Most Solar Systems Don’t Fail at Once. They Slowly Underperform.

Solar systems rarely collapse overnight. Performance usually erodes gradually: a few percentage points lost here, a string offline there, a recurring inverter issue that goes unnoticed for months.

For distributed fleets, especially residential, small commercial, and older systems, monitoring is often limited. Many installations lack detailed DC-side sensors and rely primarily on inverter AC data. That makes fault detection reactive rather than proactive.

A research team in Australia may have found a scalable way to close that gap.

A Simpler Way to Detect Underperformance

Scientists from the University of Technology Sydney, the University of New South Wales, and Diagno Energy developed a rule-based method that detects and classifies PV underperformance using only AC-side inverter data.

No additional sensor buildout. No high-resolution DC monitoring. The approach uses five-minute AC power data combined with basic system information such as size, location, tilt, and orientation.

How the Five-Step Method Works

The system gathers five-minute inverter AC power data along with basic system details. It estimates expected production under current weather conditions and compares that to actual output. Underperformance is flagged using clear thresholds: major issues are identified when output falls below 60% of expected for three consecutive clear days, while minor underperformance is recorded when output remains below 80% for seven days.

The method also checks for recurring patterns such as weekday-weekend differences or seasonal shifts. It analyzes five-minute production signatures to classify likely fault types, including inverter tripping, clipping, sustained zero output, or repeated generation drops. Finally, it generates a summary report outlining severity and probable cause.

Real-World Testing Across 1,000+ Systems

The method was tested across 1,089 PV systems and 2,213 inverter monitors in Australia between 2021 and 2023.

The results were strong:

• ●92% accuracy in detecting major underperformance
• ●88% accuracy in detecting minor underperformance
• ●Lower accuracy for ambiguous cases, such as clipping, with room for refinement

The key takeaway is scalability. The model works with data that most systems already collect.

Why This Matters for Distributed Solar

Distributed solar is no longer niche. Australia has more than four million rooftop PV systems. The United States has nearly five million distributed installations. Across fleets of that size, even small performance losses compound quickly.

Improved detection means earlier fault response, reduced lost generation, better maintenance prioritization, and stronger long-term reliability. As installed capacity continues to expand, performance management becomes as important as deployment itself.

The next phase of solar growth is not just about building more systems. It is about managing them intelligently at scale.