Why Inverters Broke the Physics of Grid Simulation
· Arshi Chadha
India now runs hundreds of gigawatts of non-fossil capacity. Earlier this year, a single grid event tripped several gigawatts of generation at once. Those two facts belong in the same sentence. The first is the result of a decade of building. The second is what happens when the way we build changes faster than the way we check.
This is our view of what the problem actually is, and what it will take to fix it.
The grid used to stabilise itself
For most of its history, the Indian grid was held together by spinning steel. Large synchronous generators carry physical inertia. When a fault hits, they resist the change in frequency and push large fault currents into the network without being asked. Nobody had to program that behaviour. It came from physics.
Solar and wind plants connect through inverters. An inverter does exactly what its control software tells it to do, and nothing more. Most inverters in the field today are grid-following: they read the voltage of the grid and synchronise to it. They do not set the voltage. They follow it.
One grid-following plant on a strong network is easy to live with. Thousands of them clustered in the same corridors in Rajasthan, Gujarat or Karnataka are a different system. In those pockets, inverters increasingly follow a voltage that is being shaped by other inverters. The reference they are tracking is no longer firm.
Why "weak grid" is the wrong frame
Engineers describe these pockets with the Short Circuit Ratio: how strong the grid is at a connection point relative to the size of the plant. A low ratio means a weak connection.
The trouble is that the ratio is usually calculated one plant at a time. When several gigawatts of renewable plants share the same transmission paths, they interact. A plant that looks acceptable on its own can sit inside a cluster that is not. Measures that account for neighbouring plants, such as a weighted short circuit ratio, exist for exactly this reason. They need to become the default way connections are planned, and system strength needs to be published as a forecast, so developers know what they are connecting into five years from now, not only today.
The simulation trap
Before a plant connects, its developer submits a computer model of how its inverters and controls will behave during a fault. The model is run through detailed studies. If it passes, the plant is declared compliant.
The uncomfortable lesson of the last few years is that a plant can pass on paper and misbehave in the field. In real disturbances, plants have delivered reactive support after the fault had already cleared, when it no longer helped. Others have pushed reactive power in the wrong direction for seconds before tripping, making the voltage problem worse.
None of this means the models were dishonest. It means a model is a description of intent, and the field adds things the model does not contain: firmware versions that differ from the ones tested, plant controllers that respond hundreds of milliseconds slower than the inverters underneath them, settings changed after commissioning, and interactions with neighbouring plants that no single study captured.
Compliance today is mostly checked once, at the start. The plant keeps changing for twenty-five years.
The part nobody sees: the distribution network
Utility-scale plants at least get studied. Rooftop solar and other distributed resources do not, and there are now tens of gigawatts of them in India, connected behind distribution transformers and governed by state rules that differ from one another.
The national operator has almost no live view of how these systems behave during a disturbance. If a transmission event causes a large share of them to disconnect at once, the loss stacks on top of the original fault. That is the mechanism that worsened the 2025 blackout on the Iberian peninsula. A grid can be destabilised from below by equipment that no control room is watching.
Models can be wrong. Measurements cannot be.
The conversation in India is moving, and that is encouraging. Forums such as the India Smart Grid Forum have brought the system operator, regulators, planners and industry into the same room to work through exactly these questions, and the direction of that discussion matches what we see from the field.
Our position is simple. Stability cannot rest on a certificate signed at commissioning. It needs three things that are checked continuously:
1. Identity. Knowing which physical device is connected where, which firmware it runs and who is responsible for it. Without this, no measurement can be traced back to an asset. 2. Measured behaviour. Recording how plants and distributed systems actually respond to grid events, and comparing that with the model they were approved on. 3. Independent verification. Testing and certification by accredited labs, rather than self-declaration, with results that remain valid only while the field data agrees with them.
This is the layer India is missing. Not more generation, and not more simulation, but independent, field-level verification of how assets actually behave, telling the operator, the regulator and the financier what is really connected and how it really behaves.
What should happen next
- Plan connections using cluster-level system strength, not single-plant ratios, and publish how that strength is expected to change.
- Require accredited, hardware-tested certification of inverter and plant controller behaviour before connection.
- Treat model validation as ongoing: compare event recordings with approved models after every significant disturbance.
- Bring distributed resources into view with common rules for ride-through behaviour and data access across states.
- Align plant controller and inverter response times so they support each other instead of fighting.
India built hundreds of gigawatts faster than almost anyone thought possible. Keeping it stable is now a question of evidence. The plants are already producing the data. The work is to make it trusted.