void sched_history_fprintf( const sched_history_type * sched_history , const stringlist_type * key_list , FILE * stream) {
int step = 1;
time_t start_time = time_t_vector_iget( sched_history->time , 0);
int total_length = bool_vector_size( sched_history->historical );
while (true) {
if (bool_vector_safe_iget( sched_history->historical , step)) {
{
int mday,month,year;
time_t t = time_t_vector_iget( sched_history->time , step );
double days = (t - start_time) * 1.0 / 86400;
util_set_date_values_utc( t , &mday , &month , &year);
//fprintf(stream , "%02d-%02d-%4d " , mday , month , year );
fprintf(stream , " %5.0f " , days);
}
for (int ikey =0; ikey < stringlist_get_size( key_list ); ikey++)
fprintf(stream , "%16.3f " , sched_history_iget( sched_history , stringlist_iget( key_list , ikey) , step));
fprintf( stream, "\n");
} else
break; // We have completed the historical period - and switched to prediction
step++;
if (step == total_length)
break;
}
}
void enkf_main_initialize_from_scratch_with_bool_vector(enkf_main_type * enkf_main , const stringlist_type * param_list ,const bool_vector_type * iens_mask , init_mode_type init_mode) {
int num_cpu = 4;
int ens_size = enkf_main_get_ensemble_size( enkf_main );
thread_pool_type * tp = thread_pool_alloc( num_cpu , true );
arg_pack_type ** arg_list = util_calloc( ens_size , sizeof * arg_list );
int i;
int iens;
for (iens = 0; iens < ens_size; iens++) {
arg_list[iens] = arg_pack_alloc();
if (bool_vector_safe_iget(iens_mask , iens)) {
arg_pack_append_ptr( arg_list[iens] , enkf_main );
arg_pack_append_const_ptr( arg_list[iens] , param_list );
arg_pack_append_int( arg_list[iens] , iens );
arg_pack_append_int( arg_list[iens] , init_mode );
thread_pool_add_job( tp , enkf_main_initialize_from_scratch_mt , arg_list[iens]);
}
}
thread_pool_join( tp );
for (i = 0; i < ens_size; i++) {
arg_pack_free( arg_list[i] );
}
free( arg_list );
thread_pool_free( tp );
}
double sched_history_iget( const sched_history_type * sched_history , const char * key , int report_step) {
void * index = hash_get( sched_history->index , key );
if (!bool_vector_safe_iget( sched_history->historical , report_step ))
fprintf(stderr,"** Warning - report step:%d is in the prediction phase - can NOT ask for historical data! \n",report_step);
if (well_index_is_instance( index ))
return well_history_iget( index , report_step );
else if (group_index_is_instance( index ))
return group_history_iget( index , report_step );
else {
util_abort("%s: can not determine internal type of:%s - fatal internal error\n", __func__ , key);
return 0;
}
}
/* Query function: */
bool enkf_config_node_internalize(const enkf_config_node_type * node, int report_step) {
if (node->internalize == NULL)
return false;
else
return bool_vector_safe_iget( node->internalize , report_step); /* Will return default value if report_step is beyond size. */
}
bool obs_vector_load_from_HISTORY_OBSERVATION(obs_vector_type * obs_vector ,
const conf_instance_type * conf_instance ,
const history_type * history ,
ensemble_config_type * ensemble_config,
double std_cutoff ) {
if(!conf_instance_is_of_class(conf_instance, "HISTORY_OBSERVATION"))
util_abort("%s: internal error. expected \"HISTORY_OBSERVATION\" instance, got \"%s\".\n",__func__, conf_instance_get_class_name_ref(conf_instance) );
{
bool initOK = false;
int size , restart_nr;
double_vector_type * value = double_vector_alloc(0,0);
double_vector_type * std = double_vector_alloc(0,0);
bool_vector_type * valid = bool_vector_alloc(0 , false);
/* The auto_corrf parameters can not be "segmentized" */
double auto_corrf_param = -1;
const char * auto_corrf_name = NULL;
double error = conf_instance_get_item_value_double(conf_instance, "ERROR" );
double error_min = conf_instance_get_item_value_double(conf_instance, "ERROR_MIN" );
const char * error_mode = conf_instance_get_item_value_ref( conf_instance, "ERROR_MODE");
const char * sum_key = conf_instance_get_name_ref( conf_instance );
if(conf_instance_has_item(conf_instance, "AUTO_CORRF")) {
auto_corrf_name = conf_instance_get_item_value_ref( conf_instance , "AUTO_CORRF");
