Black start

What it is (precise)

After a partial or total blackout, the network is rebuilt in stages: a black-start resource energizes a section of dead bus, holds a stable voltage and frequency reference on that island, then sequentially supplies cranking power for larger generating units and progressively connects load until the island can be resynchronized to the bulk grid. The defining requirement is that the resource needs no external supply to begin — it carries its own stored energy and its own voltage and frequency reference, and the start must be self-contained down to the auxiliary systems that wake the plant up.

For a BESS, the energy is already in the racks and the reference comes from the PCS. A Grid-following inverter cannot do this, because it needs an existing grid voltage to lock onto. A Grid-forming inverter, by contrast, behaves as a controllable voltage source behind an impedance, so it can build an islanded grid from a dead start, supply limited inrush and fault current — enough, combined with soft energization, to bring transformers and motors online — and accept the inherently unbalanced, lumpy load steps of a restoration sequence.

Grid-forming inverters are also particularly good at soft energization, which they perform natively and precisely in firmware; synchronous black-start units can approximate it by ramping generator excitation while connected to the dead network, but the inverter's voltage control is finer and faster.

Instead of closing a breaker onto a dead Transformer and absorbing a magnetizing inrush of typically 6 to 12 times rated current, the inverter ramps the island voltage from zero to nominal over several seconds, magnetizing every connected transformer and cable gradually with almost no inrush. This lets a comparatively small battery energize network sections that would otherwise demand a much larger rotating machine, and it is one of the strongest technical arguments for BESS-based restoration.

Why it matters in a real grid-scale project

Black start is a system-restoration service that grid operators procure under contract, so it is a revenue stream — usually availability payments plus tested-capability terms — and sometimes a connection requirement rather than an optional extra. A BESS that can credibly provide it stacks black-start and island-forming capability on top of energy arbitrage and Frequency response, improving the commercial case at the Point of Interconnection without consuming many cycles, since the service is exercised rarely but paid for continuously.

The engineering consequence is significant and is decided early, because it touches the PCS firmware and controls (true grid-forming control, not merely GFM-capable hardware), the protection scheme, and the energy reservation policy. Restoration can take hours to days, so a defined block of state-of-charge must be ring-fenced and unavailable for trading.

Energizing transformers and starting large motors demands high inrush and fault current that the inverters and DC system must be rated to deliver, and the controls must ride through the resulting voltage and frequency excursions without tripping.

The service has also moved from pilot to procurement. A California BESS black-started a gas turbine in 2017 in one of the first utility-scale demonstrations, the UK's Distributed ReStart programme trialled restoration from distribution-connected resources, and system operators in Great Britain and Australia have since awarded restoration contracts to large grid-forming battery plants. For developers, that history matters: operators now have reference projects and expect bidders to demonstrate comparable, study-backed capability rather than paper claims.

Typical values and standards

Useful anchors from real projects: grid-forming fault-current contribution is typically about 1.1 to 1.5 times rated current for short durations, set by semiconductor limits rather than machine physics — roughly a quarter to a fifth of a synchronous machine's 5 to 7 times. Direct transformer energization draws roughly 6 to 12 times rated current, which soft energization largely avoids.

Cranking power for a thermal unit's auxiliaries is commonly around 5 to 10 percent of that unit's rating, so a 100 MW-class BESS can plausibly crank a much larger plant. Reserved energy is contract-specific, but restoration plans assume hours of islanded operation before resynchronization.

There is no single global black-start standard; the performance specification lives in the system operator's framework. In Europe that is the ENTSO-E network code on Emergency and Restoration; in North America, NERC EOP-005 governs system restoration from black-start resources; Great Britain adds an Electricity System Restoration Standard with regional restoration milestones measured in tens of hours.

IEEE 2800 sets general performance for transmission-connected inverter-based resources but does not itself define black start, while EPRI, NERC and ESIG guidance is converging on what grid-forming behavior must demonstrate: voltage-source response, fault-current contribution, and stable islanded operation.

Black start adds no new safety-code obligations: the plant simply carries the standard stationary-storage regime — the UL 9540 ESS safety certification (the listing NFPA 855 requires), UL 9540A fire-propagation test data feeding NFPA 855 compliance, and NFPA 68/69 for deflagration protection. The black-start-specific numbers live in the tender instead.

GB restoration tenders have typically required a resource to energize its first dead network section within a couple of hours of instruction and to sustain the island for on the order of ten hours, or to deliver several sequential start attempts; Australian (AEMO) system-restart contracts impose comparable start-time and endurance obligations, with the exact figures set per region.

How it shows up in specs, studies and contracts

On a PCS datasheet, look past the marketing line "grid-forming capable." Check whether islanded and dead-bus-energization operation is a released firmware function, what overload and fault current the unit delivers and for how long (magnitude and duration both matter to protection), how much unbalanced and non-linear load it tolerates, and its Harmonics performance while energizing cables and transformers. Four-quadrant operation matters because restoring lightly loaded cable networks means absorbing capacitive reactive power, not just injecting.

In studies, black start is an EMT-level exercise, not a load-flow one. Expect electromagnetic-transient simulation of the full restoration sequence: energization of each transformer and cable section, motor starts, load-block pickup, and resynchronization at the Point of Interconnection. Protection deserves special attention — an inverter island produces far less fault current than the bulk grid, so overcurrent relays graded for normal grid-connected conditions may simply not see faults, forcing voltage-based or differential schemes for the restoration path.

Contractually, ask what availability percentage is required and how it is tested (periodic proving tests, commonly somewhere between annual and every few years), how much energy must be held in reserve and how that interacts with trading, how long the island must be sustained, and who supplies the plant's own auxiliary power when the grid is dead. Payment is typically an availability fee with penalties for failed tests, so the warranty and control-software commitments behind the capability carry real money.

Common pitfalls

The most common trap is conflating GFM-capable hardware with a validated black-start function. Many modern inverters can run grid-forming control while connected to a live grid, yet have never demonstrated dead-bus energization, load-step rejection (riding through the sudden loss of a picked-up load block), or hours of autonomous islanded operation.

A related one is the auxiliary-power blind spot: HVAC, controls, and battery management keep drawing power during a blackout, so the site needs a self-supply path (UPS plus self-feed from the DC side, or a small generator) to remain start-ready — a plant whose auxiliaries die with the grid cannot black start.

Two more trip-wires: energy accounting and event survival. Ring-fenced restoration energy is usable energy at the delivery point, net of conversion losses and the usable SOC window, not nameplate — and it must still be there after weeks of standby. And during restoration the island will see voltage and frequency excursions far outside normal bands, so Ride-through settings tuned for bulk-grid compliance can trip the plant off exactly when it is the only source; restoration control modes need their own, wider protection envelope.