Grid-Forming Inverters

From Niche BESS Application to Standard Grid Requirement

1. Introduction: A Paradigm Shift in How Grids Are Held Together

For roughly a century, the frequency and voltage of every large power system on Earth were created by spinning metal. Coal, gas, nuclear, and hydro plants did more than deliver energy — the rotating mass of their turbine-generator shafts physically established the 50 Hz or 60 Hz heartbeat of the grid and resisted every disturbance with stored kinetic energy. Inverter-based resources (IBRs) such as solar, wind, and battery energy storage systems (BESS) upended that arrangement. Historically, these resources have been grid-following: fast, precise, and responsive — but fundamentally dependent on an externally established voltage waveform to synchronize to.

Grid-forming (GFM) inverter technology closes that gap. A grid-forming inverter does not wait for the grid to tell it what to do; it establishes voltage magnitude, angle, and frequency itself, behaving toward the network the way a synchronous machine does. In doing so, it provides the services a high-renewables grid cannot live without: synthetic inertia, fast fault current, phase-angle jump withstand, voltage stiffness, oscillation damping, and even black-start capability.

What was, only a few years ago, a niche capability proven mainly on island grids has become a mainstream requirement. Transmission system operators (TSOs) across Europe, Australia, parts of Asia-Pacific, Latin America, and a growing number of North American markets have moved from technical white papers to binding grid-code mandates and, in some cases, to dedicated revenue mechanisms that pay batteries for inertia. This article reviews the technology, the physics of inertia, the global regulatory landscape, the emerging revenue models, and — most importantly for asset owners and developers — the plant-level engineering that determines whether a grid-forming project actually works and actually gets paid.

2. Grid-Following vs. Grid-Forming: The Fundamental Difference

The most widely used shorthand in the industry is that a grid-following (GFL) inverter behaves like a controlled current source, while a grid-forming inverter behaves like a voltage source behind an impedance. That single sentence carries almost everything that matters.

2.1 The Grid-Follower: Fast, but Always One Step Behind

A grid-following inverter continuously measures the grid voltage — typically through a phase-locked loop (PLL) — and injects current aligned to that measured waveform to hold its commanded active and reactive power setpoints. When the grid frequency sags after a generation trip, the GFL inverter does exactly what it was designed to do: it tracks the moving waveform and holds its megawatt output constant. Every corrective action (frequency response, voltage support, oscillation damping) requires a measurement, a control decision, and a dispatch — and that measurement-control-dispatch cycle introduces dead time. A useful mental model is that a grid-following inverter is a grid copier: when the voltage steps, it steps with it; when the frequency accelerates, it accelerates with it; when the phase angle jumps, it snaps to the new angle. It faithfully reproduces whatever the grid does, including the disturbance itself.

2.2 The Grid-Former: Holding Its Ground and Letting Physics Work

A grid-forming inverter inverts that logic. During the transient and sub-transient window immediately after a disturbance, it holds its internal voltage — both magnitude and angle — steady. Because the plant now looks like a stiff voltage source connected to the disturbed network through an impedance, basic circuit physics does the rest: whatever current is needed to oppose the disturbance flows naturally and instantaneously, with no measurement latency and no dispatch delay. In that sense a grid-forming inverter is a grid corrector rather than a grid copier. Consider four canonical grid events:

• Voltage magnitude step. The grid-former sources or absorbs reactive current instantly to partially restore the voltage, exactly as a synchronous machine's field-driven EMF would.
• Rate of change of frequency (RoCoF). As frequency accelerates or decelerates, the grid-former initially maintains its own angle trajectory, exchanging active power with the grid just as a decelerating flywheel releases kinetic energy — this is synthetic inertia.
• Phase-angle jump. When the network voltage angle suddenly shifts (a common consequence of line switching or faults), a large angle difference appears across the coupling impedance and a substantial power flow results, kicking the grid back toward alignment. Grid codes commonly require withstand capability for angle jumps of up to 60 degrees.
• Voltage imbalance. Where a grid-follower typically regulates negative-sequence current to zero (preserving current headroom for its day job), a grid-former naturally provides bespoke per-phase current that counteracts the imbalance.

