Huawei alarm 2034 «Grid Overvoltage»: when only one phase is high (real case)

Alarm 2034 «Grid Overvoltage» is one of the most frequent alarms in FusionSolar and one of the most misread. The manufacturer's text talks about a grid with high voltage and suggests, if it keeps happening, asking the grid operator for permission to raise the threshold. But there is one case where that would be a mistake: when only one phase is high. Then the problem is not the grid, it is the neutral.

This article uses a real case: a 15 kW plant on a farm in central Spain, with a three-phase Huawei SUN2000-15KTL-M5 inverter, that has not produced for three days. The data come from FusionSolar, read through Suntropy Connect, which is where Alexandria gets them. We removed the name and exact location; voltages, times and production are the original ones.

What alarm 2034 means

It is the grid overvoltage protection: the inverter measured on its AC output a voltage above the threshold of the configured grid code, or a high voltage for longer than allowed, and stopped exporting. On European 230/400 V grids the usual settings are around 253 V as a 10-minute average (230 V + 10 %) and around 264 V as a fast trip, but the exact values depend on the grid code loaded in the inverter (each country has its own).

The manufacturer's recommendation, which comes with the alarm itself:

• If it happens occasionally, the grid had a passing spike: the inverter reconnects by itself once the voltage is back to normal.
• If it keeps happening, check whether the grid voltage is within the allowed range. If it is not, contact the grid operator; if it is, adjust the threshold with their consent.

What the alarm does not tell you is which phase or why. For that you need the readings.

The case: from 90 kWh a day to zero

Daily production of the plant, in kWh:

The drop is abrupt: not a cloud, not soiling, but an inverter that stops working. Between 4 and 10 November the plant lost more than 500 kWh.

Alarm 2034 was raised on Tuesday the 10th at 8:25. But the readings tell a story that started earlier.

What the voltages say

Some inverter readings from Monday the 9th and Tuesday the 10th (local time):

Frequency, in all of them, between 49.97 and 50.03 Hz. And power, zero.

1. The grid is not high. Line-to-line voltages are between 404 and 414 V, perfectly normal on a 400 V grid. If the grid operator were delivering too much voltage, all three line voltages would rise, not one phase.

2. What is wrong is the split. Two phases are slightly low (218-225 V) and one is very high (260-273 V). The average of the three, about 236 V, matches what 407 V line-to-line implies (407 / √3 ≈ 235 V). In other words, the voltage triangle is fine but the neutral point has shifted towards phases A and B. That is the typical fingerprint of a loose, high-impedance or broken neutral.

3. The inverter does what it should. With phase C above 260 V, its protection will not let it connect. On Monday it spent the whole morning trying to start without exporting; at 12:05 phase C went above 265 V and from 12:10 the inverter stayed shut down on fault and stopped trying. The 2034 alarm raised on Tuesday is just the last of that series.

The diagnosis

The hypotheses, from most to least likely:

1. Faulty neutral inside the installation: a loose terminal in the main panel, in the PV protection or on the inverter's own N terminal. With unequal single-phase loads, a neutral that does not conduct well lets the neutral point «float», and one phase rises while the others drop.
2. Faulty neutral upstream: in the service connection, the main fuse box or the grid operator's low-voltage network. It is more common in rural areas, with long lines and many single-phase loads. If neighbours on the same line notice lights that are too bright or equipment that switches off, that points here.
3. Measurement skewed by the inverter's N terminal: if the neutral only fails at the inverter connection, the rest of the installation would be fine and only the inverter would see the imbalance. Rule it out by measuring phase-to-neutral at the main panel.

What it is not: a grid with high voltage. That is why you should not ask the grid operator to raise the overvoltage threshold: the inverter would stay stopped and the problem would remain.

Why it is urgent

A shifted neutral does not only stop the PV system. Single-phase loads on the high phase (lighting, electronics, small motors) are getting more than 260 V and can fail. On a farm, with pumps and control equipment, the damage can cost more than the lost production.

What to check on site

• Measure phase-to-neutral on all three phases at the main panel, with the PV system disconnected. If one phase is still above 250 V, the problem is upstream of the inverter.
• Measure neutral-to-earth voltage. With a healthy neutral it should be a few volts; several tens of volts confirm the shift.
• Check the tightness of the neutral terminals in the panel, the PV protection and the inverter, and look for signs of overheating.
• If the panel is fine, report the phase-to-neutral imbalance to the grid operator with the readings of the last few days: it is a fault on their network, not an inverter setting.

And before you go, one question for the customer: «Since the 4th, have you noticed flickering lights, equipment restarting or anything breaking down?» If the answer is yes, the visit is for today.

Telling the cases apart at a glance

If you are an AI agent: how to repeat this diagnosis

1. With list_alarms, find alarm 2034 and its plantId. With list_devices, the inverter id (kind: inverter).
2. Call get_plant_energy with metric: pv_yield and granularity: day for the last two weeks to date the start of the problem.
3. Call get_device_history for the inverter on the day of the alarm and the day before. On Huawei, phase-to-neutral voltages are a_u, b_u and c_u; line voltages, ab_u, bc_u and ca_u; frequency, elec_freq; state, inverter_state (768 is shut down on fault).
4. Always compare phase-to-neutral with line-to-line. If line voltages are between 390 and 420 V and one phase is more than 20 V away from the average, say it is a shifted neutral and that the threshold must not be raised.
5. Warn about the risk to the loads on the high phase and propose the on-site measurements.

Frequently asked questions

What does alarm 2034 mean on a Huawei inverter?

Grid overvoltage: the inverter measured a grid voltage above its protection threshold, or high for longer than allowed, and disconnected. If it is a one-off, it reconnects by itself.

Is the Grid Overvoltage alarm serious?

On its own, no. It is serious if it repeats every day or if the plant stops producing. And it is urgent if only one phase is high, because that points to a neutral problem that also affects the rest of the installation.

Can I raise the inverter's overvoltage threshold?

Only with the grid operator's consent and only if the grid really is within limits. If the cause is a shifted neutral, raising the threshold does not fix it and leaves the equipment exposed.

How do I know whether the problem is the grid or my installation?

Measure phase-to-neutral at the main panel with the PV system disconnected. If the imbalance remains, it is upstream of the inverter; if it disappears, it is in the inverter connection. If the panel terminals are tight and the imbalance remains, it is on the grid operator's network.

Why does the inverter show 400 V line-to-line if one phase is at 270 V?

Because line-to-line voltages do not depend on the neutral. With a shifted neutral the phase triangle stays the same and what moves is the centre point: some phases drop and another rises.

Sources

• Readings, production and alarms from the FusionSolar platform of the plant in this case (1 to 10 November 2026), read with Suntropy Connect.
• Text of alarm 2034 «Grid Overvoltage»: cause and recommendation from the manufacturer, included in the alarm itself.