auto_corrf_param = conf_instance_get_item_value_double(conf_instance, "AUTO_CORRF_PARAM");
if(conf_instance_has_item(conf_instance, "AUTO_CORRF_PARAM"))
auto_corrf_param = conf_instance_get_item_value_double(conf_instance, "AUTO_CORRF_PARAM");
else
util_abort("%s: When specifying AUTO_CORRF you must also give a vlaue for AUTO_CORRF_PARAM",__func__);
}
// Get time series data from history object and allocate
size = history_get_last_restart(history);
if (history_init_ts( history , sum_key , value , valid )) {
// Create the standard deviation vector
if(strcmp(error_mode, "ABS") == 0) {
for( restart_nr = 0; restart_nr < size; restart_nr++)
double_vector_iset( std , restart_nr , error );
} else if(strcmp(error_mode, "REL") == 0) {
for( restart_nr = 0; restart_nr < size; restart_nr++)
double_vector_iset( std , restart_nr , error * abs( double_vector_iget( value , restart_nr )));
} else if(strcmp(error_mode, "RELMIN") == 0) {
for(restart_nr = 0; restart_nr < size; restart_nr++) {
double tmp_std = util_double_max( error_min , error * abs( double_vector_iget( value , restart_nr )));
double_vector_iset( std , restart_nr , tmp_std);
}
} else
util_abort("%s: Internal error. Unknown error mode \"%s\"\n", __func__, error_mode);
// Handle SEGMENTs which can be used to customize the observation error. */
{
stringlist_type * segment_keys = conf_instance_alloc_list_of_sub_instances_of_class_by_name(conf_instance, "SEGMENT");
stringlist_sort( segment_keys , NULL );
int num_segments = stringlist_get_size(segment_keys);
for(int segment_nr = 0; segment_nr < num_segments; segment_nr++)
{
const char * segment_name = stringlist_iget(segment_keys, segment_nr);
const conf_instance_type * segment_conf = conf_instance_get_sub_instance_ref(conf_instance, segment_name);
int start = conf_instance_get_item_value_int( segment_conf, "START" );
int stop = conf_instance_get_item_value_int( segment_conf, "STOP" );
double error_segment = conf_instance_get_item_value_double(segment_conf, "ERROR" );
double error_min_segment = conf_instance_get_item_value_double(segment_conf, "ERROR_MIN" );
const char * error_mode_segment = conf_instance_get_item_value_ref( segment_conf, "ERROR_MODE");
if(start < 0)
{
printf("%s: WARNING - Segment out of bounds. Truncating start of segment to 0.\n", __func__);
start = 0;
}
if(stop >= size)
{
printf("%s: WARNING - Segment out of bounds. Truncating end of segment to %d.\n", __func__, size - 1);
stop = size -1;
}
if(start > stop)
{
printf("%s: WARNING - Segment start after stop. Truncating end of segment to %d.\n", __func__, start );
stop = start;
}
// Create the standard deviation vector
if(strcmp(error_mode_segment, "ABS") == 0) {
for( restart_nr = start; restart_nr <= stop; restart_nr++)
double_vector_iset( std , restart_nr , error_segment) ;
} else if(strcmp(error_mode_segment, "REL") == 0) {
for( restart_nr = start; restart_nr <= stop; restart_nr++)
double_vector_iset( std , restart_nr , error_segment * abs(double_vector_iget( value , restart_nr)));
} else if(strcmp(error_mode_segment, "RELMIN") == 0) {
for(restart_nr = start; restart_nr <= stop ; restart_nr++) {
double tmp_std = util_double_max( error_min_segment , error_segment * abs( double_vector_iget( value , restart_nr )));
double_vector_iset( std , restart_nr , tmp_std);
}
} else
util_abort("%s: Internal error. Unknown error mode \"%s\"\n", __func__, error_mode);
}
stringlist_free(segment_keys);
}
/*
This is where the summary observations are finally added.
*/
for (restart_nr = 0; restart_nr < size; restart_nr++) {
if (bool_vector_safe_iget( valid , restart_nr)) {
if (double_vector_iget( std , restart_nr) > std_cutoff) {
obs_vector_add_summary_obs( obs_vector , restart_nr , sum_key , sum_key ,
double_vector_iget( value ,restart_nr) , double_vector_iget( std , restart_nr ) ,
auto_corrf_name , auto_corrf_param);
} else
fprintf(stderr,"** Warning: to small observation error in observation %s:%d - ignored. \n", sum_key , restart_nr);
}
}
initOK = true;
}
double_vector_free(std);
double_vector_free(value);
bool_vector_free(valid);
return initOK;
}
}