frflib.forecast_methods.forecast_wf

Forecast Workflow module

Module containing the workflow object and related methods.

Workflow are the core orchestrating object of FRFLib, organizing the data distribution and sequentiation of work to all chosen Forecasters

Classes

ForecastWF

ForecastWF

Functions

get_f_methods_conditions(method_name)

function that return the list of fluid forecasted by the method

valid_forecaster_params(forecast_params[, verbose])

check that a list of conditions is validated by a given parameters set

naming_phaser(df)

reformulate dataset from asset model to phaser format

compute_init_manual(→ ForecastWF)

Run a manual parameter initializer workflow

get_new_wells_list(df_dynamic, bt_date)

Return the list of the wells which started after the bt_date. for the blind test they are assumed as new wells

extract_wf_parameters(→ pandas.DataFrame)

For a given fluid, return the parameters of the forecast : score, method, parameters, method_category

Module Contents

class frflib.forecast_methods.forecast_wf.ForecastWF(field_data, wf_param, forecast_list, wells_for=None)

ForecastWF

Object in charge of instantiating, running and saving all forecast methods in a workflow.

Parameters:

  • field_data (InputData) – Field data

  • wf_param (dict) – workflow parameters, dictionary containing all forecast parameters, including forecast mode

  • forecast_list (List[dict]) – list of dictionary with forecaster to be called in the workflow by order with their parameters

  • If running in ‘blind_test’ mode, it will realize a forecast on the last ‘blind_test_duration’ month of data.

  • If running in ‘forecast’ mode, it will forecast for ‘forecast_period’ months

the forecast modes and parameters are defined in the wf_param dictionary argument

Workflow parameters:

  • name (str): name of the forecast

  • forecast_fluids (List[str]): list of fluids to be forecasted (e.g. [‘oil’, ‘water’])

  • forecast_period (int) : number of month to forecast from data last date

  • forecast_mode (str): mode of forecast, ‘blind_test’ or ‘forecast’

  • blind_test_duraion (int): if in forecast_mode, number of month to remove from data

  • eco_cut_off (float): Value under which to close a well

  • max_bsw (float): ratio between water and liquid under which to close a well

  • max_liq_rate_stbd: maximum liquid constraint

  • max_gas_rate_mscfd: maximum gas constraint

  • short_term_forecast_adjustment: number of month to join data to trend

  • uptime_selection: uptime hypothesis (‘constant_value’ or ‘average_uptime’)

  • constant_value: if uptime is ‘constant_value’, value to consider for uptime

  • period_for_avg_months : if uptime is average_uptime, number of previous month to consider in average

  • abandonned_wells (int): number of month without production to consider a well closed

Example of Workflow execution ::
>>> general_forecast_params = {
>>>  'forecast_name': 'wf_exponential',
>>>  'forecast_end_date': "01/01/2040",
>>>  'forecast_fluids': ['oil', 'water'],
>>>  'forecast_mode': 'blind_test',
>>>  'abandonned_wells': 12,
>>>  'uptime_selection': 'constant_value',
>>>  'constant_value': 1.0,
>>>  'short_term_forecast_adjustment': 24,
>>>  'blind_test_duration': 48,
>>>  'max_liq_from_history': True
>>> }
>>> param_list = [
>>>  {'forecaster_name': 'dca_oil',
>>>   'methods': 'exponential',
>>>   'fluid': 'oil'},
>>>  {
>>>   'type': 'analog_wells_oil',
>>>   'fluid': 'oil',
>>>   'learn_x': 'cum_active_days',
>>>  },
>>>  {'type':'log_fo_fit',
>>>   'fluid': 'water',
>>>  }]
>>> # Execution
>>> ForecastWF.execute(data, general_forecast_params, param_list, nproc=12)
classmethod get_forecaster_form_dict(*args) → collections.OrderedDict
Abstractmethod:

Returns the class required parameters description for generating a form

Returns:

method parameters dictionary

Return type:

OrderedDict

classmethod get_forecaster_form(*args) → frflib.forecast_tools.forecaster_param.ForecasterParam
Abstractmethod:

Returns the form for class required parameters setting

Returns:

Forecaster Parameter object

Return type:

ForecasterParam

classmethod execute(field_data, wf_params, forecast_list, nproc=1, wells_for=None, taskobj=None, df_scenario=None)

Creates a workflow and run it.

