WIPA Water Overview

Project Objective

• Understand what influences cumulative water production and build predictive models for water volumes at fixed intervals (1, 3, 6, and 12 months) using geological, reservoir, and well design features.
• Create water forecasts for budgeted wells to optimize water handling operations and evaluate the need for new infrastructure.
• Improve the accuracy of water production forecasts across ARC assets.
• Ensure water forecasts are available for all budgeted wells to support future water optimization initiatives.
• Support budgeting and planning by delivering water forecasts (Water Gas Ratio, WGR) for new wells, loading them into Mosaic, and enabling tracking of water-related costs and well economics.

For additional project background and detailed documentation, please refer to the WIPA Water Forecast Project Charter.

Model Performance Assessment

In order to evaluate model performance against existing water forecasts:

• Model performance was assessed by comparing results to budget forecasts entered in Mosaic for 2023 and 2024, specifically for the 3- and 6-month water forecasts. The results indicate a clear improvement over previous methods.
• The machine learning model demonstrates superior forecasting accuracy for water measurement in Dawson, Kakwa, and Parkland. In Attachie, performance is comparable to the baseline forecast at the time of writing this report; however, future model retraining could improve results for this field as more data becomes available.

Below are the Median Absolute Percentage Error (MDAPE) results for the 3-month forecast:

And here are the MDAPE results for the 6-month forecast:

Leveraging WIPA Data and Architecture

The water forecast model is built on the WIPA model infrastructure and utilizes the same feature and data engineering tables. For more information about WIPA, please refer to the WIPA documentation.

Location

Model Feature Tables:

• prd_zone3.wipa.geology_reservoir_features
• prd_zone3.wipa.well_design_features
• prd_zone3.wipa.well_detail_features
• prd_zone3.wipa.volume_features

Model Results Tables:

• prd_zone3.wipa.water_regression_mosaic_inference (inference results for Mosaic budget wells)
• prd_zone3.wipa.water_regression_scenario_prediction (model predictions on historical datasets used for training)
• prd_zone3.wipa.water_regression_metrics (model performance metrics from historical training)
• prd_zone3.wipa.water_regression_shap_values (SHAP values demonstrating feature importance from historical training)

Repository:

• ARCRes/ARC Data Analytics Team/Volume Analytics

Databricks Workspaces:

• PROD: arcPrdDbwDAL
• UAT: arcUatDbwDAL
• DEV: arcDevDbwDAL

Data Sources

• prd_zone3.wellanalytics.dim__well
• prd_zone3.geoanalytics.dim__layer_hierarchy
• prd_zone3.geoanalytics.canonical__regional_attributes
• prd_zone3.wellanalytics.canonical__well_spacing_depletion_features
• prd_zone3.wellanalytics.fact__productionvolumesdayson

Contacts

See the contacts page for project leads and support.