KEENTEL INSIGHT — Why Latency Is the Whole Story

The distinction is not that grid-forming responds faster than grid-following — modern GFL controls are extraordinarily fast. The distinction is that a grid-former's stabilizing response requires no control action at all in the first instants. The response is a property of the physics of a voltage source behind an impedance, not the output of a control loop. That is what makes it functionally equivalent to rotating inertia.

3. Inertia: What It Is, How It Is Measured, and Why It Is Vanishing

3.1 The Inertia Constant

System inertia is quantified through the inertia constant H, expressed in seconds. Conceptually, take the kinetic energy stored in a machine's rotating mass — measured in gigawatt-seconds, the same energy unit as gigawatt-hours, only smaller — and divide it by the machine's rated active power in gigawatts. The result answers a simple question: how many seconds' worth of full-power energy is stored behind every gigawatt on the grid? The higher the aggregate H, the shallower the rate of change of frequency (RoCoF) after any generation-load imbalance, and the more time slower frequency-response resources have to act.

3.2 Not All Generators Are Equal

Different technologies bring dramatically different inertia contributions. Nuclear units, with their massive high-speed turbine shafts, sit at the top of the range. Combined-cycle and open-cycle gas plants contribute meaningfully but less. Biomass fleets average lower still, in part because many are internal-combustion engine plants with modest rotating mass. Wind and solar operating through conventional grid-following inverters contribute exactly zero — and, crucially, every megawatt-hour of GFL renewable output that displaces a synchronous machine removes that machine's inertia from the system in the same moment. Purpose-built solutions exist at the other extreme: synchronous condensers are, in effect, freely spinning flywheel machines built for nothing but inertia and fault current, and they deliver high inertia constants. Grid-forming BESS is unique in the lineup: because its 'flywheel' is a set of equations in software, its effective inertia constant is a design parameter — tunable across a wide range, subject to the hardware sizing consequences discussed in Section 7.

3.3 The Declining Floor

System planners do not design for average conditions; they design for the worst credible hours. A revealing metric is P10 system inertia — the level of inertia at or below which the grid operates during the worst ten percent of hours in a year. Across European systems, that floor has been trending steadily downward as synchronous plant retires or is displaced in the dispatch merit order. Low-inertia hours are exactly the hours when renewables output is highest, thermal units are decommitted, and the system is most exposed.

3.4 Why Planners Care: Cost and Catastrophe

There are two distinct drivers. The first is economic: on a low-inertia grid, frequency moves faster and farther after every imbalance, so more — and faster — frequency response must be procured to arrest it. Running a low-inertia grid is simply more expensive hour by hour. The second is security: interconnected continental systems can, under extreme stress, split into separate frequency islands along internal transmission bottlenecks. Such a system separation has happened before on the European interconnection, and the fault lines of a future split are broadly predictable. Notably, the planning need is asymmetric: regions prone to surplus generation after a split require negative inertia capability (resisting over-frequency), while import-dependent regions require positive inertia (resisting under-frequency). Grid-forming resources inherently provide both — a decelerating virtual flywheel absorbs energy just as readily as an accelerating one releases it.

KEENTEL INSIGHT — Inertia Is a Locational Product

Even where procurement frameworks accept inertia from anywhere in a control area, stability studies identify clear regional needs — and regions can even need over-frequency (negative) inertia rather than under-frequency support. Developers who understand the locational stability picture can anticipate where future locational price signals, or future bid rejections in saturated zones, are most likely to emerge.

4. How a Battery Emulates a Flywheel

An inverter has no rotating mass, but it can carry the equations of one. Virtual synchronous machine (VSM) and related grid-forming control structures embed the swing equation of a rotating machine in the inverter's voltage-reference generator. When grid frequency begins to fall, the virtual rotor 'decelerates' according to its programmed inertia constant, and the resulting angle difference between the inverter's internal voltage and the grid pulls real power out of the DC link and into the network — precisely the behavior of a flywheel giving up kinetic energy. The injected power in the first hundreds of milliseconds reduces the initial RoCoF, which produces two system-level benefits: primary frequency response resources gain time to act, and the frequency nadir they must arrest is shallower.