It is the main designated interface to create and run new workflows

Parameters:

  • wells_for –

  • field_data (InputData) – Field data

  • wf_params (dict) – workflow parameters

  • forecast_list (List[dict]) – list of dictionaries containing forecaster names and parameters to execute

  • nproc (int) – number of process available (default 1)

  • df_scenario – scenario - only for new wells

  • taskobj – for frf web app

update_manual(manual_param)

Update a workflow with manual parameters

Parameters:

manual_param – manual parameters for wells to update

Type:

dict

enrich_data()

function that recompute enriched data (data + forecast as a dataframe) This is function is usefull as the enriched data is not stored in the hdf wf file

Returns:

post_process()

post process the workflow so far mainly - complete with wells no producing / not forecasted / new wells for blind test

Returns:

get_well_forecast(wellname: str)

Returns the forecast of a well

Parameters:

wellname –

Returns:

DataFrame with all forecasted data for this well

Return type:

pd.DataFrame

get_well_forecast_param(wellname: str, fluid: str) → dict

Returns the parameters of a forecaster

Parameters:

  • wellname –

  • fluid –

Returns:

parameters of the forecaster for given well and fluid

get_forecast_plot_data_json() → dict

generate json for FRF front to plot forecast data

Returns:

forecast json

Return type:

dict

get_param_value(fluid: str, param_list=None, **kwargs) → pandas.DataFrame

extract values of a list of parameters for all possible well

Parameters:

  • fluid (str) – fluid to look parameter for

  • param_list (list) – list of parameters to extract

Returns:

Dataframe of parameters

Return type:

pd.DataFrame

get_well_fit(fluid: str, wellname: str) → pandas.Series

returns the fit of a well and fluid

Parameters:

  • fluid – fluid of the fit

  • wellname – name of the well

Returns:

Fit series indexed by date

Return type:

pd.Series

get_all_fit()
Returns:

a dataframe with all fits

Return type:

pd.DataFrame

get_all_forecast(well_list=None, method=None)

obtain all the valid forecast for the field or a well list

Parameters:

  • well_list – list of wells to fetch info from

  • method (str) – if not None, give only forecast of this method

Returns:

Dataframe with all forecasts

Return type:

pd.DataFrame

get_phaser_format(last_date_forecast=None)

Extract forecast and format them in a phaser-format-compatible dataframe

Parameters:

last_date_forecast (datetime) – last date of the forecast to extract

get_sum_forecast(well_list=None, df_forecast=None)

give the sum of the forecast for a well list or a field

Parameters:

  • well_list – list of wells

  • df_forecast – if forecast has been extracted it can be given to avoid re-extraction of the data

Returns:

aggregated dataframe

Return type:

pd.DataFrame

get_well_num_per_fmethod(fluid, well_list=None)

function that allows to get the number of well per method forecast

Parameters:

  • well_list – list of wells

  • fluid – considerd fluid

Returns:

dataframe with methods vs number of wells

get_closed_wells()

simple function which return the closed wells in the worflow

Returns:

list - closed wells

blind_test_results(fluid: str = 'oil', compute_r2=True, only_forecasted=False)

generate results of blind tests - store internally in self.bt_results and return a dataframe

Parameters:

fluid – fluid to consider

Returns:

dataframe with well by well comparison for the blind test

Return type:

pd.DataFrame

save_to_hdf(path: str)

Save workflow in a hdf file

Parameters:

path (str) – path of the file

classmethod load_from_hdf(path: str, field: frflib.data_class.input_data.InputData | str)

Load data from a HDF and instantiate a new workflow

Parameters:

  • path (str) – path to file

  • field (InputData or string) – field dataset or path to data file

Returns:

Loaded workflow

Return type:

ForecastWF

plot_well_num_per_fmethod(fluid, well_list=None)

function that plots the number of well per method forecast

Parameters:

  • well_list – list of wells

  • fluid – considerd fluid

Returns:

pie chart

compare_blind_test(other, fluid='oil', well_list=None)

make a comparison scatter plot to compare forecasted values with another forecast

bt_get_worst_off(fluid, well_number, error='absolute', worst=True)

get the worst performing well during a blind test

Parameters:

  • fluid –

  • well_number –

  • error –

  • worst –

Returns:

export_to_df(export_type='frf_format', aggregation_cat='wellname')

extract results, transform them in export format and aggregate them

Parameters:

