Grid-Forming Inverter Case Study

This case study follows a realistic microgrid upgrade: a critical-water campus wants a battery inverter to support islanded operation, reduce diesel runtime, and allow onsite photovoltaic generation to remain online during grid outages. The inverter is advertised as grid-forming, but the engineering team must prove what that means in the actual network.

The case is not about whether grid-forming inverters are useful. They are. The case is about the difference between a control capability and a validated power-system function. A grid-forming inverter must hold voltage and frequency, share load, respect current limits, coordinate with protection, manage battery state of charge, and recover from abnormal events inside a specific microgrid.

Case Summary

Item	Engineering relevance
Site	Water treatment, emergency operations, communication, and shelter loads on a campus feeder.
Objective	Island the site for short outages and reduce diesel operation during longer outages.
Main asset	5.0\ \text{MVA} grid-forming battery inverter with 12\ \text{MWh} battery.
Other resources	3.0\ \text{MW} PV plant, 2.5\ \text{MVA} diesel generator, controllable loads.
Dominant risks	Motor starting, low fault current, SOC reserve, PV curtailment, protection settings, black-start sequence.
Useful outcome	A validated operating envelope, not a generic claim that the inverter is grid-forming.

The central engineering question is:

Can the inverter form a stable island while preserving battery reserve, coordinating PV, starting critical motors, and clearing faults safely?

The answer depends on system studies and field tests, not on the inverter label alone.

Initial Architecture

The site contains a medium-voltage campus feeder with water-treatment loads, operations buildings, communications equipment, emergency shelter loads, and photovoltaic generation. Before the upgrade, the site can run selected loads from a diesel generator, but transitions are manual and PV is normally disconnected during islanded operation.

The proposed upgrade adds:

• a battery energy storage system with a 5.0\ \text{MVA} inverter;
• grid-forming control mode for islanded operation;
• a microgrid controller;
• transfer and isolation control at the point of common coupling;
• PV curtailment and ride-through coordination;
• updated relays and breaker settings;
• operating modes for grid-connected, planned island, unplanned island, black start, and resynchronization.

The owner expects three benefits: faster transition during outages, lower diesel runtime, and better use of onsite PV. The engineering team reframes those expectations into testable requirements.

Operating Requirements

The microgrid requirements are:

1. support 2.2\ \text{MW} of critical load for at least 4\ \text{h} without utility supply;
2. maintain voltage and frequency inside site limits during normal load steps;
3. start one 450\ \text{kW} pump without collapsing voltage;
4. keep PV online when stable, but curtail it when battery SOC or voltage limits require;
5. preserve 2.4\ \text{MWh} minimum SOC energy and 1.2\ \text{MWh} restart reserve;
6. clear feeder faults selectively in both grid-connected and islanded modes;
7. black-start selected loads from a de-energized island;
8. resynchronize to the utility only under approved voltage, frequency, phase, and protection conditions.

These requirements deliberately mix energy, power, controls, protection, and operation. Grid-forming behavior cannot be evaluated from steady-state energy capacity alone.

Steady Apparent-Power Check

During islanded operation, the expected critical load is 2.2\ \text{MW} with 0.9\ \text{MVAr} reactive demand before power-factor correction. The inverter apparent-power requirement is:

This is below the 5.0\ \text{MVA} inverter rating. The steady load therefore appears feasible.

This result is necessary but not sufficient. The inverter must also handle transient load steps, current limits, thermal limits, PV interaction, and protection behavior. A grid-forming inverter can be underloaded in steady state and still fail a motor-start or fault-clearing case.

Battery Reserve Check

The battery has 12.0\ \text{MWh} current usable DC energy. The outage watch operating policy requires 80\% SOC before a forecast event:

Protected energy is:

Energy available for the islanded event is:

If the critical load averages 2.2\ \text{MW} for 4\ \text{h} and PV is not credited, delivered load energy is:

With discharge efficiency of 94\%, required battery-side energy is:

The battery cannot support the full critical load for four hours without PV, diesel, or load shedding. This is the first major design correction. The inverter can form the grid, but the energy source behind it cannot satisfy the resilience objective alone.

