Forecasters

Forecasters are an individual methods that encompass an algorithm. Each forecast workflow is composed of various forecasters that will workf together to generate a full forecast for the whole field. For example in a given field, not all the wells will be declining. Therefore, FRF will select some wells for decline curve analysis through some internal acceptance criteria and some others will be forecasted by some alternative methods as analog wells. Also, it allows to forecast all the different fluids sequentially. Various forecasters have been developed internally in order to address different situations and needs.

There are two main families of forecasters:

Fit Forecasters:

Fit forecasters are based the well data. we assume that for wells with enough data, that the wells is auto-sufficient to determine its own forecast. Therefore the forecast will be derived based on fit (DCA, FO vs NP, Liquid Constant, Segmented DCA,…). Basically any equation can be transformed into forecasters.

Here is a list of fit forecasters.

• Automatic DCA
• BSW Fit
• Constant Forecasters
• Linear Forecaster
• Arps Segmented Decline

In FRF lib fitters are developed in order to add an abstract layer and to allow basically and fit and any equation easily. Here is the link the API reference fitter. (#TODO add the link the frf library)

All the fit forecasters are used for existing wells

Learn Forecasters:

The learn forecasters are forecaster that will derive the well forecast based on similar wells byut with longer history and that were already treated by fit forecaster. This list will evolve continuously based on the different studies.

• Analog Wells
• Type Curve
• Group Segmented Arps

The learn forecasters are used for existing wells and for new wells. The versions for existing and new can differ in the parameters that are used