Because the inertia constant is a software parameter, a grid-forming BESS can present anywhere from a token H to values at or beyond the highest synchronous machines — some markets accept inertia bids up to 12.5 seconds. But software generosity must be backed by hardware honesty: the active power that a 12.5-second virtual flywheel must deliver during a severe RoCoF event is very real, and it must flow through real semiconductors, real batteries, real transformers, and real cables. That tension between what the controls promise and what the plant can physically deliver is the central engineering theme of grid-forming design, and we return to it in Section 7.

5. The Global Regulatory Landscape: Three Ways to Require Grid-Forming

Grid-forming requirements are proliferating worldwide, but they are far from uniform. Broadly, every framework we have reviewed falls into one of three archetypes — and a growing number of jurisdictions combine them.

5.1 Archetype 1: Blanket Mandate Within Inherent Limits

The simplest approach: every new BESS must operate in grid-forming mode whenever it has headroom or footroom within its rated capability envelope. No specific inertia constant, overcurrent margin, or droop characteristic is prescribed. Nordic system operators, the Texas market (with requirements phasing in during 2026), and a major Midwestern U.S. market (effective 2025) have all taken this route, and the pan-European transmission operators' framework points the same direction, with individual national codes to follow. The strength of this model is speed of deployment; its weakness is that, absent quantified requirements, developers rationally build the minimum. Most projects in mandate-only markets are configured with the lowest stable inertia setting — nobody adds hardware margin out of goodwill.

5.2 Archetype 2: Mandate Plus Explicit Overcurrent Capability

Some grids — often islanded, weakly interconnected, or topologically stressed — go further and require capability above 1.0 per unit current. One Pacific island utility has long required substantial overcurrent for up to five seconds. Chile's requirement, effective early this year, calls for significant overcurrent sustained for several seconds. A recent Iberian proposal specifies overcurrent for roughly half a second. Most demanding of all, one Latin American code requires plants to ride a 30-degree phase-angle jump without entering current limiting — that is, remaining in linear operation. Whether intended or not, that single sentence drives a very large short-duration overcurrent rating, because when two stiff voltage sources separated by a modest impedance suddenly develop a 30-degree angle difference, the resulting power flow is enormous.

5.3 Archetype 3: Market-Based Incentives — Carrots and Sticks

The third archetype attaches money to grid-forming behavior. Three national implementations illustrate the spectrum:

• Great Britain — stability tenders. The system operator first ran pathfinder-style stability tenders, in which grid-forming batteries offered inertia at the lowest cost of any technology — logically, since a battery earns its keep in energy and ancillary markets and sells inertia as an incremental service, whereas a synchronous condenser must recover its entire capital cost from stability payments alone. The transition from pathfinder contracts to an enduring tender framework has, however, been turbulent: in a recent tender round, every participating battery was excluded on technical eligibility grounds — including assets already delivering inertia under earlier contracts — while requirement documents evolved through multiple versions with limited transparency. The episode underlines that market design maturity matters as much as technology maturity.
• Australia — the avoidable charge. Rather than paying for grid-forming, the market charges for its absence: adding grid-following capacity at designated nodes triggers a locational system strength charge, avoidable by installing grid-forming capability. The result is one of the highest grid-forming BESS penetrations in the world, with announced projects suggesting the share will rise further. One caution: because the charge regime emphasizes fault-level contribution over inertia, many plants procure grid-forming inverters that are minimally tuned for inertial response — a reminder that grid-forming is not a single homogeneous product. Network companies in the region have also begun procuring system strength from grid-forming batteries directly, converting the stick into an emerging carrot.
• Germany — availability-based inertia contracts. A newly opened procurement lets projects lock in inertia revenue on contracts of two to ten years, with prices administratively set (reviewed on a two-year cycle) and payment tied to annual availability. A premium product requires at least 90% availability — suited to grid-forming BESS, synchronous condensers, and must-run units — while a second tier provides an uplift for partially available resources so that every new project has some incentive to install grid-forming capability. Contracts can include a run-up period, letting projects under development secure pricing before commissioning. Operators have signaled leniency on location in the near term, accepting projects anywhere rather than only in study-identified need zones — a deliberate choice to avoid missing the current construction wave.
  

