Labels

Friday, February 20, 2015

Weather Station Project #15 - Calibration Procedures for the DIY Solar Radiation Sensor

First of all, I am an engineer, not a statistician. I am trying to come up with a way of calibrating a DIY solar radiation sensor on the cheap. The original question that I asked can be referenced at the My Weather Station subreddit.

First some background: I am building a solar radiation sensor (measures watts/meter2) from a TSL2561 luminosity sensor. The sensor will be placed under a pane of glass, coated with white translucent paint for light dispersion. The output of the sensor will be a number, representative of lux received at my house. I have 5 personal weather stations within a 15 mile radius who have expensive solar radiation sensors from which I can obtain data readings via Weather Underground. Given that cloud cover will alter the readings in each case, I am assuming that a statistical analysis is warranted. The data is also at the same time somewhat spatially distributed. I need to get the value from my DIY sensor into the ballpark of the other sensors.

My approach so far:
  1. Gather the data from the 5 sites each day for an extended period of time. 
  2. Use a Chi-square test to check if the data samples are likely from within the same distribution set (note that solar radiation results tend to follow an almost bell-shaped curve over the course of a day). 
  3. Use a sample T-test to establish a mean and standard deviation representative of the set which varies throughout the day. 
  4. Use the mean applied against the value determined by my DIY sensor to formulate an equation (assume a linear fit for right now). 
  5. Check the altered output of my DIY sensor against values from the other 5 sites in a Chi-square test to determine if I am likely from the same distribution. 
  6. Run some sort of correlation against the data set to further solidify the equation that I have found.
Question: is this approach viable? Are there any glaring issues with what I am trying to do? Is there anything else that might solidify the accuracy of the calibration?

To Be Continued ...