8 min read
In my short time frame of working with this data, I’ve found the results to be very encouraging. Days that were sunny and optimal for solar collection saw the optimized array obtain almost exactly, if not a little more electricity than the optical eye tracker. On very cloudy days, they also produced the same amount. Two specific types of weather were the clearest indicators of success for me: partly cloudy days and days with an exceptionally heavy frost.
When looking at the power production results, it’s important to note that at this time, I don’t have an elegant solution to obtaining this data; however, it is a question I am currently working on.
The challenge is that, while I love SolarEdge inverters, the information about module level performance (i.e., what each panel is producing), although collected by SolarEdge, is not easily available in an programmatically obtainable way. SolarEdge has an API I could use; however, the public API does NOT provide this level of detail. I currently have some parts on order that will allow me to connect to the RS485 port on the SolarEdge inverter to hopefully obtain this information. That, as well as ongoing results, will come in a future blog.
For now, I will have to use screen shots and manually obtained power numbers to show the results. Essentially, I took a screenshot of the monitoring portal or handwrote the results and then manually added up the nine panels in each array as my methodology. It’s manual and frustrating but still allows me to show the impact in this blog with hopefully better graphs in the future, once I solve my SolarEdge problem. SolarEdge, if you are listening: please make it easier for enthusiasts to get to their data – that’s a wonderful way to build a community of advocates!
November 17th was a perfect example of a very cloudy day in November; we had some peaks of sunlight, though, and did get some mild solar production. That said, the optical sensor on the south array was not helpful.
The power totals for the day were not great; it was a very cloudy day, but even in the poor conditions, when the sun did peek out, it allowed the north array to achieve 6% more performance over the south array.
North Array: 2200.25 Watt Hours (2.2 kWh)
South Array: 2069.75 Watt Hours (2.07 kWh)
The north array is roughly 6% over the south array in KWh production. This is obviously a small amount; however, it’s worth noting in these specific examples.
Sunday, November 18th was a day that by all accounts should have been an optimally sunny day; however, we started the day with an extremely beautiful and relatively common scene we see here in Wisconsin: The Hoarfrost (https://weather.com/science/weather-explainers/news/hoarfrost-explained/).
While this picture is beautiful, do you see the issue?
This photo makes it clear. At 8:20 in the morning, the north array (left) was pointed where it needed to be pointed. The south array was pointing a little to the west. My theory is that the hoarfrost on the cover of the optical sensor caused a refraction that confused the tracking.
How did this look from a tracking perspective?
North Array: 371.5 Watt Hours
South Array: 176.75 Watt Hours
North Array: 3913.5 Watt Hours (3.9 kWh)
South Array: 905.5 Watt Hours (0.9 kWh)
It’s clear that even at an early morning hour, the north array is greatly outperforming the south array. Almost a 3 kWhs difference in an hour!
These examples, and the relative parity of performance on optimal days, lets me feel good about my data journey. At the time of this writing, I have one array being controlled by the Raspberry Pi and the other still using the optical eye. With this in place, I want to monitor the differences in production between the two arrays over a larger period of time and provide meaningful results in a follow-up post.
As stated above, I also need to work with my SolarEdge inverters to get accurate readings per array. My intention is to both automate the collection of module level data and then track this over time to write another follow-up post on the results observed over different seasons.
Even without seeing the results in power production over time, the project itself – using a Raspberry Pi to sense the angle of an array, to control an array, to log data to a MapR cluster, to work with that data test and show results, check math, and validate hypothesis – was extremely rewarding. I’ve made my GitHub repo available, including sample data, for people to see how I’ve worked through the IoT and data science here: https://github.com/johnomernik/solarpi.
Even more importantly, I approached this problem as a curious individual who wanted to see “if” I could do it. I could, but, as you can see, there were some flubs along the way, and I welcome those of you with more skills in various areas to show me where my thought process, my code, my testing, or any part of my project could be improved upon.
If the improvement can be incorporated into my code, I will gladly work on issues submitted there. This has been a great process that allowed me to take my passions – green energy, data science, data engineering, and basic curiosity – and mix them all together into something where I, and hopefully others, have a better understanding of the world around them.
Stay ahead of the bleeding edge...get the best of Big Data in your inbox.