What is a hybrid inverter and how does it work?

On-grid, off-grid and hybrid inverters

An on-grid inverter follows the utility waveform and normally stops exporting when the grid fails. An off-grid inverter forms an AC supply for isolated loads. A hybrid inverter coordinates solar and storage, and can switch from grid following to grid forming when its hardware and firmware support backup.

These categories describe functions. A solar power conditioning unit can accept utility AC through a charger or a bypass relay while remaining incapable of synchronised export. A grid-interactive hybrid may export but power only its emergency power supply (EPS) terminal after a failure. A storage retrofit can use a separate battery inverter with an existing PV inverter, provided the system supports the intended AC-coupled operating mode. An older grid-tie inverter cannot be assumed to keep producing behind an arbitrary backup source.

A purchase specification should distinguish PV start without a battery, continuous PV-plus-grid sharing, synchronised utility export and backup with the grid physically isolated. A product may pass one test and lack another capability. Also distinguish an inverter’s AC input or bypass voltage window, its grid-tie protection window and its regulated EPS output range. These are separate operating limits.

Inside the inverter

A common DC-coupled design uses one or more MPPT-controlled DC–DC stages, a bidirectional battery converter, a DC link, and a PWM bridge with an output filter. Grid sensing and transfer protection govern which AC terminals can be energised. Component choices and electrical isolation differ by model.

PV strings → MPPT DC–DC ──→ DC link ↔ PWM bridge + filter ↔ AC bus → protected loads
                              ↕                              ↕
Battery + BMS ↔ bidirectional DC–DC                     interconnection relay
                                                             ↕
                                                       CT/meter + utility

MPPT is a control function implemented through a switching stage. It samples string voltage and current and adjusts operating voltage to maximise their product. Perturb-and-observe changes the setpoint and infers direction from the power change. Incremental conductance compares the local current slope with the maximum-power condition dI/dV = −I/V. The converter may switch at tens of kilohertz or higher while the tracking algorithm updates on a slower timescale. A shadowed string can have several local peaks because of bypass diodes. Global-search algorithms and independent trackers can help, but each input retains its own voltage and current limits.

A reversible battery stage typically raises battery voltage to feed the DC link in discharge and steps it down while charging. A non-isolated synchronous buck–boost uses controlled switches and an inductor. A transformer-based dual-active-bridge is an example of an isolated bidirectional stage. A battery management system (BMS) can command voltage and current limits; the converter must still enforce hardware limits, precharge the DC link as designed and safely stop or ramp down when the battery becomes unavailable.

A single-phase full H-bridge uses four switch positions to synthesise an alternating voltage by pulse-width modulation. A three-phase implementation commonly uses three half-bridge legs, with other multilevel arrangements possible. The inductor-capacitor filter attenuates switching components, and a feedback loop controls AC current or voltage depending on the mode. IGBTs can suit particular voltage and switching regimes. SiC MOSFETs can reduce switching losses in some designs, but their faster edges make gate drive, common-mode current and electromagnetic compatibility important. A “pure sine wave” label alone gives no test conditions for voltage or current distortion.

Galvanic isolation separates circuits so there is no direct conductive path across the barrier. An inverter may isolate only its battery stage, only control signals, or neither power input. An isolation transformer, protective earth, residual current monitoring and anti-islanding solve different problems. A transformerless design requires the protective scheme specified by its manual and approved site design.

MPPT efficiency, voltage and temperature

MPPT tracking efficiency measures how much available PV energy the controller captures. It differs from DC-to-AC conversion efficiency. String design must keep cold open-circuit voltage below the absolute DC limit, hot operating voltage inside the MPPT window and current within each tracker’s rating.

A claim such as 99.5% tracking efficiency must identify the method and irradiance profiles. IEC 62891 treats static and dynamic tracking separately. Over a measurement period, ηMPPT = ∫Poperating(t)dt / ∫PMPP,reference(t)dt. At steady state a 99.5% tracker and a 97% conversion path would yield about 96.5% from available maximum-point power to AC for that operating point, before any external wiring losses. Dynamic tracking under rapidly moving clouds can give a different result. IEC 62891

A range such as 90–500 V DC is an example of a possible tracker window, with startup voltage, full-power range and absolute maximum DC voltage listed separately. The Nexus Solar Energy DEV HYD 6 kW 48 V IP54 library entry gives 60–450 V tracking and 500 V maximum PV open-circuit voltage. The Devsol HN15KT-AH2GT manual lists a much wider 120–950 V window and a 1,000 V maximum input. These windows are model specifications, with no implication that either delivers full output at the endpoints. Nexus Solar Energy technical library

For N modules in series, with signed coefficients in fraction per °C, a first-order check is Voc,string(T) = N Voc,STC[1 + βoc(Tcell − 25°C)]. Use the analogous expression with Vmp,STC and βmp for operating voltage. Use the exact module coefficients, site design temperatures, tolerance, applicable design margins and inverter instructions. STC values correspond to 25°C cell temperature. Cells on a hot rooftop can run considerably hotter than the air.

For illustration, eight modules with 49 V open circuit, 41 V at maximum power, β_oc = −0.28%/°C and β_mp = −0.33%/°C give about 419 V open circuit at 0°C cell temperature and 279 V at maximum power at 70°C cell temperature. At 70°C the operating voltage is roughly 15% below its 328 V STC value. A ninth module would increase both voltages; the cold check must be repeated with margin. Series strings add voltage, whereas parallel strings add current. Bifacial gain and high irradiance can raise input current, and low irradiance can affect start and tracking.

