Locus’ analytics team is constantly striving to improve the methods that we use to detect underperforming systems in the fleet of residential PV systems that we monitor. This led us to investigate relative performance metrics in 2011-2012 and to develop our own in-house Virtual Irradiance system and associated GEER metrics in 2013-2014 in order to efficiently analyze the weather-adjusted performance for very large fleets of PV systems. One of our latest efforts has been to improve upon those analyses by developing the “Waterfall” analysis for the residential systems in our fleet.
Why, and by how much, is a system underperforming?
The drivers of system underperformance can be issues that require corrective maintenance action, such as partial or complete system downtime; issues that could potentially require action, such as soiling or snow cover; or issues for which it is unlikely that action would be helpful, such as shading, clipping, or degradation.
Where the GEER metric addresses the question of how much a system is underperforming, the Waterfall asks both how much it is underperforming and why the underperformance is occurring.
To successfully address these questions, we need the most accurate data possible on both the actual system performance and the modeled system performance (under the assumption that the system is working exactly as intended). We achieve this by combining the highest-quality, most comprehensive weather and satellite irradiance data with best-practice performance modeling techniques and power and energy data from Locus LGate monitored systems at a 5-minute granularity. This ensures the highest fidelity of measured and modeled power and energy time series data.
Using proprietary, physically-inspired statistical methods, we were then able to analyze the differences between modeled and measured energy to estimate the extent to which systems were affected by seven primary causes of PV system underperformance: clipping, downtime, snow cover, partial capacity loss, soiling, shading, and degradation.
Why is it called the Waterfall?
The origin of the name "Waterfall" comes from one of the visualizations that we utilized extensively during the development phase of this project. Below is an example of a waterfall for a site with some downtime issues and some shading issues.
One question arises quickly with a site like this: How do we account for the energy losses when the system is shaded and isn’t producing due to downtime?
If we suspect that a system should be producing 100 kWh, but is only producing 80 kWh due to an unfortunately-located tree, then it seems justifiable to say that we should have a loss of 20 kWh due to shading. At the same time, if the system is totally down due to an inverter failure, it may actually produce nothing, making the system's total downtime loss 100 kWh. However, both of these can’t be correct at the same time; it doesn't make any sense to have 120 kWh of energy losses for a system when it only would have produced 100 kWh under ideal conditions.
We are left with a situation where one can select any shading loss less than or equal to 20 kWh, which is arguably correct as long as the downtime loss plus the shading loss is no more than 100 kWh. On the bright side, at least the system isn’t soiled too, as that would make even more of a mess of the accounting!
Conveniently, the Waterfall chart, above, shows how we resolved this very clearly (and why we call the whole tool the Waterfall). We model each specific type of loss consecutively (left to right in the chart) and take into account previously allocated losses. So, since downtime is to the left of shading on the chart, we would have calculated a downtime loss of 100 kWh and then had a shading loss of 0 kWh for the scenario described above. Thus, our solution is consistent, relatively easy to explain, and squarely within the space of solutions that are arguably correct.
What can you do with categorized loss data for 70,000 residential sites?
A lot! One of the first thing that comes to mind is that we can tell to what extent fleets of residential systems are underperforming with a granularity that was not previously possible. This will allow us come up with compelling answers to questions like "How different are soiling rates in Southern California as compared to the East Coast?" This will have to wait for its own blog post that is coming out soon!
Locus' Virtual Irradiance tool provides highly accurate solar irradiance data for asset managers, performance engineers, and operators without on-site sensors. To learn more about Virtual Irradiance, please click here. The Virtual Irradiance Performance Waterfall is a game-changing loss-estimation feature in the LocusNOC platform that allows asset managers, owners, and operators to reach new levels of efficiency and cost-reduction. Click here to request more information on the VI Performance Waterfall.