SunPower harnesses the power of NVIDIA V100 GPU-enabled virtual machines on the Google Cloud Platform to help homeowners design their ideal solar solution in seconds, and bring clean, sustainable energy to homes globally.
A faster route to solar power
SunPower is always searching for new ways to help homeowners go solar—but despite the clear environmental benefits, solar can be complicated and it’s not always easy to win them over.
Going solar is a big decision that involves researching solar panel solutions and evaluating the roof space needed for each system. Every project requires SunPower designers to figure out the dimensions of the roof, identify any obstructions like chimneys and vents, as well as figure out the costs and potential savings. And to add to the challenge, every roof is different and requires a unique customized design.
It’s not an easy task—in fact, if SunPower solar designers were to make these calculations for each of the 100 million solar homes in the U.S., it would take them over a century.
SunPower knew that if it could make solar simple and develop a solution that showed homeowners the value of solar—and enabled them to design their own solar solution—all in seconds rather than hours, it could transform the buying experience. Not only would it reduce the barriers to going solar, but it would also enable SunPower to serve customers on a much larger scale.
SunPower just needed the right technology to handle the compute-intensive workloads the design process involves.
The value of solar in seconds
SunPower created Instant Design, a solution that allows homeowners to create a design automatically in less than 60 seconds using just a home address—and it all runs on the Google Cloud Platform using NVIDIA V100 GPU-enabled virtual machines.
The SunPower Instant Design API service uses the one petabyte of satellite and digital surface (DSM) data across the world, along with a database of manually-generated designs as a base. Using this data, the service can develop machine learning models through a quick process of training, validation, and analyzing test results.
These machine learning models cover each step of the solar design process, including everything from roof segmentation to obstruction identification.
Using NVIDIA V100 GPU-enabled virtual machines alongside Google Machine Learning Training solutions through the Google Cloud Platform, SunPower was able to train and test new models in just one day—a process that previously took a week. And by enabling the company to train multiple models simultaneously, the platform also allowed SunPower to iterate and improve upon its models. Once the models were deemed suitable for homeowner use, SunPower was able to quickly integrate the models with its web application.
An entirely new solar experience
With the flexibility of NVIDIA GPUs delivered through the Google Cloud Platform, SunPower can choose the processing power that best fits its needs. Working with a combination of CPUs, GPUs and Cloud TPUs, it can ensure it has the best performance for its Instant Design API service.
Now, the Instant Design API service has the computing power necessary to carry out all the calculations involved in assessing homeowners’ rooftops—from light exposure and potential energy production, to estimated costs and potential savings.
What can now be done in a few seconds with just a couple of clicks, previously took hours. And not only does it speed up the design process for SunPower, creating the potential to generate hundreds of unique designs every minute, but it also enables homeowners to understand the value of solar power almost immediately.
With easy access to NVIDIA GPUs and recent advances in aerial imagery, SunPower has been able to change the way it offers solar power to homeowners, quickly transforming the solar industry today.
Try SunPower Design Studio, powered by Google and NVIDIA, on your home at www.sunpower.com/design.