  • export_type – ‘frf_format’ or ‘phaser_format’

  • aggregation_cat – category in df_static to aggregate against

Returns:

dataframe with results in asked format

detail_export()

Export details about the parameters of each forecaster

Returns:

Dictionary of dataframes, first one is “summary”, the rest is one dataframe for each forecaster

cut_like_wf(another: ForecastWF) → None

Cut forecasts to keep only fluids and end_date such as defined in another workflow Works only for forecasts in forecast mode

Parameters:

another – Model workflow to mimic

cut_forecasts_to_end()

cut forecasts to match required end date

get_result_summary(fluid=None, include_stat=False, results_only=False)

function that gives a dataframe summary of the results of the workflow main data are the parameters and the ultimate cumulative per fluid

Parameters:

  • fluid – oil water gas liquid

  • include_stat – bool - if true will concat the df with static data and production ind

Results_only:

if false give the parameters of the forecast

Returns:

dataframe

__repr__()

Return repr(self).

classmethod get_f_methods_fluid(method_name)

function that return the list of fluid forecasted by the method

classmethod get_f_methods_default_param(method, field_data=None)

return default parameter of the method

Parameters:

method – method name

Returns:

dictionary of default parameters

classmethod get_available_f_methods(fluid_list, available_ux=True)

function that given the list of the fluid selected will give the available forecast methods in frflib

:param available_ux if true show only list of method validated for UX :return: dictionary with method_name : method name : method :rtype: dict

classmethod get_f_methods(method_name)

function that return the object forecast method with the given name

Returns:

Forecaster object

Return type:

AbstractForecast

classmethod get_f_methods_form_dict(method_name, field_data=None)

function that return the dictionary necessary to build the form for the given forecast name

See also

forecast_methods.abstract_forecast.AbstractForecast.get_forecaster_form_dict()

classmethod get_f_methods_form(method_name, field_data=None)

function that return the dictionary necessary to build the form for the given forecast name

setup_blindtest(bt_duration=None, new_wells_split=1)

allow to setup the main parameters for the blnd test (the input_data,the bt_data for comparison purposes the new wells list and the bt_start up It allows to run for both existing and new wells

:param data : (input_data) the dataset for the workflow :param bt_duration: (int) duration of the bt in months :param existing_tag: :param new_well_split: (float) - between 0 and 1 allows a subset selection for the new wells

Returns:

enriched_data: (input_data) - subset of field data bt_data: (input_data) - subset of field data dt_bt_start: (date) blind test date start new_wells: (set) new wells set

get_fluid_dict()

function that return a dictionary of fluids and variables available in the forecast

Returns:

frflib.forecast_methods.forecast_wf.get_f_methods_conditions(method_name)

function that return the list of fluid forecasted by the method

frflib.forecast_methods.forecast_wf.valid_forecaster_params(forecast_params, verbose=False)

check that a list of conditions is validated by a given parameters set

frflib.forecast_methods.forecast_wf.naming_phaser(df)

reformulate dataset from asset model to phaser format

frflib.forecast_methods.forecast_wf.compute_init_manual(data: frflib.data_class.input_data.InputData, nproc=1, duration=LONG_DURATION, forecast_parameters=None) → ForecastWF

Run a manual parameter initializer workflow

Parameters:

  • data – Field data to compute initialization for

  • nproc (int) – number of processes to use (default 1)

  • duration (int) – number of months to forecast if no forecast parameters specified (default to 480)

  • forecast_parameters (dict) – forecast parameters to use (default is None).

Returns:

workflow executed with ManualInitializer

Return type:

ForecastWF

frflib.forecast_methods.forecast_wf.get_new_wells_list(df_dynamic, bt_date)

Return the list of the wells which started after the bt_date. for the blind test they are assumed as new wells and therefore not treated

Parameters:

  • df_dynamic – dynamic dataframe

  • bt_date – pd.datetime

Returns:

set of new wells

frflib.forecast_methods.forecast_wf.extract_wf_parameters(dict_wf_param: dict, dict_wlist_fmethod: dict, fluid: str, fluid_list: List[str], main_fluid: str) → pandas.DataFrame

For a given fluid, return the parameters of the forecast : score, method, parameters, method_category Wells not forecasted are return with score=0 and empty parameters If fluid is not forecasted return the parameters of main fluid (ex : liquid)