Thoughts on Solar

On Sunday, May 20, 2012, an annular eclipse was visible in large parts of the United States, including here in Pasadena. To the data geeks here at Run on Sun, that gave us a view to the eclipse that was just a bit different - a Data Geek’s view, if you will.

One of the great things about having a solar power system that is monitored - particularly with a system like Enlighten from Enphase - is that it automatically captures a great deal of data and sometimes that data can be used to look at some surprising things - like a solar eclipse! So this will be a brief tutorial on how to take a raw data set and turn that into a graphical representation of the solar eclipse. (We used Excel 2010 for our analysis, but any tool that allows for the creation of pivot tables should allow you to do the same thing.)

Hypothesis

We started out with a simple hypothesis - since the annular eclipse covered as much as 80% of the sun’s surface (as seen from SoCal), we would anticipate that there would be a significant impact on the output of a solar power system during the eclipse. If we looked at the data from the days prior to the eclipse (excluding any days that were dominated by cloud cover during the relevant time when the eclipse occurred) we should see a noticeable decline in system output relative to the other days. If we graphed those days on the same graph, the eclipse should be an obvious outlier.

But to test our hypothesis, we would need to gather some data!

Data, data, data…

The Enphase Enlighten website allows a system owner to access a variety of reports about how the system has been performing. One of those reports is titled “Site Recent Power Production” and it provides a comma separated values (csv) file that contains the site’s power output over five minute intervals for the preceding seven days.

Excel will open the .csv file and once you have it opened, you can begin the process of massaging the data into the form needed to “see” the eclipse. (If you care to follow along, you can download the dataset that we used by clicking on this link.)

The raw data consists of two columns: a time stamp value and the power produced at that moment in time in Watts. The time stamp combines the full date with the time of day, and as you can see here, the time increases in five minute intervals. Altogether, our week’s worth of data consists of 2,016 data pairs which we will ultimately want to graph.

However, we want to partition the data into seven discrete days (which will then be graphed on top of each other), with hours of the day and five-minute intervals on the X-axis. To achieve that end, we need to add some columns based on the time stamp value. Fortunately, Excel has a number of built-in functions that will let us do exactly what we want to do with a minimum of effort.

By-the-way, it is interesting to note that the system is already generating a significant amount of power by 5:45 in the morning! The system being monitored has 209 solar panels which means that at 5:45 we were receiving, on average, just over 2 Watts per panel.

This ability to produce power at very low levels is another way in which a microinverter system helps to maximize overall system yield.

We continue this process by creating columns for the hour and the minute. Fortunately, this is nothing more than using the built-in “hour” and “minute” functions from Excel.  By selecting our date/time field with hour and minute, we can establish the remaining data fields that we will need.

However, we now have lots of rows of data to process and making sense of that can be difficult—difficult, that is, unless you use pivot tables.

Pivot Table to the Rescue!

Now we are ready to create our pivot table. For those of you who do not know, pivot tables are a data geek’s best friend, and the means by which massive amounts of data gets distilled into a meaningful table - and ultimately a graph. There are lots of things that you can do with a pivot table but the one we will be using here is very simple.

On the insert tab choose pivot table. For the data source select the power column and the three new columns that we just created. You can embed the table on the current worksheet or in a new one.

Then do the following:

• Select the “Day” field and drag it to the “Column Labels” box.
• Select the “Hour” field and drag it to the “Row Labels” box.
  • Do the same thing with the “Minute” field.
• Now drag the “Power Produced” field to the “Values” box. Note that it gets labeled as “Count of Power Produced (W)” in the box and all of the data values are “1″s. What Excel is doing is counting how many power values there are for each five-minute interval - not what we want.
  • To change it, click on the down arrow next to the field and select “Value Field Settings…” Where it says “Summarize value field by” change from Count to Sum and click OK.
• Click on the heading “Column Labels” and type Days.
• Do the same for “Row Labels” and enter Hours/Minutes.
• We don’t need the “Grand Total” so right mouse on that heading and select “PivotTable Options…” Click on the “Totals & Filters” tab and deselect both of the Grand Totals options.

If you’ve followed along correctly, you should have a pivot table that looks like the one on the right.

Click on the drop down filters and deselect the “blank” value for both Days and Hours/Mins. Also, we only want hours where the total power exceeded zero watts, so select Value Filters from the Hours/Minutes drop down. Select Greater Than… and enter 0.

Now we are ready to create our graphs.

A Picture is Still Worth a Thousand Words!

We have distilled our 2,016 rows of raw data into a more manageable form, but it is still hard to “see” our eclipse - time to graph this and let the data speak for itself!

From the PivotTable Tools menu, choose options and then click on PivotChart. Choose a Line chart from the pop-up. Right mouse on the chart and select “Move Chart…” to move it to a new worksheet. This looks pretty good, but Day 5 is very noisy - let’s exclude it. Click on the Day drop down and deselect Day 5. That’s our eclipse graph on the right.

Can you “see” the eclipse now? The chart is still quite busy; let’s exclude some additional days to see if we can make this cleaner.

Return to the Day drop down and exclude every day except days 3 & 7. Now our new graph just tracks two days and we excluded a lot of distractions from what we were seeking.

Here’s our new graph:

The graph clearly illustrates that these two days saw almost identical power production for the vast majority of the day, until late in the afternoon. The timing of that sole difference coincides with our eclipse - can we zoom in for a closer look?

Easily enough we can - all we need to do is limit the time span displayed to the time period associated with the eclipse. The eclipse began around 5:25, reached its maximum roughly an hour later and was over by an hour after that. By selecting the Hour drop down filter we can limit the time period to the four hour window between 5 and 8.

Here is our final “view” of the eclipse:

Our final graph clearly depicts the impact of the eclipse on the power output of our solar array.

Monitoring systems are generally considered an important part of commercial solar systems since they help the system owner maximize their investment. However, this is just one example of how a monitoring system can also be used as a tool to observe and quantify other real world phenomenon - making this an important adjunct to a solar power system at a school - like this one at the Westridge School for Girls in Pasadena.