Are you getting the right insights from terabytes of solar PV plant data?
[caption id="" align="alignright" width="300"] Utility scale solar PV plants produce terabytes of data that need to be analyzed to understand the plant's performance.[/caption]
The Challenge: Mountains of Data
Anyone involved with solar plants knows that they are instrumented with numerous weather and electrical sensors and that this pv plant data is logged every single minute. Because of the large number of sensors and frequency of collection, you may end up with a very large data set (3 million points per day on some plants). The process of installing sensors, data acquisition and communication devices can be so complicated that sometimes people forget what the data is for: understanding plant performance. Solar plant owners and operators want to know if a plant is operating normally or what they need to do to bring it back to normal. While data is important, what that data tells us is even more important.
Last month, I spoke about analyzing PV plant data at the 3rd Utility Scale PV Plant Optimization conference hosted by PV Insider in San Jose, Calif. Over the two days of the conference, the speakers include a melting pot of PV developers, EPC companies, utilities and technology providers. My full presentation is on YouTube and embedded below, but here I will summarize some of the solutions to analyzing and presenting all of this pv plant data.
The Solution: Data Presented Graphically
To get to what the data shows us quickly, I feel that data should be summarized in relevant metrics and presented in graphical format for quick comprehension. At the conference, I presented Anscombe's Quartet as an example. Research in neuroscience and cognitive science has shown that we can only store three to nine chunks of visual information at a time in short-term memory. Individual numbers on a dashboard are stored as discrete chunks, but a well-designed graphical pattern can represent a great deal of information as a single chunk. While there is sound scientific basis for using graphs and charts, unfortunately many charts fail to convey information in an easy manner or simply distort information. You will find several examples of bad charts and dashboards in my presentation.
Know Your Audience
The first users of plant information are often technicians or operators who lack the time and skill to manipulate and interpret extensive data. They may make bad decisions based on poorly presented data. All of us are familiar with false alarms that have resulted in expensive truck rolls. I believe engineers need to present information in an easy to understand way to operators if we expect to run our plants efficiently. I have demonstrated simple charts and dashboards in my presentation to answer common questions like:
Is the plant operating normally right now?
What is the progress in the last year on production targets/guarantees?
Is the loss of performance due to soiling or inverter faults or DC system problems?
How to verify reporting accuracy of irradiance sensors?
I hope that my accompanying presentation (embedded below) provides you with some insight into solar plant data analysis and display. We use these techniques at Cupertino Electric to operate plants efficiently for our customers and make sure they meet or exceed stated production targets.
[youtube http://www.youtube.com/watch?v=tR4lAzCWTT8]
Do you have examples of good charts of solar plant data that need to be shared? You can email me at:
and I will include your chart in a follow-up. I welcome your thoughts and insights on this topic in the comments below.