6. The Quantification Problem: Grid-Forming ≠ Grid-Forming

Every code says roughly the same words — 'voltage source behind an impedance' — and then quantifies the requirement completely differently. Four dimensions dominate, and each one lands directly on hardware sizing:

• Overload / overcurrent profile. Some markets mandate overcurrent magnitudes and durations outright (several seconds in the island and Chilean cases; sub-second in the Iberian proposal). Others — Great Britain and Germany — mandate nothing but allow voluntary short-term overload to be monetized as inertia or short-circuit-level revenue, with the permissible overload calculated from the design RoCoF and the bid inertia constant.
• Phase-angle jump. Withstand requirements commonly reach 60 degrees. Some codes additionally ask suppliers to declare the largest jump the plant can absorb before current limiting; at least one requires 30 degrees with no current limiting at all — a linear-operation requirement with outsized hardware consequences.
• RoCoF magnitude and duration. Design RoCoF values cluster around 1–2.5 Hz of total frequency excursion, but with very different intensities and durations. The most onerous published proof requirement asks plants to demonstrate inertial power delivery through 1 Hz/s sustained for five full seconds — a 5 Hz sweep spanning essentially the entire operational frequency band.
• Power oscillation damping (POD). Damping obligations range from very slow inter-area modes (fractions of a hertz, relevant in long, weakly meshed systems) up to 10–20 Hz control-interaction modes in other networks. A plant tuned for one duty is not automatically fit for the other.

KEENTEL INSIGHT — Read the Test, Not the Title

Two markets can both label a product 'grid-forming inertia' while one implies roughly 25–30% overload for five seconds and the other implies 100% overload for one second. Procurement teams that compare vendor offerings on the label rather than on the specific overload-duration-angle-RoCoF envelope routinely buy equipment that cannot pass the commissioning test. We recommend building a compliance matrix per target market before the PCS shortlist is drawn up — not after.

7. Plant Engineering: Where Grid-Forming Projects Are Won or Lost

7.1 The Whole Plant Forms the Grid — Not Just the Inverter

The grid-forming control lives in the power conversion system (PCS), and it is natural to start there. But the service is delivered by the plant. Providing an active-power surge for an angle jump or RoCoF event raises immediate questions across the entire one-line: Does the battery have the instantaneous power capability at the state of charge where the event finds it? What happens to cell balancing and state-of-charge estimation? Are protection settings, transformers, and cables secure under repeated short-duration overloads? And at the top of the stack, the power system plant controller (PPC) must adapt: a grid-following fleet does exactly what it is dispatched to do, but a grid-forming fleet has been granted autonomy, and the PPC's role shifts from commanding to supervising — keeping a continuous watchful eye on assets that are now allowed to act on their own.

7.2 Sizing for the Revenue You Want

Every plant's capability is bounded by its PQ envelope, and the binding corner is almost always discharge-overexcited. Real projects always carry some unintentional overbuild — inverter capacity installed above nominal plant rating, driven by reactive power obligations and design margin. Markets that monetize inertia let owners sell that margin. The economics get much better with deliberate short-term overload capability — a boost factor. With, say, 25% overload available for around five seconds, a plant can guarantee inertial power delivery without reserving any trading headroom: the battery trades its full nominal range while the overload band stands permanently ready for inertia. The energy involved is trivial (a few seconds); the power capability is everything.

The arithmetic compounds quickly at high inertia bids. Delivering the maximum biddable inertia constant of 12.5 seconds at a 2 Hz/s design RoCoF requires roughly 100% additional power above nominal — 200% total — for about one second. One second is nothing to a battery energetically and a long time to power electronics thermally. With legacy central-inverter topologies, reaching that capability means additional inverter blocks, additional footprint, additional transformers, and additional medium-voltage switchgear — costs that erode the very revenue being chased. Modern AC-block architectures with distributed string inverters and inherent overload headroom change that calculus, allowing meaningful oversizing within the existing footprint.

