It was seriously foggy this morning, so I went out for lunch at 12:20pm without my sun glasses on. What a big mistake… As soon as I reached the Berkeley library, I was blinded by the drenching sunlight!

In the past, whenever I tried to explain the powerful SF fog to my friends and family back in Taiwan, they would just think that I exaggerated it. But now I got the data from our solar system to back me up!

Here’s the fun part: The readings went from 1,014w at 12:15pm, jumped to 2,775w at 12:20pm, and then shot up to 3,562w at12:25pm! Wow! We actually hadn’t seen it passed 2,800w for a while! David guessed that the cloud probably had reflected the sunlight at the right angle and intensified it. And then the cloud probably just blocked the sun, so the reading dropped down to 548w at 12:30pm. (It started to feel like the stock market…) But no worry, it quickly came back again and touched down on 3,009w 20 minutes later. By looking at the above diagram, you got a pretty good idea of today’s weather in our neighborhood. Pretty cool, huh?

The following Power.now summary provides more information. While the median was only 548w, the maximum was a whopping 3,562w. Also the day is getting shorter…

Notice any difference between today’s and the previous days‘ diagrams?

First of all, today was the end of 2009 daylight Saving time, and our solar system reflected that in the scatter plot diagram on the right — instead of waking up at 8am, now it wakes up at 7am. I also replaced the pair diagram with a more meaningful histogram, so now we can tell how many readings per 50w. But the more significant differences are the new look. Yup, they are prettier than the previous ones!

Thanks to Hadley Wickham’s ggplot2 package*, I was able to easily produce pretty diagrams after reading his (free!) well-written qplot chapter. Hopefully once I’m done with his book, I’ll be able to produce more meaningful, beautiful statistics diagrams.

But what was the driving force behind this “beautification“? Well, David got intrigued by my R diagrams, and this afternoon he started investigated gnuplot. An hour or so later, his shell script was able to produce a whole week of data in smooth lines! Man, I’d better hurry up, or  I won’t have my job much longer!

*: For those who are interested: The ggplot2 talk he gave at the Bay Area userR Group in Sep. was impressive! You can find his data, code and slides here, and the details in “Beautiful Code” — “Chapter 18 Bay Area Blues: The Effect of the Housing Crisis.”

David saw the LED kit at the Evil Mad Scientist Laboratories about a month ago and that gave him the idea of a robot pumpkin for the Halloween (萬聖節). Even though his Robot Pumpkin looked very cute during the day (but not as cute as my baby pumpkin), it actually looked pretty scary in the yard at night! But apparently not scary enough for those little trick-or-treaters.

We got about 8 kids dressed in all sorts of different cute customs; too bad that my camera didn’t work that well to catch them. Our last visitor was an adorable 2-year-old Elvis. He looked rather confused and didn’t realize that he got a role to play, so instead of questioning, “Trick or treat?“, he extended his tiny left hand, looked at me, and asked frankly, “Give me candies.” David, I, and his young parents next to him all burst out laughing! What was I supposed to do besides following his cute demand?

My baby pumpkin, David's Robot Pumpkin, and David

Come see David’s new pet — The Evil Domo Santa! Today as we walked by the window display of Kingston Bookstore, this little monster grabbed David’s attention and he immediately fell in love with it. At first I thought that he was getting it for his little buddy, JC, but no no no, it was for himself! How cute! So I said, “Nah, I’ll get it for you as your Xmas present. Merry Xmas!”

The Domo Santa is scary in a very cute way and fits perfectly with the Halloween spirits. See how protective it is of David? I can almost hear it barks, “Go away! David is working! Or I’m gonna bite you!“

Today I cleaned up my scripts and set up the data storage structure (to get ready for decades of solar power data! ). I’m sure that I’ll wanna tweak our daily output later as I learn more about R graphics, but as for now, I’m happy enough with them and want to work on my long To-do list first. Plenty of Python/R/shell scripts to write in the future! And once I’m happy enough with the daily and weekly data analysis (I’ll add the monthly and annual ones later on), David will setup the cron jobs to fully automatically produce graphs and data daily/weekly/monthly/annually. Pretty neat, huh?

	It turns out that our solar system is an amateur weatherman, too. Even though it couldn’t tell us that today was a windy day, it did show us that it was cloudy around noon.   The sun: Can you see me now? Our solar system: Yes! I can still see you! Last Sunday evening, it even recorded the power outage as well (and dropped the daily total power statistics as a result).
	The pair plotting graph looks cute, but the most important one is the Power vs Time graph as shown above. I might drop this eventually when its cuteness wears out.

From the summary statistics, we learned that our power system was happier than yesterday even though it was cloudy around noon. What does that tell us? To get the full picture about the daily power generation, we really need both the graph and the summary.

Continue our Solar Power Adventure…

Last week David spent his vacation writing a C software package called fslurp at SourceForge, to “read and display data from Fronius IG and IG Plus inverters, using a serial port to talk with the Fronius Interface Easy Card.” Because now we have a way to read our inverter’s beautiful mind, it’s time to start collecting data and do something about it!

First I wrote a tiny Python script to clean up the dataset generated by David’s fslurp. Next I loaded the cleaned dataset to R for the data visualization step. However, things didn’t go as well as I had in mind; the plotted graph looked strange — each reading was a short horizontal line instead of a circle.

After investigating the problem for a while, I realized that I got tripped over by the infamous Factor object. Fortunately, plenty of experienced R users have already solved this problem and posted their solutions along with detailed explanations online. Even better, William Revelle actually implemented the same solar power plotting project back in Feb. 2006! While he had 38 months of data, I just had 1 day to play with. But, still!

Tomorrow I’ll need to add a little shell script to have the whole process automatically run at the end of the day. Frankly, it’s more interesting to have at least 1 year of data, so we can see the amount of power generated following the changes of season. But we’ll get there, and we’ll have a lot more data to track!

[10/27/2009 UPDATE] Yesterday I forgot to post the daily summary: