Comparing Energy3D's prediction of solar panel yields with real data from a house in Massachusetts

Fig. 1: Street view before solar panel installation

How accurately can Energy3D predict the energy generated by solar panels? This is critical for Energy3D as our goal is to provide a reliable engineering tool for modeling and designing solar energy applications. Even if our primary target users are students, there is no room for complacency, simply because engineering is all about accuracy and it is important that we pass this spirit to the next generation.

Fig. 2: Comparing predicted and real data

We have compared Energy3D's results with sensors placed on the horizontal plane and the vertical south-facing plane and concluded that Energy3D predicts satisfactory results. But we haven't compared Energy3D's predictions with output data from real solar panels.

My colleague Dan Damelin of the Concord Consortium has recently had solar panels installed for his house (Figure 1 shows his house before solar panels were installed). His solar system, which consists of 34 SunPower panels estimated to have a total power output of 11 kW, went into operation last December. By the time I am writing this blog post, he has accumulated six months of data, providing a good basis for a case study. So I asked our summer intern, Guanhua Chen, a PhD student from the University of Miami, to conduct a case study that uses Energy3D to analyze Dan's solar system.

The solar company that Dan hired came to his house, surveyed the site, and gave him a proposal that detailed the layout of the 34 panels. They also provided him a projection of monthly outputs, juxtaposed with his monthly electricity bill. The solar company's estimate is shown in Figure 2 as the gray line, whereas the bar graph represents the monthly electricity usages.

Fig. 3: An Energy3D model of the house

Guanhua used Energy3D to create a 3D model of the house and put 34 SunPower panels following the actual layout done by the solar installer (Figure 3). The dimension of the SunPower panels is slightly different from that of most other brands (which is approximately 3 by 5 feet). They have great solar cell efficiency, which is about 21% -- one of the highest in the market.

Comparing with the real production data from December to June (represented by the red line in Figure 2), the solar company's projection overestimates a bit of the yields in the winter months but significantly underestimates those in the summer months. By comparison, Energy3D's predictions (represented by the green line) for the spring and summer months agree much better with the real data. Like the solar company's predictions, Energy3D seems to overestimate the winter production. This may be due to the fact that we haven't incorporated the effect of snow and ice in the simulation core of Energy3D. Should we factor this effect in the calculation, the results would be more accurate. In our next iteration of the computational core, we will build a mathematical model of the snow effect.

If Energy3D can outperform the production software used by the solar installer in this case -- as Figure 2 seems to suggest, the implication could be enormous, because this is a free tool so easy that every student can use. With it, we now have a serious chance to engage and enable students to solve critical energy problems. And there are million of students out there! If a fraction of them can be turned into little solar engineers by Energy3D,  the world could be a better place sooner.