The University of Massachusetts Amherst
University of Massachusetts Amherst

Search Google Appliance


Pioneering Computer Model Detects Faults in Residential Solar Arrays

solar array

Solar Array

College of Engineering faculty and students were part of the UMass Amherst research team that developed Sundown, a computer model for residential solar fault detection featured in a recent article in PV Magazine

Residential-scale solar systems lack sensing and instrumentation for performance monitoring and fault detection, as do larger utility-scale arrays, so these smaller solar arrays can often experience costly faults that go undetected for long periods. SunDown detects and classifies per-panel faults in residential solar arrays without the need for additional expensive sensors.

The research team members published their research findings in the journal Compass '20, with lead author Menghong Feng. Feng graduated with an M.S. from the Mechanical and Industrial Engineering Department (MIE) in 2020 and now works for Apple in China. 

Other researchers involved were Associate Professor David Irwin of the Electrical and Computer Engineering (ECE) Department and his doctoral student Noman Bashir; Dragoljub (Beka) Kosanovic, the assistant director of the Center for Energy Efficiency and Renewable Energy in the MIE department; and Associate Dean of Computing and Facilities and Distinguished Professor Prashant Shenoy of the College of Information and Computer Sciences.

"Sundown is a great example of the ongoing collaborations between the College of Engineering and the College of Information and Computer Sciences," says Shenoy, who supervised Feng's M.S. thesis that resulted in the Sundown work and led to the Compass '20 paper

As Irwin explains, "The research in the paper uses two years of solar data from a real home and a manually generated dataset of solar faults, which show that this approach can detect and classify faults (including from snow, leaves, debris, and electrical failures) with 99.13 percent accuracy. The computer model can also detect concurrent faults with 97.2 percent accuracy."

As background for this innovative research, the Compass '20 paper notes that recent technological advances and falling prices have led to a significant increase in deployments of both large utility-scale and smaller residential solar arrays.

According to Kosanovic, "Large utility-scale solar farms tend to be instrumented with sensors for monitoring real-time generation to identify production issues, and these techniques can be employed in commercial and utility solar farms as well."

However, as the paper explains, "Due to cost reasons, smaller residential-scale systems lack such sensing and instrumentation and may only level faults. Thus, it is not uncommon for residential solar arrays to encounter power anomalies or other local faults that go undetected for long periods, resulting in generation and revenue losses."

(December 2021)