Which wells sustain field production, when and why did they lose performance (water, pressure, downtime), and how much oil was lost to downtime?
Status: data cleaning, SQL KPIs, production analysis and Power BI dashboard done. Decline-curve forecast in progress.
| # | Finding | Number |
|---|---|---|
| 1 | Two wells carried the field | F-12 + F-14 = 85% of the oil |
| 2 | Water arrived early and took over | Water cut passed 10% 13–23 months after first oil in the original wells; field water cut above 70% from 2011 |
| 3 | Downtime was the largest controllable loss |
Details: notebooks/02_analysis.ipynb · Power BI report (PDF) · water cut by well · downtime loss by year
A production data analyst at an operator has to answer three questions every day:
- Contribution: which wells drive production?
- Well health: how are water cut, GOR and pressures evolving?
- Losses: how much production is lost to downtime?
This project answers those questions with real daily well data from the Equinor Volve field, as a Daily Production Report style analysis.
- Source: Equinor Volve open dataset, Volve production data.xlsx (daily production sheet), 2008–2016, North Sea.
- Licence: Equinor Open Data Licence. Educational and research use, with attribution to Equinor and the Volve licence partners.
- The raw file is not stored in this repo. Download it and place it in
data/raw/.
A four-page report built on a star schema (fact_production, fact_downtime, dim_well, dim_date), with Year and well slicers on every page. Full export: dashboard/volve_dpr.pdf.
| Page | What it shows |
|---|---|
| 1 · Field overview | KPI cards (oil, water cut, producer efficiency, lost oil, water injected), monthly oil by producer well (injectors F-4 and F-5 are excluded from the production chart) and oil share by well |
| 2 · Well health | Water cut by well with the 10% breakthrough line, GOR over time, and a per-well table with first oil, volumes, water cut and GOR |
| 3 · Downtime & losses | Lost oil by year split into full-day stops vs partial days, lost oil by well, an efficiency matrix (well × year) and Lost vs Potential % |
| 4 · Daily report | Pick a day on the date slider and see each well's on-stream hours, volumes, water cut, choke, wellhead pressure and lost oil, like an operator's morning report. Wells with fewer than 24 on-stream hours are flagged |
Model, DAX measures and theme live in dashboard/ (see its README). The input tables are regenerated with python src/export_powerbi.py.
Cleaning (Python) → KPIs (SQL, DuckDB) → Analysis and decline curves (Python) → Dashboard (Power BI)
KPIs (implemented in sql/kpis.sql and as DAX measures in dashboard/measures.dax):
| KPI | Definition |
|---|---|
| Oil, gas and water volumes | Per well and field, daily and monthly |
| Water cut | water / (oil + water) |
| GOR | gas / oil |
| Producer efficiency | on-stream hours / calendar hours, producers only |
| Estimated downtime loss | lost hours × the well's median full-day rate in the same month (upper bound) |
| Lost vs potential | lost oil / (produced oil + lost oil) |
Python (pandas, scipy, matplotlib) · SQL (DuckDB) · Power BI
data/raw/ original data (not tracked: download from source)
data/processed/ cleaned data
notebooks/ 01_cleaning → 02_analysis → 03_decline
sql/ KPI queries
src/ reusable Python functions
dashboard/ Power BI file (.pbix), PDF export, DAX measures, theme
dashboard/data/ star-schema CSVs (not tracked: regenerate with src/export_powerbi.py)
images/ charts and dashboard screenshots
pip install -r requirements.txtThen download the data into data/raw/ and run the notebooks in order.
Aguinaldo Miguel — Automation & Control Engineer | Data Analytics for Oil & Gas Operations and Supply Chain · Luanda, Angola · LinkedIn
Code: MIT License. Data: © Equinor and the Volve licence partners, under the Equinor Open Data Licence.