The familiar reference point of a power system is disappearing. For most of the history of electricity, synchronous machines — large rotating generators — set the voltage and frequency, and everything else followed. A growing share of generation now comes from inverter-based resources that were originally designed to follow that reference rather than create it. The grid-forming inverter changes that arrangement: it can establish voltage and frequency instead of merely tracking them. As synchronous machines retire, the question is how much of that role can move to power electronics, and what changes when an inverter stops following the grid and starts setting it?
The Difference Between Following and Forming
A grid-following inverter measures the existing voltage waveform at its terminals and injects current into it. This approach works well as long as synchronous machines or another strong source provide the voltage reference, because the inverter never has to establish the waveform itself. A grid-forming inverter, by contrast, synthesizes its own voltage and frequency reference, behaving more like a voltage source than a current source. Grid-forming capability removes one of the constraints on operating a network without synchronous machines, but the island case also depends on available energy sources, protection settings, control tuning and how the system responds to a frequency disturbance.
What an Inverter-Dominated Grid Loses
The first loss is inertial response. Synchronous machines store kinetic energy in their rotating mass, and when frequency changes, that kinetic energy is exchanged almost immediately, giving the system time to respond. Inverters have no rotating mass, so this physical response is absent unless control software is deliberately designed to emulate it. In systems where coal plants are retiring, gas plants often remain as the synchronous backbone. Grid-forming controls can inject power quickly based on frequency deviation, but that is a converter action, not stored kinetic energy.
The second loss is fast frequency response. Synchronous machines respond to frequency changes through governor action and inertial response, while gas peaking units contribute through fast-start reserves and ramping capability. Inverter-based resources can respond very quickly if programmed to do so, but grid-following units generally do not contribute unless their controls are set for frequency response. A grid-forming inverter, because it sets frequency internally, can be configured to inject or absorb active power almost immediately when system frequency deviates.
The third loss is fault current. A synchronous machine can deliver several times its rated current during a fault, which helps protection systems detect the fault and clear it. An inverter’s current is limited by its power electronics, so a grid-forming unit contributes much less fault current than a comparable synchronous machine, even if it is programmed to supply some. This has direct consequences for protection coordination, because existing overcurrent schemes may no longer see the fault current they were designed around.
Three Ways to Set the Voltage Reference
Droop control is common in practical deployments. Each grid-forming unit adjusts its active power output in proportion to frequency deviation and its reactive power in proportion to voltage deviation, using predefined droop slopes. When multiple units operate in parallel, these slopes determine how they share load without any external reference or communication. The trade-off is that frequency and voltage deviate from nominal even in steady state, which grid codes may constrain.
A virtual synchronous machine controller emulates the swing equation of a synchronous generator, including an inertia-like term and damping in the control software. The inverter’s active power setpoint adjusts according to the same dynamic equations a rotating machine would follow, so the electrical behaviour at the terminals resembles that of a synchronous machine. No kinetic energy is stored, but the control loop can be tuned to provide a power response shaped by that inertia-like term. The approach can ease integration with existing system studies, though it may introduce slower dynamics than droop if not carefully designed.
Virtual oscillator control takes a different route, deriving the inverter’s voltage and frequency from the equations of a coupled nonlinear oscillator. Units synchronize with each other through the network itself, so parallel operation emerges without a central controller or explicit droop slope assignment. The approach has been demonstrated in microgrids and research settings, but it is less established in utility-scale deployments than droop or virtual synchronous machine controls.
Operational Difficulties: Stability, Black Start, Protection
Load sharing among parallel grid-forming units can be coordinated through supervisory control, though where units operate without a central controller the arrangement raises stability questions. With many independently synthesizing voltages, weakly damped interactions can develop between converters, particularly when network impedance or control gains are not carefully coordinated. The exposure is greatest when only a few gas plants remain synchronous in a given area. Grid codes are still working out how to specify stability requirements for plant-level grid-forming behaviour, which is one reason the distinction between capability and utilization matters.
Black start changes the problem. A grid-forming unit can energize a dead network and build voltage from nothing, but restarting a whole system requires a sequence. The unit needs an energy source to support its own auxiliary loads, and the network must be sectionalized to avoid energizing too much load at once. A grid-forming inverter can be part of a black-start resource, but it is not by itself a complete restoration plan.
Protection coordination becomes harder when fault current is limited. Overcurrent relays that depend on high fault current to detect and isolate faults may need to be replaced or supplemented with schemes that do not rely on large fault currents, such as differential or directional protection. This is a system-level cost that is easy to overlook when the focus is only on the inverter.
Standards and the Commercial Picture
IEEE Std 2800-2022 is the first edition of a voluntary IEEE standard setting consensus technical minimum capability and performance for inverter-based resources interconnecting with transmission systems. NERC has adopted parts of IEEE 2800-2022 into reliability standard PRC-029 for ride-through capability, responding to FERC Order 901. The associated requirements become mandatory for inverter-based resource plants above 75 MVA connected at 100 kV or higher on October 1, 2026, and above 20 MVA at 60 kV or higher on January 1, 2027.
A separate IEEE recommended practice, P2800.1, is intended to specify technical minimum functional and performance capabilities for grid-forming inverter-based resource equipment, including verification test procedures and criteria. It is framed as complementing IEEE 2800, but it explicitly places the actual utilization of those capabilities in any given application outside its scope. In parallel, grid codes in jurisdictions such as Great Britain and Australia have moved toward specifying grid-forming capability for inverter-based resources, although the requirements differ. An amendment, P2800a, modifies IEEE 2800-2022 to reduce potential technical barriers for inverter-based resources containing grid-forming equipment where justified considering reliability and security needs.
On the commercial side, grid-forming capability generally adds cost relative to a grid-following unit, but the premium is driven more by controls, modelling and compliance testing than by a fundamentally different power stage. The same manufacturers that dominate utility-scale inverter supply are adding grid-forming firmware and certification rather than designing new hardware from scratch. The cost difference tends to narrow as requirements become more standardized, though early adopters face additional engineering and grid-study expenses.
Where Systems Have Run Without Spinning Machines
Isolated microgrids and small island systems have already operated without synchronous machines, with one or a few grid-forming units establishing voltage and frequency. Larger interconnected systems have tested high instantaneous shares of inverter-based generation, often keeping gas peaking units online as a safety net, which limits what those trials reveal about fully inverter-dominated operation. These demonstrations are distinct from routine commercial operation of an entire large grid with no synchronous generation anywhere.
What remains unresolved is how to manage the transition across a full control area. Plant-level grid-forming requirements may be considered for integration into the ongoing revision of IEEE 2800, but actual utilization remains outside P2800.1’s scope. Protection, stability across many units, and black-start sequencing still depend on system-specific engineering. The direction is toward greater specification of capability, but the operational experience at scale is still being built.
References
- IEEE Power & Energy Society — “Inverter-Based Resources Standards in Electric Power Grids”: background on IEEE 2800-2022, NERC PRC-029, FERC Order 901, and mandatory compliance dates.
- IEEE Standards Association — “IEEE P2800.1 Recommended Practice”: scope and relationship of P2800.1 and P2800a to IEEE 2800.
