Energy and utilities

Solar PV plants

In a PV plant failures show up by comparison: an inverter or tracker that drifts from its twins under the same sun.

In operation

Request the private demo The five failure modes Download the brief (PDF)

The failures that matter

Five components, in order of impact.

Sibilla has Blueprints in operation for each of these. For every one: how it fails, what announces it, how early you can see it, and what it costs you.

  1. 1

    Inverter: IGBT power module overheating

    Downtime Cost

    Thermal cycling and poor cooling stress the power modules and DC-link capacitors. Internal temperatures creep up against load and conversion efficiency drops before a trip.

    Signals
    Internal temperatures against load and ambient, conversion efficiency, derating events, alarm frequency.
    Typical warning time
    Days to weeks. (indicative)
    Teneuro
    Partly. Extra temperature points. The main data comes from the inverter logger.
  2. 2

    Inverter cooling: fan and filter failure, derating

    Downtime

    Clogged filters and worn fans reduce airflow. The inverter runs hotter and derates, losing yield long before it stops.

    Signals
    Internal temperatures against load and ambient, derating events, fan noise and vibration.
    Typical warning time
    Days to weeks. (indicative)
    Teneuro
    Yes. Fan noise and vibration, plus extra temperature points.
  3. 3

    DC side: insulation loss, ground faults

    Downtime Safety/Env

    Damaged cables, moisture and ageing connectors lower insulation resistance until the inverter trips on a ground fault.

    Signals
    Insulation resistance trend measured in the morning, ground-fault alarms, temperature in field combiner boxes.
    Typical warning time
    Days to weeks. (indicative)
    Teneuro
    Partly. Temperature in combiner boxes.
  4. 4

    Trackers: actuator and sensor faults

    Downtime Cost

    Worn gearmotors, faulty angle sensors or lost communication leave rows at the wrong angle. That costs yield every day and risks damage if stow is missed in high wind.

    Signals
    Actual against commanded angle, motor current, travel time, communication losses.
    Typical warning time
    Days to weeks. (indicative)
    Teneuro
    Yes. Motor current and vibration where the tracker controller does not expose them.
  5. 5

    MV transformer: overheating

    Downtime Cost Safety/Env

    Load peaks, high ambient temperature and poor substation ventilation heat oil and windings. Heat ages the insulation and shortens transformer life.

    Signals
    Load current and power, oil or winding temperature, ambient temperature.
    Typical warning time
    Months to years: it is life being used up, not a sudden alarm. (indicative)
    Teneuro
    No. Load and temperatures come from the transformer monitor and the plant SCADA.

Also relevant

Outside the five above: not covered by a Blueprint today, we can work on them as a project.

  • Strings and modules: hot spots, bypass diodes, PID and accelerated degradation.
  • Soiling and shading losses.
  • Overheating field connectors and string fuses.

Systems and connectors

What we usually find in the plant, and how we read it.

Read-only, without touching the machines. Three levels, stated plainly: Native, 1-day activation, Project-based.

SystemViaLevelPrerequisite
AVEVA PIPI Web APINative Read access to the customer's PI server.
Huawei SmartLoggerModbus TCP, via Teneuro or a gateway1-day activation —
SMA Data Manager, Sungrow LoggerModbus TCP / SunSpec, via Teneuro or a gateway1-day activation Logger register map.
Monitoring platforms (meteocontrol VCOM, Power Factors Drive)REST API1-day activation API access enabled on the customer's account.
Plant SCADA with OPC UA or SQLOPC UA or SQL1-day activation OPC UA or SQL access opened on the SCADA.
Tracker network controllersModbus TCP1-day activation Only where the controller exposes Modbus TCP. Closed controllers are project-based.
Plant SCADA on IEC 60870-5-104 or proprietary protocolsGatewayProject-based —

Trademarks belong to their respective owners and indicate technical compatibility only.

Limits

Where predictive maintenance does not pay off.

  • Single modules: they are replaced when they fail.
  • Small plants without string current measurement.

There, good preventive maintenance costs less. We will tell you before we start.

Standards

The references we work with.

IEC 61724-1
PV performance monitoring: system classes and KPIs such as performance ratio.
IEC 62446-2
Preventive, corrective and performance maintenance of PV systems.
IEC 62446-1
Documentation, commissioning tests and inspection of grid-connected systems.
SunSpec Modbus
Open data model for inverters and meters.
ISO 14224
Taxonomy of components and failure modes behind the Sibilla ontology.

Our solutions on this asset

Sibilla

Blueprints in operation on the five components above: KDIs, expected degradation patterns, links to the fault tree.

Discover Sibilla →

Teneuro

Where the data is not there: vibration and sound, fieldbuses and the signals of silent machines, sent to Sibilla.

Discover Teneuro →

Before Plasmiam, our team designed and ran predictive maintenance for a major international utility: more than 1,000 power plants in 22 countries and over 40 GW of installed capacity, starting from hydro. That experience shaped Teneuro, designed, built and patented by Plasmiam to bring vibration monitoring to large fleets of small plants at a low cost per point.

Tell us what you run. We'll show you how it fails and when you'd notice.

Request the private demo Share it with your team (PDF)

Sources

Last updated 3 October 2026. Warning times are indicative orders of magnitude from literature and practice: they depend on the duty profile and are validated on your fleet.