Power balance and motor starting

PV, grid import and battery discharge supply the load, battery charging, export and losses. Motor starting is constrained by RMS current, apparent power, source impedance, duration and voltage sag. A kW rating on its own cannot show that a compressor or pump will start.

Take all power terms as positive in their named directions and use one consistent system boundary. Then PPV + Pimport + Pdischarge = Pload + Pexport + Pcharge + Ploss + dEbuffer/dt. The last term accounts for changes in short-term stored energy such as the DC-link capacitor. Over a steady interval it averages near zero. Curtailment appears as lower actual PV input relative to available PV, not as a dissipated loss. Grid import and export in this equation refer to net power at the metering boundary.

Single-phase apparent power is S = VrmsIrms. For a balanced three-phase load use S = √3 VLLIL. For a 230 V compressor drawing 6 A in normal operation and 18 A on start, an illustrative 3× current case gives 4.14 kVA during start. If other coincident loads draw about 1.2 kVA, a conservative arithmetic budget is about 5.34 kVA before voltage sag and margins. Check each phase separately where the loads are uneven. Actual locked-rotor amperage can be higher than 3×. Variable-speed compressors can behave differently from direct-on-line motors.

Where a load schedule assumes 3× for 5–10 seconds, require a published overload curve or a measured start trace for that entire interval. A brief peak rating may last only a fraction of that time. Power factor changes during start, so battery current cannot be inferred from kVA alone. Include battery BMS peak-current duration, cable voltage drop, concurrent loads and the difference between grid bypass and islanded EPS performance. An inverter that starts a motor on bypass may trip on the same motor in backup mode.

Five operating modes and what changes in each

The controller changes power direction and limits according to PV availability, battery state, load, grid status and export permission. Battery-less, self-consumption, zero-export, scheduled charging and backup modes have distinct prerequisites. Several functions can operate together, subject to protection and equipment limits.

1. Battery-less daytime mode

First the inverter verifies the PV startup conditions and detects any utility supply it requires. It then tracks the solar array and delivers available PV power to AC loads. In a model with actual grid blending, the utility supplies the deficit when sunlight falls or a load starts. In other designs a relay transfers the entire load to the AC source. If grid and battery are absent, sustained operation requires an explicitly supported PV-only island mode, adequate irradiance and enough instantaneous power to handle transients. With no storage, shade or sunset can interrupt supply.

2. Solar with battery priority

The dispatch controller may serve load first, then charge the battery with surplus PV, and finally export permitted excess or curtail the array. If PV falls short, it discharges the battery down to its configured reserve unless a tariff rule, BMS limit or backup policy takes precedence. BMS constraints can reduce available charge power abruptly near a temperature or cell-voltage limit. A higher reserve leaves less capacity for shifting daytime solar to evening use.

3. Metered zero export

An external CT or compatible meter measures net power at the point of common coupling. The inverter reads the signed measurement, compares it with the export limit, then reduces generation, charges storage where possible or maintains a small import bias. When a load suddenly disconnects, meter filtering, communications latency and inverter slew rate can permit brief export. A claim of less than 50 ms must specify whether it measures sampling, initial response or settling, and under which load step. Zero export depends on meter placement and polarity, phase mapping, communications fail-safe behaviour and utility requirements.

4. Time-of-use charging and peak shaving

A schedule can allow battery charging when energy is cheaper and discharge during expensive intervals or above a target demand. Before dispatch, the controller checks reserve, charge and discharge limits, grid permissions and local time. If the tariff bills maximum demand over an interval, the controller must target that interval and avoid creating a new peak while charging. For example, delivering 3 kWh through a storage path at 85% round-trip efficiency requires about 3.53 kWh of charging energy. The tariff spread must cover those losses and any cycling cost.

5. Backup and intentional islanding

Upon loss of utility, a backup-capable system isolates the permitted loads from the public grid, establishes local voltage and frequency, and supplies those loads from supported sources. It later resynchronises or transfers back under the approved conditions. A stated transfer under 4 ms or within 10 ms is a model-dependent test result or specification with a defined load and operating state. The Nexus Solar Energy library publishes below 10 ms for a named 12 kW three-phase entry. Even a 10 ms break may reset some IT or control equipment. Medical and other critical loads require their own qualified backup design, transfer validation and applicable approvals. Nexus Solar Energy technical library

Single-phase or three-phase supply

Single-phase hybrid inverters suit compatible 230 V circuits and defined essential loads. Three-phase units suit three-phase equipment and sites requiring coordinated 400 V-class supply. Selection depends on sanctioned connection, phase-by-phase load and backup limits, neutral arrangement and permissible imbalance.

At unity power factor, 6 kW at 230 V corresponds to roughly 26 A on one phase. A balanced 15 kW load at 400 V line-to-line is about 21.7 A per phase. An installation with one phase carrying most of its household load can hit a phase limit even when the site total is below the advertised rating. Ask for the model’s asymmetric output capability separately in grid-connected and EPS modes. Specify backed-up circuits and check the protection and earthing design with the installer.

Verified Hardware Test Data (YouTube Lab Series)

These first-party videos document the setup and readings shown in each recording. The summaries distinguish demonstrations, stated ratings and sizing examples. View counts are dated YouTube snapshots.

Frequently asked questions