Forecasts and schedules that keep small solar plants within deviation bands
TL;DR Bandkeeper forecasts output for 1–20 MW solar plants and files their schedules every day, so owners pay less in deviation charges. Open-access and captive plant owners pay a monthly fee per MW. It works because tolerance bands tightened in April 2026 and small plants have no in-house forecasting team.
Every grid-connected solar plant of size has to tell the grid operator in advance how much it will inject, block by block. When clouds make actual output miss the schedule, the plant pays deviation charges. From 1 April 2026 the central regulator cut the penalty-free band for solar and hybrid plants from ±10% to ±5%, and for wind from ±15% to ±10%[1]. The calculation also shifts year by year toward scheduled generation, reaching full weight from 2031[1].
States run their own versions. Madhya Pradesh, for one, ordered solar and wind generators to appoint a coordinating agency for scheduling[2]. Large developers have analysts for this, but a factory owner with a 5 MW captive plant usually does not.
What it does, in regulatory terms: a forecasting and scheduling service, acting as the plant's qualified coordinating agency (QCA) where the state's rules allow it. It does not trade power or hold a trading licence.
The seed proposed peer-to-peer trading to steady the grid. In India that has only run as a supervised pilot - a private Delhi distributor tested it with about 140 buildings until June 2021[3]. The legal lever today is better schedules, so that's what Bandkeeper sells.
Here's how it works. Bandkeeper pulls inverter data, combines it with satellite cloud imagery and weather models, and produces day-ahead and intraday forecasts for each plant. A small desk revises schedules within the windows the rules allow and files them. Owners get a monthly statement of deviation charges paid, against what they would have paid on a flat schedule. Where a state lets a QCA pool several plants, pooling reduces charges further.
Owners of 1–20 MW open-access, captive and group-captive solar plants pay a fixed fee per MW per month, with a bonus tied to charges avoided. Commercial and industrial developers who run many such plants for factory clients are the bigger buyer. I found no public price list for QCA services in India, so get quotes during validation.
The April 2026 tightening is the trigger[1]. The Green Energy Open Access Rules of June 2022 also cut the minimum for buying green power through open access from 1 MW to 100 kW[4], which has pushed more mid-size plants onto the grid. Every one of them now has a scheduling job that someone must do.
No public count exists of intra-state solar plants under scheduling rules, nor of what they pay for forecasting. The adjacent number is PM-KUSUM, a farm solar scheme, which had 13.11 GW installed by 31 March 2026[5]. That's part of a growing base of small grid-connected plants, though not all of them fall under deviation rules. If you want a real count then build it state by state from each load despatch centre's list of scheduled generators.
One Bengaluru energy-analytics firm sells sensing and AI tools for asset and grid management and has worked on scheduling for years[6]. A few other Indian scheduling firms also serve this field. Global forecasting vendors sell forecasts but do not file Indian schedules. The gap is the small-plant owner who wants one fee, a filed schedule and a monthly penalty report, without a sales cycle built for gigawatt developers.
| Difficulty | Moderate (46/100) |
| Starting capital | ₹60 L–1.2 Cr |
| Time to launch | 6–9 months |
Forecasting a single small plant is harder than a large portfolio, because one cloud covers the whole site. You also need a desk that files revisions on time every single day, and data links to inverters made by a dozen brands. Each state's rules differ, so expansion is slow.
Bandkeeper needs no trading licence, but in many states it must register as a qualified coordinating agency (QCA).
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
Don't let a model file schedules with no human watching, and don't promise savings you haven't measured. Weather surprises happen, and deviation charges hit the owner's money directly.
Want to work out the AI layer for Bandkeeper? Talk to Yash →
Here's what can go south -
I like this as a sound, narrow business for a founder with power-market and data-science experience who will learn one state's rules deeply. If you're hoping to build a trading platform then this isn't it. First prove, on five real plants over three months, that your schedules cut deviation charges by at least 30% against what each owner paid before.
Plan for roughly ₹60 L–1.2 Cr, mostly on data scientists, a small scheduling desk and weather data.
No trading licence, but many states require you to register as a qualified coordinating agency before filing schedules. Read your state commission's regulation first.
Yes, plants in many grids pay penalties for missed schedules. Check how your local grid operator sets forecasting rules and who may submit schedules.
The free Bandkeeper validation kit has the line-by-line capital breakdown, the effort scoring, the regulatory checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for plant owners and developers, and 5 assumptions to test. Download the kit →
Thinking about building Bandkeeper or something close to it? The make-or-break work is beating incumbents on accuracy for single small plants on cloudy days, and I'm happy to pressure-test your plan for that. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.