formation properties corresponding to the k well property vectors, and
adjusting the coefficient vector such that the adjusted coefficients yields the optimal predicted formation property that best matches the measured formation property.
2. The method according to claim 1 , wherein the well properties are determined in the lab or at the well.
3. The method according to claim 1 , wherein the formation property comprises unconfined compressive strength (UCS).
4. The method according to claim 3 , wherein the UCS is determined in a lab from a core sample using the triaxial test.
5. The method according to claim 1 , wherein the formation property further comprise in-situ stresses & pore pressure, elastic properties, poroelastic properties, rock strength properties, Young's modulus, Poisson's ratio, undrained Poisson's ratio, cohesion, friction angle, minimum horizontal stress, pore pressure, permeability, Biot's coefficient, and volumetric thermal expansion coefficient.
6. The method according to claim 1 , wherein the well property comprises gamma ray, porosity, density, resistivity, compressional wave transit time, shear wave transit time, temperature, and caliper.
7. The method according to claim 1 , wherein the well properties, the formation properties, the correlation functions, and the coefficients are stored in a database of a computer.
8. The method according to claim 1 , wherein several correlation functions are determined, and one correlation function is selected that estimates a predicted formation property from the well properties.
9. The method according to claim 8 , wherein the selected correlation function is selected based on a ranking of the correlation functions.
10. The method according to claim 9 , wherein the ranking is performed by a score based on a performance metric assigned to each correlation function.
11. The method according to claim 10 , wherein the performance metric comprises mean squared error (MSE), mean absolute error (MAE), mean absolute percentage error (MAPE), or coefficient of determination (R 2 ).
12. The method according to claim 1 , wherein the coefficient vector is adjusted by an adjustment vector, and the adjustment vector is determined by δθ=(A T A) −1 A T b, where
A
=
[
∇
θ
f
(
x
1
,
θ
)
∇
θ
f
(
x
2
,
θ
)
⋮
∇
θ
f
(
x
k
,
θ
)
]
,
and
(
4
)
b
=
[
y
1
M
-
y
1
C
y
2
M
-
y
2
C
⋮
y
k
M
-
y
k
C
]
.
13. The method according to claim 12 , wherein the coefficient vector is updated iteratively by repeating Eq. (3) and (4).
14. The method according to claim 13 , wherein Eq. (3) and (4) are repeated until the length of the adjustment vector is smaller than a threshold, or until the improvement to a performance metric measuring the match between the predicted formation properties and the measured formation properties is smaller than a threshold.
15. The method according to claim 14 , wherein the performance metric comprises mean squared error (MSE), mean absolute error (MAE), mean absolute percentage error (MAPE), coefficient of determination (R 2 ) are used.
16. The method according to claim 1 , wherein the measured formation property is determined from core plugs or drill cuttings.