7.3 Degradation, Micro-Cycling, and the SOC Accuracy Trap

A common owner concern: if inertial response tracks every wiggle of grid frequency, will the battery micro-cycle itself to death? The evidence says no — with one important caveat. Analyses of historical frequency data, projected onto weakening grids, indicate that the energy throughput attributable to inertial response is small, and the oscillation frequencies involved fall outside the ranges that battery literature flags as accelerating lithium-ion degradation. Long-term wear is a manageable, minor line item.

The nearer-term issue is state-of-charge accuracy. Inertial duty means continuous low-magnitude, bidirectional current — precisely the regime where current transducers are least accurate. Small measurement errors around zero current integrate into meaningful SOC estimation drift, degrading dispatch accuracy and available-energy reporting well before any electrochemical degradation appears. Mitigations exist — periodic rebalancing routines, coulomb-counting correction strategies, and control filtering — but they must be engineered in deliberately. A related and underappreciated stress case: merchant strategies increasingly sweep the full SOC window, meaning plants spend hours near 0% SOC — exactly where a sudden full-plus-overload discharge kick from a grid event is hardest on cells and most likely to expose inadequate short-term peak discharge ratings. Cell and module suppliers should be told, contractually and explicitly, what grid-forming duty will ask of their product.

7.4 Retrofits and Topology Risk: The Shared Steering Wheel

Retrofitting grid-forming onto existing sites is attractive on paper and hazardous in practice. Legacy topologies that parallel multiple inverters on a shared low-voltage transformer winding — excellent for granular DC-side control and DC fault segregation — carry inherent common-mode voltage and circulating-current risks, and grid-forming operation elevates both significantly. The intuition: in grid-forming mode each inverter holds the voltage steering wheel. Put four inverters on one winding and four drivers are steering the same wheel; the result is inter-unit oscillation risk that tends to trip the plant precisely when the grid needs it most. Safer retrofit paths include single-DC-bus central architectures, multi-winding transformers that give each grid-forming unit its own winding, and — the direction the industry is moving — AC-block string-inverter designs using silicon-carbide devices, whose much higher switching frequencies provide the inter-unit decoupling that shared-winding operation demands.

7.5 Validation: From Datasheet to Site Test

Grid-forming claims must be proven, and a mature validation ladder has emerged: paper due diligence against the target code; EMT (electromagnetic transient) simulation of the vendor controls; factory witness testing; hardware-in-the-loop (HIL) testing at unit level or with project-specific network models — now effectively compulsory in the Australian model-acceptance process and spreading elsewhere; and finally staged site testing at commissioning. Failure modes are predictable and recurring: inverters without genuine overcurrent margin misbehave at the edge of capability (in the worst documented cases responding in the wrong direction during angle jumps); units fail asymmetric-condition tests; and some controls attempt to 'intelligently' prioritize current phase during limiting, when the pan-European guidance is explicit — retain the voltage-source character, scale the current vector proportionally at the limit, and do not get clever about which component the grid supposedly needs.

KEENTEL INSIGHT — EMT Models Are the Product

In a grid-forming procurement, the EMT model and its validation records are as much the deliverable as the hardware. Projects that arrive at the interconnection queue with benchmarked EMT models, HIL results, and a clean compliance matrix move through operator review in a fraction of the time — and avoid the re-study loops that quietly consume a year of revenue.

8. The Business Case: What Inertia Is Worth

8.1 Headroom Is Expensive; Overload Is Cheap

The naive way to guarantee inertial availability is to reserve trading headroom — hold, say, 10% of nominal power out of the market at all times. On representative numbers, that trade is terrible: a megawatt might earn on the order of €20k per year in an inertia contract versus an order of magnitude more in wholesale and ancillary markets. Overload capability breaks the dilemma. A plant with a c. 25% five-second boost factor is always available for inertia while trading its full nominal rating, converting the inertia contract into nearly pure incremental revenue against modest, largely opex-side costs (typically software licensing).

8.2 Representative Economics

On a representative 100 MW / 4-hour project in an availability-based inertia market, three revenue tiers emerge. Grid-code-driven unintentional oversizing alone (reactive power margin) supports roughly €3k/MW-year of inertia revenue. Deliberate 25% overload capability lifts that to a headline figure of roughly €10k/MW-year. A third strategy — layering overload on top of reserved battery headroom — models well in backtests (optimizer studies suggest a quarter of plant capacity could historically meet 90% availability) but is dangerous over a ten-year horizon: as ancillary markets saturate and merchant strategies shift toward wholesale trading, the top megawatts get used far more often, and a decade-long availability commitment against them becomes a genuinely risky position.

