Updated: Nov 12, 2021
Hey Everyone, I’m Joe and I lead Product here at OpenSolar. I got my first job in solar in 2009 at Sungevity - at the time a local, Berkeley-based installer that was pioneering a novel technology that promised to significantly reduce the cost of solar: Remote Solar Design.
The appeal of Remote Solar Design (RSD) was straightforward enough: If you had reliable remote site assessment, you could eliminate the initial site visit and save a truck-roll, which typically costs a couple hundred dollars. If you do this early in the sales process, when conversion rates were still in the single digit percent, you could cut your customer acquisition costs by a few thousand dollars per job - a lot of money in an industry where fat margins are hard to come by.
But making Remote Solar Design a reality was anything but straightforward. Sungevity had a mysterious super-developer named Adam, who was based in Sydney, Australia and built the first-of-its-kind RSD software, leveraging newly available, high-resolution oblique aerial imagery to construct 3D models of each site. But even with that software, we still relied on a team of very sharp, highly trained “RSDers” to painstakingly complete these designs. I sat next to our RSDers for about a year and got to watch them in action.
For any given site, they’d trace out the ridges and gutters, hips and valleys of each roof face in multiple oblique views. They’d then map each edge that they defined in one view to the corresponding edge that they defined from another view. Then they’d layer in obstructions. For trees, for instance, they had to decide whether each tree looked more like a lollipop or a traffic cone and then add one of those shapes to the 3D model. Only after all this would we add the modules and run the energy production simulations and savings calculations. (Then, given the myriad of opportunities for human error, we’d have someone else on the team run the design through a quality assurance process to ensure there hadn’t been any mistakes, so we could confidently offer a performance guarantee against the production estimates.)
It was a lot of work! But, happily, times have changed.
Adam from Sydney, who I now know as my friend Adam Pryor, has been hard at work refining RSD technology and taking advantage of the latest datasets which are far richer than anything we had 10 years ago. Rather than relying on 2D oblique images, or LiDAR-based point clouds, the latest generation of the software uses photogrammetry-based Digital Surface Maps (DSMs). This DSM-based approach allows us to instantly generate a high-resolution 3D model of any site without any input from the user. No defining roof edges in multiple views, or raising roofs off the ground. Just enter the address and the 3D model of the site is there. This offers huge benefits to the user:
It allowed us to cut the time it takes to complete a design from 15-20 minutes to under 2 minutes
It’s dramatically reduced the training and expertise required to use the software
It allows for a level of shading accuracy that cannot be achieved with other technologies
It has virtually eliminated the potential for human error
Crucially, it produces incredibly accurate estimates.
Given my understanding of the tool - and my faith in Adam to do amazing things! - I’ve been confident from the start that the reliability and accuracy of OpenSolar’s design tool is second to none. But since not everyone knows Adam or has the time to dig into the technology, we’ve engaged with some of the most trusted experts in the industry to provide independent, 3rd party assessments of the accuracy of OpenSolar’s design tool.
Accurate production estimates are ultimately a function of two things: project inputs (e.g. pitch, azimuth, shading, scale, weather and hardware parameters) and the energy model (i.e. the math that converts those inputs into outputs). So we set out to get each of these assessed by third parties.
We asked NREL (the U.S. National Renewable Energy Laboratory), which has been benchmarking the accuracy of shading assessments in solar design tools for years, to evaluate the accuracy of OpenSolar’s analyses. They identified 80 locations on homes in Los Angeles and Denver and we used OpenSolar to judge the Sun Access Values (SAVs) at each location. They then compared those values to measurements that they had made on the roof using a SunEye. We’re thrilled to report that no other solar design tool scored better than OpenSolar did:
OpenSolar’s assessment of Sun Access Values was accurate to within +/- 3%
In a separate study using the same methodology, NREL found that the current US market leader was only found to be accurate to within +/- 5%
Then we engaged PVEL (PV Evolution Labs), the leader in PV bankability testing, to assess the accuracy of our 3D site models. PVEL sent a couple of their engineers onto roofs of all shapes and sizes where they meticulously measured the slope and scale of each roof. A separate engineer took the same measurements in OpenSolar, and then compared their results with the on-the-roof measurements. Again, we are happy to report that they found the accuracy of our models to be best-in-class:
OpenSolar’s assessment of scale was accurate to within 1 ⅓ feet 100% of the time (no other tool provides as narrow a tolerance)
OpenSolar’s assessment of pitch was accurate to within 4 degrees over 97% of the time (no other tool provides as narrow a tolerance OR such high compliance)
Finally, we had our energy modeling validated. OpenSolar’s energy modeling is based on NREL’s most advanced System Advisor Model (SAM) and we asked PVEL to validate the integrity of our implementation of SAM. PVEL estimated production for sites in a range of environmental conditions (e.g. hot, unshaded; cool, shaded) in OpenSolar, and compared those estimates to estimates for the same sites made in industry standard energy modeling tools like PVsyst or standalone implementations of NREL’s SAM. PVEL found:
OpenSolar’s energy modeling is consistent with standard engineering practices
OpenSolar’s production figures are within 0.0%-0.2% of standalone SAM estimates
OpenSolar’s production figures are within 0.1%-1.9% of PVsyst estimates
The last 15 years have been good for Remote Solar Design. Being able to quickly, easily and reliably produce highly accurate designs was always the dream during my Sungevity days and that dream is now a reality.
We’re very excited about where we go from here!