AISPRY · GMR POWER TRADING
Featured Innovation

AI forecasts for India's electricity market.

AiSPRY built GMR Power Trading for GMR Energy — a per-block ensemble that forecasts next-day MCP across all 96 fifteen-minute blocks of the IEX Day-Ahead Market.

Explore the Platform Read the Case Study
DAM AUCTION · LIVE 96 · 15-MIN BLOCKS
₹4,287/MWh ▲ 2.4% FCST · TOMORROW · DAILY AVG
B61 · ₹6,940
₹7K ₹5K ₹3K ₹1K
B1 · 00:00 ───── B96 · 24:00
Night
₹3.1K
M.Ramp
₹4.6K
Daytime
₹3.8K
E.Ramp
₹6.9K
Close
₹4.2K
96
Blocks Per Day
15-min Granularity
< 10%
MAPE Target
Per-Block Accuracy
87
Engineered Features
Across 10 Datasets
18 ms
Inference Latency
FastAPI
AISPRY · GMR POWER TRADING
Data Foundation · 10 Datasets

770K rows. One price.

Ten datasets power 87 engineered features to forecast tomorrow's MCP — block by block, across all 96 fifteen-minute windows.

See the Data Read the Case Study
⚡
Market Trading
IEX · DAM + RTM
191K rows
🔋
Grid & Supply
NLDC · NPP/CEA
232K rows
🌡️
Weather
Open-Meteo · 247 cities
~237 MB
🌊
Hydrology
NPP · CWC reservoirs
52K rows
⛏️
Coal & Fuel
Coal India · plant-level
276K rows
📅
Calendar Events
IPL · festivals · polls
1K rows
Target Variable MCP · ₹/MWh · next-day · per block
10
Authoritative
Datasets
770K
Total Rows
Across All Sources
1,040+
Trading Days
Of Training Data
Apr '23→
Coverage
Continuously Refreshed
AISPRY · GMR POWER TRADING
The Platform · MLOps Stack

Ensemble forecasts. End-to-end MLOps.

V6 Ensemble (XGBoost + LightGBM + Lag-96), trained per block, served through production MLOps: Airflow, MLflow, Feast, FastAPI — plus lineage and drift monitoring.

Talk to AiSPRY Read the Case Study
01
Data Sources
IEX · NLDC · NPP · Open-Meteo · CWC · Coal India
10datasets
02
Storage & Validation
S3 / GCS · Great Expectations · ClickHouse
770Krows
03
Features (Feast)
Offline + Online stores · 4 ms retrieval
87features
04
V6 Ensemble (MLflow)
XGBoost + LightGBM + Lag-96 · per-block
96models
05
Serving
FastAPI · PostgreSQL · React Dashboard
18 mslatency
+
Cross-cutting
Airflow · Prometheus · Grafana · OpenLineage · Argo
5tools
V6 Ensemble
XGBoost + LightGBM
+ Lag-96
96
Per-Block
Models
4 ms
Feature Store
Retrieval
±1σ
Confidence
Bands Per Block