At the ~€10k/MW-year level, the project-level effects are material: approximately one percentage point of IRR uplift — decisive for projects squeezed toward single-digit returns by restrictive flexible-connection agreements — and, because lenders prize long-term contracted revenue, a debt-sizing benefit. A ten-year inertia contract at that level can service roughly 7% of capex on its own, unlocking around three percentage points of additional gearing. Combined with tolling structures, the contracted layer meaningfully reshapes the capital stack.

8.3 Bankability and the 90% Cliff

The availability construct is where financing risk concentrates: contracts with a sharp payment drop below 90% availability put a cliff under the revenue line. Three facts make the risk manageable. First, real fleet data shows batteries averaging around 95% availability — measured market data, not a projection. Second, unavailability is front-loaded: commissioning-era teething problems dominate, and availability improves as plants mature. Third, contract design can absorb the risk — run-up periods that start the inertia commitment only once the plant has demonstrated stable availability, vendor warranty alignment, and conservative initial commitment levels. Lenders already underwrite availability risk in every tolling agreement; the inertia cliff is a new shape, not a new species.

9. Standards and What Comes Next

Component-level guidance is still catching up to grid-code ambition. Almost everything published to date regulates behavior at the point of interconnection; it does not tell an engineer how to design the plant behind it or type-test individual units. IEEE 2800 reaches deepest into system and equipment selection on the North American side, though its center of gravity remains the PCS. In Europe, national connection-rule documents are beginning to blur the line between grid requirement and plant design guidance, a formal European standard on grid-forming capability is in preparation, and the pan-European framework report gives national operators a common technical foundation that each must now transpose into binding code. On the DC side, no standard yet substitutes for direct engagement: cell and module suppliers must be told explicitly that their product will see grid-forming duty — short-duration peak discharge across a wide SOC window — and must confirm, in writing, that they are comfortable with it.



The direction of travel is unmistakable. Requirements that today apply only to BESS will migrate toward PV, wind, and eventually large converter-based loads. Markets that today mandate grid-forming without quantities will begin quantifying — and paying — as operators discover that mandate-only regimes deliver minimum-compliance plants with little usable inertia. Locational signals will sharpen. And the plants positioned to benefit are the ones engineered from day one with overload capability, validated EMT models, and topologies that can actually deliver what the controls promise.

10. How Keentel Engineering Supports Grid-Forming Projects

Keentel Engineering LLC provides independent, vendor-neutral engineering across the full grid-forming project lifecycle:

• Grid Interconnection & Compliance. Interconnection studies, grid-code compliance matrices, and requirement interpretation across U.S. ISO/RTO and international frameworks.
• EMT Modeling & Studies. EMT model development, benchmarking, and validation for grid-forming and grid-following IBRs; PSCAD and PSS/E study execution; HIL test planning and witness support.
• Plant Design & Sizing. PQ envelope and overload sizing against target inertia bids, PCS topology evaluation, retrofit feasibility screening (including shared-winding oscillation risk), protection coordination, and transformer/cable adequacy under overload duty.
• Owner's Engineer Services. Independent review of vendor grid-forming claims, factory and site test witnessing, commissioning support, and NERC compliance program development for IBR fleets.
• Business Case Support. Inertia revenue strategy assessment, availability risk analysis, and technical support for financing due diligence.
  

Frequently Asked Questions 

About Keentel Engineering 

Keentel Engineering LLC is a power systems and grid interconnection consulting firm headquartered in  Tampa, Florida, with offices in Austin, Sacramento, and Baltimore. Our practice spans POI and grid  interconnection engineering, power system studies at EHV/HV/MV, substation and transmission design,  EMT modeling, utility-scale renewables and BESS engineering, NERC compliance, and owner’s engineer  services. Deliverables are prepared under the responsible charge of licensed Professional Engineers.