Revised Source Strategy

The team revises the islanded operating strategy:

• the battery inverter forms voltage and frequency;
• PV follows the island voltage and contributes when available;
• the diesel generator starts if SOC falls below a defined threshold or PV forecast is poor;
• noncritical water-processing loads are shed during the first islanded minutes;
• the 450\ \text{kW} pump starts only after the island is stable;
• PV is curtailed when load is low and battery SOC is high;
• restart reserve remains unavailable for normal dispatch.

The revised design treats the grid-forming inverter as the reference source, not the only energy source.

PV Coordination

The PV plant uses grid-following inverters. In islanded mode, those inverters need a stable voltage and frequency reference. The battery inverter can provide that reference, but only within its current, voltage, frequency, and DC-link limits.

PV coordination requires:

1. ride-through settings compatible with microgrid transitions;
2. active-power curtailment command or frequency-watt response;
3. reactive-power settings that do not fight the battery inverter;
4. anti-islanding settings appropriate for intentional islanded operation;
5. behavior defined for low load and high PV output;
6. reconnection delays after trips;
7. telemetry showing whether PV is following commands.

The team discovers that PV curtailment speed matters. If cloud cover clears while load is low and the battery is near high SOC, the microgrid must curtail PV before frequency rises or the battery hits charge limits.

Motor-Start Event

The critical pump is rated 450\ \text{kW}. With a soft starter, the expected starting apparent power is 1.6\ \text{MVA} for several seconds. During the start, the island also carries 1.8\ \text{MVA} of other load.

A first apparent-power screen is:

This is below the 5.0\ \text{MVA} inverter rating, but the event is still not automatically acceptable. The inverter current limit, voltage sag, control bandwidth, motor torque requirement, and other loads must be tested together.

The field test initially shows an unacceptable voltage dip when the pump starts immediately after islanding. The correction is operational and control-based:

• wait until the island voltage and frequency settle;
• block simultaneous large load starts;
• pre-curtail noncritical loads;
• confirm battery SOC and temperature margin;
• use a soft-start profile with verified acceleration time;
• alarm if the pump fails to accelerate within the allowed window.

The lesson is that grid-forming controls do not remove the need for motor-start engineering.

Fault Current and Protection Problem

Protection is the hardest issue in the case. In grid-connected mode, the utility source provides high fault current. In islanded mode, the inverter limits current quickly to protect power electronics. A legacy overcurrent relay that worked with the utility source may not trip fast or selectively with inverter-limited fault current.

The original feeder protection expects high fault current for instantaneous operation. The grid-forming inverter can provide only limited current for a short duration. The team therefore cannot assume that existing protection will clear islanded faults correctly.

The protection revision includes:

• mode-dependent relay settings;
• differential or zone protection for selected feeders;
• breaker trip logic coordinated with inverter fault behavior;
• inverter blocking and ride-through rules;
• grounding review for islanded operation;
• test cases for line-to-ground and phase faults;
• clear fallback state if protection state is uncertain.

The practical lesson is direct: an inverter can form voltage and frequency without providing synchronous-machine fault current. Protection must be engineered for that reality.

Control Hierarchy

The microgrid uses a layered control hierarchy:

1. local inverter controls regulate voltage and frequency;
2. battery management protects cells, SOC, temperature, and DC limits;
3. PV controllers follow voltage and curtail active power when commanded;
4. relays and breakers protect feeders and isolate faults;
5. the microgrid controller schedules sources and loads;
6. operators supervise modes and authorize maintenance or recovery.

The hierarchy avoids giving one supervisory controller unsafe authority. If communication to the microgrid controller fails, local protection and inverter safety functions must still work. If the battery management system declares a limit, the dispatch controller must respect it even if the grid service is valuable.

Black-Start Sequence

The team defines a black-start sequence for a fully de-energized island:

1. verify utility isolation and safe switching state;
2. energize battery auxiliary systems;
3. start the grid-forming inverter in voltage-source mode;
4. energize the critical control and communication bus;
5. energize protection, metering, and supervisory controls;
6. pick up the first small load block;
7. connect PV only after voltage and frequency are stable;
8. start larger loads in approved sequence;
9. start diesel if SOC, PV forecast, or load requires it;
10. record voltage, frequency, SOC, alarms, and operator actions.

