-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtest_kerinci_map_estimate.py
257 lines (202 loc) · 9.74 KB
/
test_kerinci_map_estimate.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
import numpy as np
import timeit
import pickle
import os
import errno
from lib import GeothermalCore as GC
from lib import InverseCore as IC
from scipy.optimize import minimize
import pandas as pd
#---truth
#perm_powers_truths = np.log10(np.array([5.00000000e-14, 1.00000000e-14, 1.00000000e-15,
# 5.00000000e-16, 1.00000000e-15, 2.00000000e-14,
# 2.50000000e-14, 1.00000000e-14, 5.00000000e-16,
# 5.00000000e-16, 5.00000000e-16, 1.00000000e-14]))
# tighter cap
#perm_powers_truths = np.log10(np.array([5.00000000e-14, 1.00000000e-14, 1.00000000e-16,
# 1.00000000e-16, 1.00000000e-15, 2.00000000e-14,
# 5.00000000e-16, 5.00000000e-16, 1.00000000e-14]))
# 2.50000000e-14, 1.00000000e-14, 5.00000000e-16,
perm_powers_precal = np.log10(np.array([
1.00000000e-15, 1.00000000e-15, 1.00000000e-15,
2.00000000e-15, 2.00000000e-15, 8.00000000e-15,
5.00000000e-16, 5.00000000e-16, 5.00000000e-16,
1.00000000e-16, 1.00000000e-16, 1.00000000e-16,
5.00000000e-14, 5.00000000e-14, 5.00000000e-16,
5.00000000e-14, 5.00000000e-14, 5.00000000e-16,
5.00000000e-14, 1.00000000e-14, 3.00000000e-14,
5.00000000e-15, 5.00000000e-15, 5.00000000e-15,
5.00000000e-14, 5.00000000e-14, 5.00000000e-16,
5.00000000e-14, 5.00000000e-14, 5.00000000e-16]))
#---coarse process model
process_model_coarse = GC.GeoModel(name='test_process_model_kerinci',
datfile_name='input-files/kerinci/coarse-model/Keriv0_027',
incon_name='input-files/kerinci/coarse-model/Keriv0_027',
geom_name='input-files/kerinci/coarse-model/gKerinci_v0')
list_of_obs_wells = ['LP002','LP001','KRDB1']
process_model_coarse.rename_wells_as_obs(list_of_obs_wells)
process_model_coarse.set_rock_permeabilities(perm_powers=perm_powers_precal)
#list_of_obs_wells = ['LP002'] #['LP002','LP001','KRDB1']'KRDB1']
#---fine process model
process_model_fine = GC.GeoModel(name='test_process_model_kerinci_fine',
datfile_name='input-files/kerinci/fine-model/Keriv1_027',
incon_name='input-files/kerinci/fine-model/Keriv1_027',
geom_name='input-files/kerinci/fine-model/gKerinci_v1')
process_model_fine.rename_wells_as_obs(list_of_obs_wells)
process_model_fine.set_rock_permeabilities(perm_powers=perm_powers_precal)
# ---Layered
# perm_powers_truths = np.log10(np.array([
# 1.00000000e-15, 1.00000000e-15,
# 2.00000000e-15, 8.00000000e-15,
# 5.00000000e-16, 5.00000000e-16,
# 1.00000000e-16, 1.00000000e-16,
# 5.00000000e-14, 5.00000000e-16,
# 5.00000000e-14, 5.00000000e-16,
# 5.00000000e-14, 3.00000000e-14,
# 5.00000000e-15, 5.00000000e-15,
# 5.00000000e-14, 5.00000000e-16,
# 5.00000000e-14, 5.00000000e-16]))
#
# process_model_coarse = GC.GeoModel(name='test_process_model_kericini',
# datfile_name='input-files/kericini/coarse-model/Keriv0_027',
# incon_name='input-files/kericini/coarse-model/Keriv0_027',
# geom_name='input-files/kericini/coarse-model/gKerinci_v0',
# islayered=True)
# --- data model object
real_data_model = GC.GeoModel(name='test_data_model_kerinci',
datfile_name='input-files/kerinci/coarse-model/Keriv0_027',
incon_name='input-files/kerinci/coarse-model/Keriv0_027',
geom_name='input-files/kerinci/coarse-model/gKerinci_v0')
real_data_model.rename_wells_as_obs(list_of_obs_wells)
#---load real data of appropriate resolution and store in above.
real_data_model.d_obs_well = {}
real_data_model.ss_temps_obs_well = {}
for i,welli in enumerate(list_of_obs_wells):
df = pd.read_csv('./saved_data/kerinci_data/Temp_' + welli + '.dat',header=None,sep=' ')
df.rename(columns={1:'d',0:'T'},inplace=True)
real_data_model.d_obs_well[i] = df['d']
real_data_model.ss_temps_obs_well[i] = df['T']
#---create a basic comparison model (basis of likelihood function)
measurement_space = IC.MeasurementSpace(bias=0.0, sigma=10.0)
#---create a parameter model
parameter_space = IC.ParameterSpace(mu=-15, sigma=1.5)
#-----create a basic process space model
process_space = IC.ProcessSpace()
#-----set up discrepancy info
load_discrepancy = True
map_coarse_discrep_to_data_grid = False # - default should be False?
