adding example jupyter notebook

[GregoryKimball]

This request adds a jupyter notebook and a "docs" folder to the repository. PVLIB uses a similar structure to provide tutorials to the functionality.

[GregoryKimball]

This pull request may address issue #10
[Provide usage examples #10 ]

[mdeceglie]

@GregoryKimball I made a few minor changes, please take a look and make sure they look good to you.

[mdeceglie]

This is a great start to illustrate the workflow. Let's make sure to keep it updated as we build out the tools.

[GregoryKimball]

Thank you for improving the notebook.

I added a centering step so that the data is initially centered around 1 before the analysis begins. I recommend doing the centering step between filtering and analysis to prevent outliers from off-centering the data.

[mdeceglie]

Thanks for illustrating the centering step. I suggest it might be more general and readable to use DataFrame.rolling().median()

[GregoryKimball]

I did some experimenting with DataFrame.rolling(), but for sites with seasonality in the performance metric (i.e. a lot of them), I think the DataFrame.resample('12M).median().max() may be more appropriate.

[cdeline]

I reviewed this branch, and really like it. I think it's important to get some usage examples up quickly, and this is a great start. The fact that the Kyocera system isn't a clean dataset may lead to issues - e.g. -8% per year degradation rates, but hopefully this will be resolved as we improve the filtering.

I did include a few additional plots to visualize what the output of the YoY and OLS regression looks like.

[mdeceglie]

The data from the BP system (12) system may make for a cleaner example: http://dkasolarcentre.com.au/source/alice-springs/dka-m5-b-phase

[cdeline]

Good suggestion Mike - I like system 12 very much!

[dirkjordan]

We could also consider one of the NREL systems that nobody cares about anymore like the ASE.

[GregoryKimball]

Thank you Chris and Mike for adding the new BP system. I made some updates to the metadata and filter to make sure this site runs OK.

Ready to merge?

[cdeline]

I vote for merge. Chris Deline National Renewable Energy Laboratory 303-384-6359 From: Gregory Kimball [mailto:notifications@github.com] Sent: Monday, May 22, 2017 9:58 AM To: kwhanalytics/rdtools Cc: Deline, Chris; Comment Subject: Re: [kwhanalytics/rdtools] adding example jupyter notebook (#17) Thank you Chris and Mike for adding the new BP system. I made some updates to the metadata and filter to make sure this site runs OK. Ready to merge? — You are receiving this because you commented. Reply to this email directly, view it on GitHub<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fkwhanalytics%2Frdtools%2Fpull%2F17%23issuecomment-303142972&data=02%7C01%7Cchris.deline%40nrel.gov%7C6f4595d01e1640f14f5a08d4a12b6084%7Ca0f29d7e28cd4f5484427885aee7c080%7C1%7C0%7C636310655064273697&sdata=SQ9rxxp5U4RAyt8rfXyS8e8GPATwA9zbzto8znaqKm4%3D&reserved=0>, or mute the thread<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAWKtOBriJ2nnHOJx3MZ0oY05jBg1mdfzks5r8bCegaJpZM4NCE3Y&data=02%7C01%7Cchris.deline%40nrel.gov%7C6f4595d01e1640f14f5a08d4a12b6084%7Ca0f29d7e28cd4f5484427885aee7c080%7C1%7C0%7C636310655064273697&sdata=QrsntepnLhWugMj7WrbRm414vEfQ3HH3Fb%2FrdumGVoY%3D&reserved=0>.

[mdeceglie]

I will update the example to work with the newly merged versions of the degradation functions.

[mdeceglie]

The previous version was using ambient as cell temperature leading to significant seasonality. I updated the normalization step to model cell temperature based on POA irradiance. But there are still remaining problems:

-OLS is not a good idea with 15 minute data, there are a lot of low data points that cause a highly suspect fit
-The new version of YOY won't accept anything more frequent than daily.

The solution to both these problems is to aggregate with an irradiance weighted sum daily. I'll keep working on this in the example.