From Forecast to Dispatch: How Indian Renewable Portfolios Cut DSM Exposure

Sep 30, 2026 · @Sidharth

A forecast tells you what a plant may generate. A dispatch decision tells you what to schedule with the SLDC or RLDC, what to hold in the battery and what to sell on IEX, for each of the 96 time blocks in a day. The gap between those two is where Indian renewable generators lose money: in DSM charges, in market revenue left on the table and in missed peak-supply obligations.

Arkco is built around that gap. This article explains how forecast-to-dispatch orchestration works for Indian renewable portfolios: why physical limits come first, why a grid curtailment must never be booked as an outage, why the most accurate forecast is not always the cheapest schedule, and how to prove the result in settlement.

Key takeaways

• A schedule must fit the plant's physical envelope (available capacity, connectivity quantum, battery state of charge) before any commercial logic ranks it.

• DSM charges are settled block by block, so the schedule that minimises expected deviation cost can differ from the most accurate point forecast.

• Grid-initiated curtailment and plant outages need separate records. Mixing them distorts both DSM analysis and any compensation claim.

• Storage turns scheduling into a trade-off across time: energy used to cover a daytime shortfall is energy not available for the evening peak, whether for a PPA obligation or a high-priced RTM block.

• Judge orchestration on every value stream it touches: forecast accuracy, DSM charges, trading revenue and PPA dispatch compliance, each against an agreed baseline. In a 126 MW wind pilot, Arkco's engine cut DSM charges by about 45%.

What is renewable portfolio orchestration?

Renewable portfolio orchestration is the coordination of generation, storage, flexible load and market positions across the same 15-minute time blocks, so every block has a schedule that is physically deliverable and commercially sound. The output is not a forecast. It is a decision: a schedule to submit, a revision to make, a battery action or a bid to place.

In India, three things make this harder than it looks:

• DSM is settled block by block. A good day on average can still carry heavy charges from a handful of blocks around a wind ramp or a cloud front.

• Schedules can only be revised with a lag. Intraday revisions under IEGC 2023 take effect a few time blocks after submission, so a forecast update is only useful if it arrives before the revision deadline for the block it affects.

• Hybrids and storage create obligations across time. RTC, FDRE and peak-supply contracts tie what you do at 2 pm to what you must deliver at 8 pm.

Orchestration connects these into one decision chain: forecast the resource, check what the plant can physically deliver, then choose the schedule, battery action or market position with the best expected commercial outcome.

Why do physical limits come first?

No optimiser can schedule energy the plant cannot deliver. Engineering data sets the feasible range for each block; commercial logic only chooses within it.

Available capacity, not nameplate. Wind speed and irradiance set potential output. Turbine availability, inverter derating, feeder faults and planned maintenance set how much of it reaches the pooling substation. A weather forecast can be right and the schedule still wrong if the availability picture is stale.

Connectivity is often the binding constraint. Many Indian hybrids are built with combined wind and solar capacity well above their connectivity quantum. When both resources are strong at once, export is capped at the substation limit and the extra energy is either stored or lost. A schedule that ignores this will over-commit in exactly the blocks where the weather looks best.

Storage has its own envelope. A battery's state of charge, power rating, round-trip losses and charging source limit what it can do. Some projects may only charge from their own generation, not the grid. Orchestration has to track these limits block by block rather than treat the battery as an unlimited buffer.

Keeping resource shortfalls and asset failures separate matters too. A resource miss points to the forecast model. An asset failure points to O&M and an availability update. Both change the schedule, but they need different fixes.

How is grid curtailment different from a plant outage?

A curtailment is an instruction from the SLDC or RLDC that limits injection even though the plant is available. An outage is a plant condition that limits what it can generate. Both reduce export, but they have different causes, different paperwork and different commercial consequences.

Grid curtailment. Wind and solar have must-run status in India, so curtailment is meant to happen only for grid security or network constraints. When it does, the priorities are to comply, to record the instruction (time, quantum, effective blocks, reason), and to check the schedule was revised in line with it so the curtailed energy isn't treated as a deviation. Curtailment for reasons other than grid security may be compensable under the Must-Run Rules, 2021, but only if the lost generation is documented block by block.

Plant outage. An inverter trip, turbine fault or feeder failure lowers deliverable output below the submitted schedule. Here the questions are how fast the fault reached the scheduling desk, whether a revision can still take effect in time, and when availability will be restored. A fault picked up after the revision deadline for a block cannot save that block, but it can still protect the ones after it.

The common failure is misclassification. Booking a curtailment as a forecast miss makes the forecasting look worse than it is and weakens any compensation claim. Booking an outage as weather hides an O&M problem. Orchestration should tag every shortfall with its cause.

Why isn't the most accurate forecast always the best schedule?

Because DSM does not charge for average error. It charges for deviation beyond a tolerance band, block by block. A schedule built to minimise expected DSM cost, which Arkco calls commercially-biased forecasting, can beat a schedule built to minimise mean absolute error.

The idea is simple. A forecast is really a range. When that range is skewed, for example ahead of a likely wind ramp-down, scheduling at the centre of the range exposes the plant to a large under-injection if the downside arrives. Shifting the schedule towards the risk costs a little in the blocks where the downside doesn't come, and saves a lot where it does.

