IndiaEnergy and utilities2023 – ongoingPTC India
Day-ahead and intraday electricity demand forecasting with Amazon Forecast
PTC India raised day-ahead demand forecast accuracy from 82% to 95% with Amazon Forecast
Problem
Power distribution companies in India pay large deviation penalties when actual demand differs from the amounts they schedule in 15-minute blocks. They needed more accurate demand forecasts.
Approach
Power Trading Corporation of India chose Amazon Forecast over custom models because its team was lean and lacked deep machine learning expertise. It built two predictors, one for day-ahead forecasts (96 blocks) and one for intraday forecasts about two hours ahead, using demand, weather and holiday data. Forecasts run on a schedule or on demand, and dashboards compare them with actual demand for stakeholders.
Results
Takeaway
A managed forecasting service let a small team skip building models. The accuracy figures come from AWS's own write-up, which also lists them as targets.
Categories
- Industry
- Energy and utilities
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor case study
Sources
Source summaries
AWS describes how Power Trading Corporation of India built day-ahead and intraday demand predictors on Amazon Forecast. It reports intraday accuracy of 98% against 85% before, and day-ahead accuracy of 95% against 82% before. It also describes the scheduled pipeline and the dashboards.
Full details are in the original sources.