Let's appreciate what it takes for a battery inverter to impersonate a heavy spinning machine: a neural network, a real-time grid estimator, and a controller that retunes its own settings on the fly.
Pumped storage hydropower skips the impersonation. Its generators are heavy spinning machines. Hundreds of tons of rotating steel, synchronized with the grid, resisting disturbances the instant they hit. No settings to drift out of tune. No model to retrain.
The best virtual synchronous generator is still an actual synchronous generator.
Batteries have an important role to play, but as the grid adds more inverter-based resources, it needs more of the real thing.
The research
New work from the University of Michigan-Dearborn and Pacific Northwest National Laboratory, published in Electric Power Systems Research in 2026:
- Inverters programmed to imitate spinning generators, known as virtual synchronous generators, showed oscillations, long settling times, and overshoot as grid conditions changed.
- Counterintuitively, the fixed-setting designs performed worse as the grid got stronger.
- The researchers' fix: a physics-informed neural network that continuously estimates the grid and retunes the controller in real time. It settled quickly with minimal overshoot across a wide range of grid conditions.
Impressive work. Also early-stage: the testing was in offline and real-time simulation, with no field trials yet.
Sources:
- Hoang et al., Adaptation of virtual synchronous generators to dynamic conditions in power grids, Electric Power Systems Research 261 (2026)
- Pacific Northwest National Laboratory, publication listing
Originally posted on LinkedIn.
Is this issue on your desk?
Start a conversation