VFM (NF ML-based) Package

This package features scripts and workflows to estimate the production rates of naturally flowing wells using machine learning models.

Usage

Run the workflows from this package to perform the calculations. The schedules may be adapted to use different contexts.

Details

Training the Models

The workflows VFM (NF ML): Reset Training Rates and VFM (NF ML): Train MLs are used to train the machine learning models. The first workflow resets all rates used as input and / or targets, the second one trains the models.

The training puts out three models: one for predicting the liquid production rate, one for predicting the water cut and one for predicting the gas liquid ratio.

The models can be retrained whenever needed by running the schedule Training of workflow VFM (NF ML): Reset Training Rates.

Using the Models

The workflow VFM (NF ML): Predict Signals is used to predict the vfm liquid production rate, vfm oil production rate, vfm gas production rate and vfm water production rate.

First, the vfm liquid production rate is predicted using the liquid production rate model. This liquid production is then supplied to the other models to calculate the oil, gas, and water rates.

Data Requirements

The following data is required:

liquid production rate
oil production rate
water production rate
gas production rate
bottomhole pressure
bottomhole temperature
choke opening
gas liquid ratio
water cut
well head pressure
well head temperature