The rupee effect in any block depends on the applicable DSM charge rate, the reference price and how under- and over-injection are treated. Weighing those against the forecast range is what the optimiser does. The bias is not a fixed offset: it is chosen per block and per asset, and it goes to zero when the forecast range is narrow.

Where do IEX markets and BESS come in?

Once the feasible range is known and contracted obligations are covered, the remaining energy and flexibility can be put to work: shifted in time with storage, or sold in the Day-Ahead Market (DAM) or Real-Time Market (RTM) on IEX. Each option competes for the same MWh.

Contracts come first. A project with a PPA profile, an RTC or FDRE commitment, or an evening-peak supply obligation must reserve energy for it before offering anything to the market. An attractive RTM price does not make an uncontracted volume deliverable, and it doesn't make a contracted one available.

The battery links every block to the ones after it. Indian exchange prices usually bottom out in the solar hours and peak in the evening, so the natural cycle is to charge from surplus daytime generation and discharge into the evening peak. But the same stored energy has competing uses: covering a wind shortfall to protect the DSM position, meeting a PPA peak-supply obligation, or selling into a high-priced evening RTM block. Using it for one leaves less for the others. Orchestration puts these on one timeline and prices them against each other, rather than letting DSM, trading and the peak obligation each claim the battery separately.

Market timing is a constraint, not a detail. RTM auctions run every half hour for delivery shortly after, and DAM bids close the day before. A bid recommendation is only useful if it can be placed, through the trader or the generator's own exchange access, before the gate closes and matches what the plant can physically deliver.

How does Arkco put this into operation?

Arkco is the decision and orchestration layer, and it fits the generator's existing set-up rather than forcing a new one. It can work through the generator's QCA or trader. Where the generator trades on its own licence, Arkco supports it directly, integrating with IEX and the load despatch centres for scheduling and bidding.

A typical deployment runs in three stages:

1. Train on the asset. Models are tuned to each plant using its own SCADA, availability and settlement history, rather than one generic model across a portfolio. The engine behind it is e2m, developed and run by Metro Power across Australian renewable and industrial assets for about 16 years.

2. Agree the baseline. Arkco and the customer fix the reference for measuring results: historical DSM charges, realised market prices and the PPA compliance record. Customers who want proof before committing can start with a time-bound pilot, in which Arkco's schedules run in parallel and are replayed through actual settlement.

3. Operate live. Arkco's schedules, revisions, bids and battery instructions go out through the agreed channel. At onboarding, the customer decides which actions run automatically and which need operator sign-off, and how exceptions such as a curtailment instruction or a data-feed failure are handled. Every action and override is logged.

Pricing follows the same logic. Arkco charges a platform fee plus a share of the measured gains against the agreed baseline, whether they come from lower DSM charges, higher trading revenue or better PPA compliance.

How do you know orchestration worked?

Measure every value stream orchestration touches, each against the agreed baseline and each from settlement or meter data rather than model output. A better forecast is not a gain until it changes a decision that shows up in a statement.

Keep the operating record complete. Log every override, curtailment instruction and maintenance event, so a gain from a decision isn't confused with a change in conditions.

Pilot result. In a six-month pilot on a 126 MW wind asset in India, run in parallel against real DSM settlement across 3,360 data points, Arkco's engine improved day-ahead MAE by about 27% and 90-minute-ahead MAE by about 30% over the incumbent forecast. Replayed through actual settlement, DSM charges fell by 44.9% on a pool-adjusted basis and 47.9% at SPV level (p < 0.001). As a standalone wind asset, the pilot measured forecasting and DSM; trading, storage and PPA-compliance gains are measured on portfolios where those streams exist.

Frequently asked questions

How can a wind or solar generator in India reduce DSM charges?

By scheduling to minimise expected deviation cost rather than average forecast error, updating schedules before the IEGC revision deadline for each block, keeping availability data current, and using storage where it has it. The biggest savings usually come from a small number of volatile blocks around ramps and weather fronts.

What are the DSM tolerance bands for wind and solar?

Under CERC's DSM Regulations, 2024, the bands from April 2026 are ±10% for wind and ±5% for solar, with further tightening on a glide path to FY2031-32. Hybrids and portfolios should check how the bands apply to their specific connectivity and scheduling arrangement.

Is grid curtailment the same as a plant outage?

No. Curtailment is an SLDC or RLDC instruction that limits injection while the plant is available. An outage is a plant or equipment condition. They should be recorded separately, because the schedule revision, DSM treatment and any compensation claim depend on the cause.

Does Arkco work with our existing QCA and trader?

Yes. Arkco can work through a generator's existing QCA or trader. Where the generator trades on its own licence, Arkco supports it directly, integrating with IEX and the load despatch centres for scheduling and bidding.

Does Arkco dispatch automatically?

It can. The customer decides at onboarding which actions run automatically, which need operator approval, and how exceptions are handled. Every action and override is logged for settlement reconciliation.

How does BESS change the scheduling problem?

Storage lets energy move between blocks, so a decision now changes what's possible later. The optimiser weighs DSM cover, market margin and contracted peak delivery on one timeline, within the battery's state of charge, power and charging limits.

How is Arkco priced?

A platform fee plus a share of measured gains against an agreed baseline, from lower DSM charges, higher trading revenue or better PPA compliance, so the customer pays mainly for results that show up in settlement.