Forecasting for energy operations.
A shared data platform supporting demand forecasting, solar forecasting and reserve management, with evaluation built around operational decisions.
- Organisation
- An energy organisation
- Sector
- Energy
- Team
- Data and platform engineering, ML and MLOps, QA and backtesting, product
- Service
- Programmes

In short
- Operational data, connectedA governed route for data held behind a restricted operational network.
- A shared data and AI platformAn Azure Databricks platform with shared data and model infrastructure.
- Three forecasting productsLoad, solar PV and reserve management forecasts for operators.
- Built and testedAll three products completed user acceptance testing.
The challenge
An energy organisation’s operators needed forecasts of electricity demand, solar generation and reserve requirements to plan operations.
No platform existed to build and run those forecasts, and the operational data sat behind a restricted network boundary.
What we delivered
A governed data-capture path
A controlled route for operational readings and weather data across the network boundary.
A shared data and AI platform
An Azure Databricks platform with shared data and model infrastructure, including catalogue, versioning and monitoring.
Three forecasting products
Load forecasting, solar PV forecasting and reserve management, each designed around the decisions operators make.
The result
All three products were built, tested and completed user acceptance testing. Each model was backtested against an agreed baseline, using only the data available at the time of each forecast.
Operators keep the final decision on how each forecast is used.
How we worked
An embedded squad brought together data and platform engineering, ML and MLOps, QA and backtesting, and product.
The team agreed baselines and success measures with operators before building, and reviewed results against them at each stage.
