Skip to main content

Predicting Oil Production

Business Need​

Forecast mid term oil and gas production from wells and fit well-known oil-industry curves to extrapolate predictions 30+ years into the future.

Solution Overview​

This solution leverages multiple components of the Abacus platform to deliver automated production forecasting:

Step 1: Connect & Preprocess Data

  • Data Source: Production data is ingested from a Snowflake table using our native connector
  • Data Structure: The raw data contains daily production metrics at the well level
  • Preprocessing: Daily production values are aggregated to monthly intervals, which provides the optimal balance between reducing noise and maintaining actionable granularity for operational decisions

Step 2: Build Forecasting Model

A time series forecasting model is developed using the Forecasting and Planning solution, which automatically handles seasonality, trends, and well-specific production patterns to generate monthly production forecasts. It predicts 12 months into the future.

💡Pro Tip

During tests, we found using a multi-output regression approach in combination with time series forecasting yielded a better outcome. We used the predictive modeling module to create an ensemble model.

Step 3: Curve Fitting

The curve fitting part was written purely in python and the code was hosted within a pipeline. The functions are then used to extrapolate predictions into the future.

How It's Used in Practice​

This solution operates as a fully automated workflow with minimal user intervention:

Automated Production Pipeline (runs monthly):

  • Refreshes production data from Snowflake, executes forecasting models
  • Executes forecasting models for all active wells
  • Retrieves predictions from the feature group storage
  • Applies curve fitting to forecasted values
  • Writes results back to Snowflake for consumption by analytics tools

Model Retraining Pipeline (runs quarterly):

  • Automatically retrains forecasting models every 3 months to incorporate recent production trends and maintain prediction accuracy

End users simply consume the forecasts and curve-fitting parameters from their existing BI tools and dashboards without needing to interact with the Abacus platform directly.

✨Key Outcomes

Key achievements:

📉 Reduced forecast variance by 15%​

Improved prediction accuracy and decision-making confidence

⚡ Fully automated forecasting​

Eliminated manual processes

Additional Information​