Each step has a hold point. The sequence stops if voltage, frequency, SOC, inverter temperature, relay state, or communication state is outside its approved limit.

Black start is not proven by energizing an empty bus. It is proven by energizing the loads, controls, and source mix needed for the defined mission.

Resynchronization

Returning to grid-connected mode is also a controlled event. The microgrid must match voltage magnitude, frequency, phase angle, and protection state before closing the point of common coupling.

Useful permissives include:

• voltage difference inside limit;
• frequency difference inside limit;
• phase-angle difference inside limit;
• utility source healthy;
• island stable for a defined time;
• no active protection lockout;
• battery and PV controllers in reconnection-ready mode;
• operator or automation authority confirmed.

The team adds a post-resynchronization recovery rule: the battery must restore its resilience SOC before normal economic dispatch resumes. Otherwise the site may reconnect successfully but remain unprepared for a second outage.

Validation Test Matrix

The final commissioning plan includes:

Test	Evidence required
Planned islanding	Voltage, frequency, phase, load, SOC, and breaker records remain inside limits.
Unplanned islanding simulation	Critical loads stay energized or shed according to priority.
Load step	Frequency and voltage recover within accepted time.
Pump start	Motor accelerates without unacceptable voltage sag or inverter current trip.
PV curtailment	PV follows active-power limit and does not overcharge the battery.
Islanded fault	Protection clears the fault selectively or enters defined safe state.
Low-SOC operation	Controller starts diesel or sheds load before reserve is consumed.
Communication loss	Local controls maintain safe voltage, frequency, and protection behavior.
Black start	Defined load blocks energize in sequence from de-energized state.
Resynchronization	Reclosing occurs only under approved synchronizing conditions.

The team also requires event records, not only pass/fail statements. Records must include time-synchronized voltage, frequency, active power, reactive power, SOC, inverter current, breaker status, relay targets, PV output, alarms, and operator actions.

Final Operating Envelope

After studies and tests, the approved operating envelope is narrower than the original sales concept:

• the battery inverter may form the island for defined load blocks;
• full four-hour critical-load support requires PV contribution, diesel start, or staged load shedding;
• large motor starts are sequenced and blocked during unstable states;
• PV must be curtailment-ready before islanded operation with high solar output;
• islanded protection uses revised settings and selected differential protection;
• economic dispatch cannot reduce SOC below the active resilience floor;
• black start is allowed only with the tested load sequence;
• resynchronization requires automatic permissives and operator confirmation.

This is a better engineering result than a vague claim of “grid-forming capability.” The site now knows what the system can do, what it cannot do, and what evidence supports each claim.

Transfer Lessons

Several lessons transfer to other microgrids and weak-grid projects:

1. Grid-forming is a system behavior, not only an inverter feature.
2. Energy reserve can be the limiting constraint even when inverter MVA is adequate.
3. PV coordination must be designed for islanded operation, not assumed from grid-connected performance.
4. Motor starting needs transient testing, not only steady-state power checks.
5. Inverter-limited fault current can invalidate legacy overcurrent protection assumptions.
6. Black start and resynchronization are distinct operating modes with separate evidence.
7. Dispatch value is limited by SOC, thermal state, protection, firmware, and validation.

The case shows why microgrid design is multidisciplinary. Power electronics, protection, controls, energy storage, operations, and human procedures all determine whether an islanded system actually works.

Common Mistakes

A common mistake is treating a grid-forming inverter as a small synchronous generator. It can establish voltage and frequency, but its current limit, fault response, overload duration, and control dynamics are different.

Another mistake is proving a microgrid at light load and then assuming the result applies to motor starts, high PV output, low SOC, or protection faults. Commissioning must test the stressful operating states, not only the easy ones.

A deeper mistake is separating energy management from protection. If the battery is low, if PV trips, or if a feeder fault is not cleared selectively, the island can collapse even though every component is individually functional. The validated operating envelope must include source state, load state, protection state, and recovery state together.

Grid-forming inverters can make resilient inverter-based microgrids possible. They do not remove the need for power-system engineering; they make that engineering more explicit.