“No Faults Found”: Why a Clean Dashboard Can Hide a Dead Inverter
A monitoring dashboard with no open faults looks like good news. Often it only means the system has nothing to say. A dead inverter does not raise an alarm — it stops reporting. A device that sends no data can still show a perfect health score. And a check that could not run looks exactly like a check that found nothing. Ellume Vector treats silence as information: it detects inverters that have gone quiet or never came online, scores a device it cannot assess as “unknown” rather than healthy, and reports how many of its 46 detection checks actually ran on your site — and why each of the others did not.
Most monitoring is built around alarms, and alarms are built around data. A reading crosses a threshold, a fault code arrives, output falls below what the weather allows — something in the data triggers a response. That works well for the faults that announce themselves. It works badly for the faults that do the opposite. When a device stops sending data, there is nothing to cross a threshold, no fault code and no output to compare. The alarm logic has nothing to act on, so it stays quiet, and quiet looks like healthy.
A dead inverter doesn’t raise an alarm. It just stops talking.
Who this matters to
Control room and O&M teams who rely on the fault list to decide where to go, asset managers who report fleet health to owners, and anyone who has been told “no issues this month” and wondered how the system could be so sure.
Three ways a dashboard says “all clear” when it doesn’t know
- •The device that goes silent. An inverter that loses communication, or fails so completely that its logger goes down with it, does not report a fault. It disappears from the data. Many monitoring systems miss it entirely, because every check they run needs a reading to work on. The same applies to an inverter that never came online after commissioning: with no history, it looks no different from a unit with a temporary gap.
- •The score that defaults to healthy. Health scores are usually built by deducting points for problems found in the data. A device that sends no data has no problems found, so it keeps full marks. The result is a dashboard where the healthiest-looking inverter on the site is the one that has stopped working.
- •The check that never ran. Some checks need signals that not every site provides — a sensor that was never installed, a value the inverter does not transmit, data too sparse to judge. When a check cannot run, most systems simply report nothing, and “nothing” reads as “no fault found”.
Why silence passes for health
The damage spreads further than one missing device. Loss calculations need output to work with, so a silent inverter reports no measurable loss and sinks to the bottom of any list ranked by energy at stake. Fleet benchmarks that still include dark devices drag the average down and make the working equipment look artificially good. And a day with a six-hour data hole, totalled as if it were complete, quietly understates production in every report built on it. None of this is a sensor fault or a software crash. It is a design choice: when the data is missing, the system fills the gap with the most reassuring answer available.
What the dashboard shows, and what is true
| Situation | What many dashboards show | What is actually true |
|---|---|---|
| Inverter stops reporting | No alarm; device absent from the fault list | A possible total failure, losing energy every daylight hour |
| Device sends no data | Health score of 100 | Health cannot be assessed |
| Check lacks the signal it needs | No fault found | The check did not run |
| Silent unit in a loss ranking | Zero loss, ranked last | Potentially the largest loss on site |
| Day with a six-hour data hole | A complete daily total | A partial day, understated |
Three quiet weeks: an illustrative case
Numbers illustrative. A 3 MW central inverter loses communication on a Friday evening. On Monday the fault list is clean, the site health summary is green, and the inverter's own health score still reads 100. Because it reports no output, it reports no loss, so it never reaches the top of the weekly priority list. Three weeks later a technician on a routine visit finds it tripped. At around 15 MWh per sunny day, the site may have lost more than 300 MWh — at an illustrative $60 per MWh, close to $19,000 — while every screen said nothing was wrong. With silent-device detection, the same inverter is flagged on Saturday morning, its health score reads “unknown”, and it is ranked on the capacity it represents, labelled as an upper bound, rather than on a loss it cannot report.
How Ellume Vector reports what it cannot see
- •Silent and never-started devices detected. Vector flags an inverter that has gone completely silent, and separately detects one that never came online at all, so neither can hide as a temporary gap.
- •“Unknown” instead of a flattering default. A device sending no data is scored “unknown”, not 100. When a site first sees this, some devices will move from looking healthy to showing unknown. Nothing has broken — the score has stopped inventing good health.
- •Ranked on what is at stake. A dead inverter reports no energy loss because it reports nothing at all. Vector ranks it on the capacity it represents instead, clearly labelled as an upper bound rather than a measurement, and non-reporting devices are excluded from peer comparisons so they no longer distort the benchmark.
- •Partial data marked as partial. A day with missing data carries its true coverage rather than being presented as complete, 30-day trends show how much data each day was based on, and data freshness is shown on screen, so stale numbers cannot pass for current ones.
- •A coverage report for every site. Vector reports how many of its 46 detection checks ran on your site and, for each one that did not, the reason. Some sites, for example, do not transmit the signal needed to detect an inverter repeatedly starting and stopping. Vector says so, rather than returning “no fault found”.
Three questions to ask your monitoring system
What does it show for a device that has sent no data for a day — an alarm, or nothing? What health score does that device get? And for a site with no open faults, can it tell you which checks actually ran? If the answers are “nothing”, “the same as before” and “no”, a clean dashboard is telling you less than it appears to.
Frequently asked questions
- Why don’t monitoring systems flag silent inverters?
- Most alarm logic is triggered by data: a reading out of range or a fault code. A device that stops sending data triggers nothing, so unless the system is specifically built to treat missing data as a signal, the failure goes unreported.
- A device now shows “unknown” health. Is something broken?
- Not necessarily. “Unknown” means the device is not sending enough data to be assessed. It is an honest answer that used to be hidden behind a default good score, and it is a prompt to check the device or its communications.
- What is a check coverage report?
- A list of the detection checks that ran on a site and those that did not, with the reason for each. It answers the question “why does my plant show no faults?” — because there are none, or because the checks that would find them could not run.
- Why would a detection check not run?
- Usually because a required signal is missing: a sensor that was never installed, a value the equipment does not transmit, or data too sparse to judge. Knowing which checks are missing tells you where your monitoring has blind spots.
