[Blog #4] Running More Experiments and Simulations

In these last two weeks, most of my time was spent outside testing sample panels, which was very exciting! All samples of the 3 different panels I ordered from Alibaba arrived one after the other, and I was able to finish almost all of the outdoor testing that I wanted to get done, which meant generating around 40 curves that display the performance of the panels under various conditions. Under full sun, I decided to test the panels horizontally, facing south and angled 52 degrees from vertical (default orientation), and partially shaded (while still at default orientation). To conduct partial shading tests, I needed to figure out exactly how I should shade the panels to acquire the most insightful results. I decided I would orient them "longwise" and compare the panel’s performance when the left half is shaded to its performance when the bottom half is shaded. This would provide me with a stronger understanding of how the cells are strung together within the panels, i.e. which cells are in parallel and which are in series. Testing all of the panels under these two shading conditions revealed some interesting discoveries. Most of the panels barely produced any power when half shaded either way, but two of the panels were still able to output around 1.7W when the bottom half was shaded, which is still about half of their rated power. This data allows me to declare with very high certainty that the cells in those two panels are strung together horizontally in series and vertically in parallel, meaning they would outperform the other panels when exposed to foliage that obstructs the panel horizontally.

Another project that the company is working on is building a test harness that can connect to one of our devices and run various tests on both the hardware and software. I was tasked with building an equivalent circuit model for our solar system to see how our device's circuitry responds to it. Thankfully, I learned how to do exactly that in my EE classes this past semester! To build an equivalent circuit model for one of our panels, I needed to extract the shunt and series resistance of the panel from it's full I-V curve. After going outside under full sun and recording data across the entire I-V range of the panel, I was able to find the shunt and series resistance by calculating the negative reciprocal of the slope at the very beginning and very end of the I-V curve, respectively. I found a shunt resistance of 1541 ohms and a series resistance of about 1.44 ohms, which seemed about right to me. Then, I began building the circuit using silicon diodes and the two resistances that I calculated. Once I hooked it up to a power supply and an electronic load, I recorded data across the entire I-V range once again to compare it to the I-V curve of the actual panel out in the sun. The result I got was simply a more idealized version of the panel's actual curve, which is exactly how I hoped it would look!

This past week, I conducted experiments in low-light conditions: around 9 am and 5:30 pm. I decided to test the panels during these times because they are equidistant from solar noon, which is about 1:15 pm. This will allow me to make a more "apples to apples" comparison when it comes to morning vs. evening performance. Although the default orientation is facing south, we may need to face some panels east or west if there is foliage that obstructs the south side of the panel. Therefore, I wanted to run these tests at the adjusted orientation to observe how they'd perform in lower light during the morning and evening. One of the challenges I've had with testing is that sometimes I don't sweep a large enough range of voltages to get an accurate maximum power point. What makes this difficult is that while I'm outside taking data, I'm not actually calculating the power yet and am instead just jotting down the current at different voltages. Doing a full power calculation for each data point while testing would take too long and run the risk of sunlight conditions changing too much during a single test. This issue has made me need to redo some tests to sweep a broader range, but I'm getting better at intuitively knowing where to start/stop along the I-V curve.

I want to take a step back for a second and talk about the motive behind running all of these different tests at different conditions. As I mentioned in my last post, one of the goals is to calculate the maximum power point, but there is another key facet of these experiments that leads to an important subsequent action item: setting the panel's voltage to achieve maximum power. Essentially, the important metric isn't actually the maximum power itself but the voltage that produces that maximum power. This is because voltage is the one value that can actually be altered and controlled within the device. What these tests are really looking at is how that maximum power point voltage changes under different conditions. When I was talking to my director about this, he suggested I convert the I-V curves that I generated to P-V curves in order to more clearly see how different voltages change the overall power output. After a few hours of very tedious Excel work, I finally converted all I-V curves to P-V curves, and they looked much better.

I only have two more weeks left in my internship, and I'll mainly be focusing on building a system that can record solar panel data across an entire day. This will be instrumental in determining which panels/orientations yield the highest power output across a full day. I've never built a system that records data across such a long span of time, so I'm really excited to dig in and learn more about the programming side of data collection.