#discrepancy_filename = 'discrepancies_combined_kerinci.p'
discrepancy_filename = 'discrepancies_combined_kerinci_map_data.p'
if load_discrepancy:
discrep = IC.ModelDiscrep(process_space=process_space, measurement_space=measurement_space,
coarse_process_model=process_model_coarse, data_model=real_data_model)
#else:
# discrep = IC.ModelDiscrep(process_space=process_space, measurement_space=measurement_space)
discrep_data = pickle.load(
open("./saved_data/" + discrepancy_filename, "rb"))
discrep.discrepancy_dist = discrep_data
discrep.compute_overall_discrep_model(
map_coarse_discrep_to_data_grid=map_coarse_discrep_to_data_grid)
#update process space based on raw discrep.
process_space.discrepancies = discrep.discrepancy_dist
#compute relevant data space quantities.
measurement_space.bias = discrep.discrepancy_mean_data
measurement_space.icov = discrep.combined_icov_data
measurement_space.map_process_to_data_scale = map_coarse_discrep_to_data_grid
#---create a Bayes model
#use pro_model_coarse for coarse, pro_model_medium for medium
#bmodel = IC.BayesModel(name='test_bayes_model',
# process_model=process_model_medium,
# data_model=synthetic_model_fine,
# comparison_model=comparison_model,
# parameter_model=parameter_model)
model_name = 'test_kerinci_bayes_model'
if load_discrepancy:
model_name = model_name + '_discrep'
bmodel_coarse = IC.BayesModel(name=model_name,
process_model=process_model_coarse,
data_model=real_data_model,
measurement_space=measurement_space,
parameter_space=parameter_space,
process_space=process_space,
fine_process_model=process_model_fine)
#---test calculate likelihood
start = timeit.default_timer()
ll = bmodel_coarse.lnlike(perm_powers_current=perm_powers_precal)
stop = timeit.default_timer()
print('time to compute lnlike (s): ')
print(stop - start)
print('ll:')
print(ll)
#---test calculate model run time
start = timeit.default_timer()
bmodel_coarse.process_model.simulate()
stop = timeit.default_timer()
print('time to run model (s): ')
print(stop - start)
#---test find best fit
f = lambda x: -bmodel_coarse.lnprob(perm_powers_current=x)
#guess = np.tile(-15.0, len(perm_powers_precal))
#guess = perm_powers_precal #if trying to improve on John's...
#use opt from last time
save_path = './saved_data/'
guess = pickle.load(open(save_path + 'test_kerinci_bayes_model_best_fit_solution.p', "rb"))
maxfevs = 10000#1000
fatol = 1e-3
start = timeit.default_timer()
res = minimize(fun=f, x0=guess, method='Nelder-Mead', tol=1e-3,
options={'disp': True, 'maxfev': maxfevs, 'fatol':fatol})
#res = minimize(fun=f, x0=guess, method='L-BFGS-B', tol=1e-3,
# options={'disp': True, 'maxfev': maxfevs})
stop = timeit.default_timer()
print('time to find best fit (s): ')
print(stop - start)
best_fit_solution = np.zeros((len(res.x)+1))
best_fit_solution[:-1] = res.x
best_fit_solution[-1] = res.fun
#---save data
save_data = True
if save_data:
#save best fit data
save_path = './saved_data/'
try:
os.makedirs(save_path)
except OSError as exc: # Python >2.5
if exc.errno == errno.EEXIST and os.path.isdir(save_path):
pass
else:
raise
pickle.dump(best_fit_solution[:-1], open(save_path + bmodel_coarse.name + "_best_fit_solution.p", "wb"))
#---do predictive check
load_data = True
if load_data:
best_params = pickle.load(open(save_path + bmodel_coarse.name + "_best_fit_solution.p", "rb"))
else:
best_params = best_fit_solution[:-1]
param_sets = np.array((best_params,))
bmodel_coarse.predictive_checks(parameter_sets=param_sets,subtract_bias=False,do_data_space=True)
bmodel_coarse.predictive_checks(parameter_sets=param_sets,subtract_bias=True,do_data_space=True)
bmodel_coarse.predictive_checks(parameter_sets=param_sets,subtract_bias=False,do_data_space=False)
bmodel_coarse.predictive_checks(parameter_sets=param_sets,subtract_bias=True,do_data_space=False)
#---test emcee
#p_start = best_params
#bmodel.run_emcee(p_start=p_start,
# n_walkers=30, n_burn=10, n_sample=30, save_data=True, run_name='_test_short')
#test find best fit with custom derivative
#f = lambda x: -bmodel.lnprob(perm_powers_current=x)
#df = lambda x: -bmodel.lnprob_grad(perm_powers_current=x)
#
#guess = np.tile(-15.0, 12)
#maxfevs = 200
#fatol = 0.01
#start = timeit.default_timer()
#res = minimize(fun=f, jac=df, x0=guess, method='BFGS', tol=1e-1,
# options={'disp': True, 'maxfev': maxfevs, 'fatol':fatol})
#
#stop = timeit.default_timer()
#print('time to find best fit with gradient (s): ')
#print(stop - start)
#
#best_fit_solution_wderiv = np.zeros((len(res.x)+1))
#best_fit_solution_wderiv[:-1] = res.x
#best_fit_solution_wderiv[-1] = res.fun