By Shilpy Dewan, Vice President – Markets, Operations and Digital, Serentica
Every monsoon, Mumbaikars joke that a red alert for rain often guarantees a bright, sunny day. It is the kind of irony we laugh about while navigating flooded streets or working from home on a day when the sky stays clear. But what feels like a harmless quirk in daily life carries much higher stakes in the renewable energy world. For wind and solar developers, predicting the weather with precision is mission-critical. A missed gust of wind or an unanticipated shower can translate into millions of dollars lost in a single day, while throwing entire grid operations off balance.
Why forecasting matters for renewables
For years, wind and solar power generators were treated as “must-run” resources – whatever they generated was absorbed into the grid. This worked when renewables made up a small share of the mix. However, as India rapidly scales up capacity, variability in its output has become too large to simply “fit in” around conventional power. Wind generation can drop significantly within minutes if the wind dies down, forcing grid operators to instantly find a backup supply equivalent to powering a small city. Conversely, a sudden burst of solar generation on a mild afternoon can flood the grid when demand is low, leading to curtailment and wasted clean power. An accurate forecast gives system operators visibility and control, and project developers a chance to plan revenues in an increasingly complex market.
From physics to Artificial Intelligence: A new era of forecasting
Since the 1950s, weather forecasting has relied on numerical weather prediction (NWP) – physics-based models and statistical techniques that mimic atmospheric processes. This requires multimillion-dollar supercomputers capable of solving complex equations that mimic atmospheric processes after ingesting massive volumes of data from satellites. Over the years, this approach has grown in scope and accuracy, but it remains computationally demanding and slower to adapt. As advanced and as complex as these models are, they have spatial and temporal limitations, leading to inaccuracies that get further magnified by microclimates, terrain and the hyperlocal weather phenomena of a particular site.
For wind and solar, even small deviations, such as a shift in wind speed by a few meters per second or cloud formation at the wrong hour, can translate into hundreds of MWs of error on the grid.
This is where artificial intelligence (AI) offers a new edge. By learning directly from historical patterns and integrating real-time data, AI models can capture micro-trends that traditional models often miss and continuously refine their outputs. Hybrid approaches that combine physics-based weather models with AI are emerging as the most promising frontier. AI-driven forecasting can provide not just day-ahead visibility but also short-term (intra-day) updates, allowing generators to continually refine their schedules. In fact, many AI engines are now trained on decades of NWP data to learn the cause-and-effect patterns that govern how weather evolves.
On the scheduling side, AI can enable dynamic optimisation in a complex future where ancillary services and congestion charges can be the norm. By combining forecasts with market signals, such as day-ahead prices, ancillary service costs and congestion charges, algorithms can recommend when to sell, when to store and when to curtail. This integration of forecasting with commercial strategy is critical as India’s market moves towards higher levels of competition and real-time trading.
While much AI effort today is focused on predicting extreme events like floods or cyclones, the same techniques can be scaled to enhance renewable forecasting. The European Centre for Medium-Range Weather Forecasts has even launched an AI Quest, inviting global players to compete and develop sub-seasonal forecasts (for two weeks or more ahead) using AI and machine learning.
Regulations running ahead of science
While advancements are slowly taking shape in the world of weather predictions, Indian regulations have already taken a stern view of renewable energy developers and the precision with which they should forecast their output.
To understand the current regulations, think of it like hosting a dinner party where you must declare the exact number of guests and their consumption well in advance. If more arrive, you cannot order extra food; if fewer come, you pay double for the wasted meals. That is how India’s scheduling regime treats renewable generators today: forecasts must be locked in, and any deviation from actual output attracts financial penalties, even when the grid may have the room to adjust.
These penalties, known as deviation charges, are already high and set to increase further from April 2026. The recently released draft regulation envisions wind and solar to be treated like any other conventional source of power, like thermal, stripping away the flexibility that acknowledges their weather-driven variability. From 2026 onwards, the margin of error allowed will shrink sharply, and penalties for deviations will rise. Without the ability to offset shortfalls or surpluses through the market, developers are left carrying the full burden of uncertainty.
While this reflects the growing role of renewables in the grid, the challenge is that forecasting wind remains inherently difficult, with weather models still far from perfect. Without parallel measures to improve prediction tools and data quality, stricter regulations risk penalising variability rather than enabling solutions.
Contrast this with international practice. In Europe, the role of Balance Responsible Parties allows generators to pool their risks. A wind farm can aggregate with a thermal unit, a gas plant or even demand-side management resources. If the wind underperforms, another source steps in, keeping the overall schedule balanced. It is a team sport – one player’s miss is covered by another’s shot on goal. The idea is not to punish variability, but to manage it collectively.
While India has allowed pooling at the substation level with other renewable generators, this is still a small concession compared to the huge penalties that are being imposed. There is still no mechanism for renewable generators to offset their imbalances through the market, leaving them to bear the full cost of forecast errors.
Rethinking the role of system operators
India relies heavily on penalising the variable generator. However, grid stability is not a one-sided responsibility. Just as traffic police do not fine every driver in a sudden jam but instead manage flow with diversions, signals and patrols, it is time that Indian regional load despatch centres are equipped to take on a more proactive role. To ensure active ancillary services, technologies such as grid-forming static synchronous compensators, battery-as-a-grid assets, frequency response markets and a more comprehensive toolkit for system operators are critical to ensure reliability, without making renewables the scapegoat for every imbalance.
Just as traffic management is not about fining drivers alone, grid reliability should not rest solely on penalising renewable generators. Globally, system operators are evolving from rule enforcers into active facilitators of flexibility; India must move in that direction if it aims to integrate renewables at scale.
Towards 24×7 clean power: A global benchmark in the making
India’s energy transition is not just about adding more MWs of wind and solar – it is about making those MWs dependable. Forecasting is the first step, but the real achievement lies in building a system that can embrace variability without punishing it. This requires blending different resources, such as solar, wind, hydro, storage and even flexible thermal, into integrated portfolios that can guarantee firm output.
“Green power-as-a-service” models are emerging, where generators take on end-to-end responsibility for supplying clean energy under service-level agreements. To his end, Serentica has signed time-block guarantee contracts with industrial consumers. This shifts the burden of variability away from the industrial consumer, allowing them to lock in predictable costs and decarbonisation outcomes.
A punitive approach to weather-dependent generation will only push developers into risk aversion, slowing the very investment India needs. However, an enabling framework – one that combines AI-driven forecasting, market-based risk pooling and proactive system operations – can unlock confidence, capital and scale.
If India gets this balance right, it will not only meet its renewable targets but also set a global precedent – demonstrating how a fast growing economy can run on weather-dependent power while keeping the lights on, costs low and emissions falling. This is the benchmark the world is waiting for us to set, and the opportunity we cannot afford to miss.
