grid-load forecast

· 1 min read

grid-load predicts India’s electricity Demand for the next 48 hours. the whole pipeline is automated from data scraping to the forecast dashboard .

workflow

workflow diagram

source code is available at: https://github.com/jitendradara12/grid-load

row notes which i wrote during development…

  1. The best data yet is found on npp.
  2. The api is unprocted and dumps json directly on curl 'https://npp.gov.in/dashBoard/demandmet1chartdata?date=2026-06-08'
  3. The Endpoints served data all the way from sep25, my scripts/fetch_yesterday script runs everyday and updates the dataset with latest rows.
  4. ICED dataset is longer and scripts/clean_datasets.py cleans and src/data_pipeline.py merges them to get Combined time series spans from 2017-01-01 00:00:00 to present.
  5. Currently, using merged data for training and NPP dataset for inference.
  6. Most reasonable way for the model is ridge regression. weather/temperature features aren’t integrated yet.