0
$\begingroup$

I am working on a machine learning problem, where i have to predict if a device connected with solar panel at a specific block has been damaged. I have many factors to consider, out of which atmospheric temperature, wind speed and a humidity are to be considered. These 3 factors are being recorded from 3 separate devices. This device helps pack cells ini case of severe windspeed or other environmental factors. What if this device is also damaged. Its to model this.

What sort of signal processing can be done on this data? Do i need to use fourier transformation, wavelet, noise removal or filtering? I am at beginner level in signal processing. I tried doing some research, however i feel that i do need some help in this matter.

Edit: Its not exactly for panel glass if you thinking about the actual panel, but a device that connects between panel and transmission. This device makes decision if panel should be folded when windspeed is high for damage prevention. What if that device has been damaged.

$\endgroup$
  • 1
    $\begingroup$ 1) Can't you just look at the production output of the panel itself? 2) what makes you think that these three factors are somehow correlated with panel damage? In my plant the only panel damage was mechanical (snow/ice, roof avalanche). $\endgroup$ – Hilmar Sep 4 at 7:04
  • $\begingroup$ That's a great question, however 1) Production can decrease based on other factors also like occasional sandstorm and other patterns. There are birds also who sometimes sit on panels and it reduced output. 2) Its not exactly for panel glass if you thinking about the actual panel, but a device that connects between panel and transmission. This device makes decision if panel should be folded when windspeed is high for damage prevention. What if that device has been damaged. I hope you got an idea. $\endgroup$ – StatguyUser Sep 4 at 7:10
  • $\begingroup$ so re 1) do you have the output data? because really, "damage" on a solar panel to me would be "produces less than it should", and you can measure that on the output alone, don't need any of the environmental factors (might use the to enhance your observation, MAYBE) . $\endgroup$ – Marcus Müller Sep 4 at 7:34
  • $\begingroup$ I do have damage data. This device is surveyed at regular intervals and i have the labeled data. Now i am trying to see how to treat input model features through signal processing. If you happen to know, please share what signal processing procedured should be processed for temperature/humidity etc recorded as hourly time series, recorded every day. Thank you!! $\endgroup$ – StatguyUser Sep 4 at 9:43
  • $\begingroup$ again, do you have the POWER OUTPUT data of the panels? the rest probably really doesn't matter. The signal processing technique I'd employ would be "ignoring it". $\endgroup$ – Marcus Müller Sep 5 at 7:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.