kW-Labs/nmecr

TOWT predictions cannot predict negative values?

Closed this issue · 5 comments

Cyy-fresh1994 commented

We used the nmecr to predict annual normalized electricity consumption.

Our training data has a lot of negative values due to solar PV production. And the model fit by using training data can capture the negatives, see the screenshot below:
image

But when we do predictions, the predictions are all positive, it looks like the negative values were cut off, see the screenshot below:
image

Cannot nmecr be used for predicting negative data?

dexterluu commented

Hi,

Do you have data and code to replicate the issue? Seems like in the first plot the values can go negative so it doesn't appear to be an issue with the coefficients.

Thanks,
Dexter

Cyy-fresh1994 commented
dexterluu commented

Sorry to have to ask again, but I don't think the attachments went through via email. Can you attach the files on the issue thread on GitHub?

Thanks,
Dexter

Cyy-fresh1994 commented

Historical Meter 2.csv
Historical temperature data.csv
TMY2 Temp - San Fran.csv
nmecr-demo - Copy.txt

The nmecr-demo file should be .rmd, but I cannot upload rmd files. Please change txt to be rmd before you open it.

Thank you,
Yiyi

dexterluu commented

Hi Yiyi,

I updated the package and included an argument to allow for negative values in the calculate_model_predictions function. Please reinstall the package to get the change.

Thanks,
Dexter