Day-ahead demand forecasting & power-market intelligence, in one platform
EPM gives grid operators and traders AI-driven demand forecasts, digitized exchange bid curves, and a unified real-time view of IEX, HPX, and PXIL — replacing manual Excel guesswork with data-backed decisions worth crores.
Power procurement has run on guesswork for too long
State grid operators buy power a day ahead against estimated demand across 96 blocks. Get it wrong and the cost is immediate — either wasted crores on unused procurement, or grid instability from underestimation.
Manual, Excel-driven forecasting
Analysts eyeballed weather and historical spreadsheets to estimate demand — slow, inconsistent, and expensive when wrong.
Bid curves trapped as images
Exchanges publish DAM, GDAM, and RTM bid curves only as charts — traders can see MCP/MCV but can't quantify or simulate them.
Market data spread across portals
IEX, HPX, and PXIL each publish separately, making cross-exchange comparison a manual, time-consuming chore.
AI-powered day-ahead demand forecasts, block by block
The operator selects a date, zone, and model with a weather toggle, and the platform pulls four-plus years of demand history and weather from BigQuery, runs multiple ML and statistical time-series models simultaneously, and blends the top performers into a Triple Hybrid Ensemble — delivering a complete 96-block forecast in seconds.
- Inverse-MAPE ensemble: each model is scored on MAPE, then blended with bias correction for an optimized composite forecast.
- Weather-similarity engine: surfaces analogous historical days as reference benchmarks.
- SetLine tool: manual scaling, shifting, or smoothing across specific block ranges with configurable taper, for edge cases operators need to control.
- Four dashboards: Demand Forecast, Forecast Analysis, Model Performance, and Pattern Analytics — with KPI cards, confidence bands, forecast-vs-actual comparisons, MAPE/MAE/RMSE/R² leaderboards, and CSV export throughout.
Confidence bands, temperature overlay, and rostering band across all 96 blocks
Original bid curve vs a saved what-if scenario, recomputed live
Exchange bid-curve charts, turned into structured market data
Indian exchanges publish DAM, GDAM, and RTM bid curves only as images — traders see the Market Clearing Price (MCP) and Volume (MCV) but can't analyze them quantitatively. The Curve module fixes that with a Selenium, OpenCV, and EasyOCR pipeline that extracts the staircase geometry, calibrates it against the official MCP/MCV, and stores it in partitioned BigQuery tables.
- Instant curve loading: pick a market, date, and block to load real demand and supply curves immediately.
- What-if scenario engine: add or remove MW/price points and rebuild the cumulative demand and supply stacks in real time.
- Staircase-intersection logic: recomputes clearing price and volume, mirroring actual exchange bidding behaviour.
- Interactive Plotly charts: compare original vs scenario MCP/MCV, and save scenarios for later comparison and audit.
Every exchange, every market, one dashboard
IEX, HPX, and PXIL each publish DAM, GDAM, HPDAM, and RTM data on separate portals — daily markets once after close, RTM every 15 minutes. Independent scraping pipelines per exchange automatically extract, validate, and store all of it in a unified BigQuery history, eliminating manual collection entirely.
- Unified monitoring: toggle between exchanges and markets from one interface — no more visiting separate portals.
- Smart date picker: auto-detects the last 30 days with available data.
- Live current-block marker: loads any 15-minute block and refreshes automatically every 15 minutes.
- Cross-exchange comparison: view IEX, HPX, and PXIL side by side across all four market segments.
DAM · GDAM · HPDAM · RTM — refreshed every 15 minutes
From hours of manual effort to seconds of measurable accuracy
EPM replaces manual, Excel-based estimation with live ML-driven forecasting and quantitative market intelligence — reducing procurement error and giving operators a defensible, data-backed demand curve they can act on with confidence.
Seconds, not hours
A full 96-block forecast runs in seconds versus hours of analyst effort.
Lower procurement error
Reduces over- and under-buying exposure on power exchanges.
Quantified market intelligence
Bid curves become analyzable data — not just charts to eyeball.
One unified view
Every exchange, market, and forecast in a single interface.