0
$\begingroup$

I am working on the LSTM time series forecasting of solar energy production. The available data is one year on a half hourly basis. More than 60% of the data values are zero as the PV stations cannot produce solar energy at night. Now my questions are as follows,

1) If this big amount of zeros somehow affecting the learning process?

2) Is there any way to deal with such sparse data set?

$\endgroup$
0
$\begingroup$

1) If this big amount of zeros somehow affecting the learning process?

Have you actually tried. Yes, there are a lot of zeros, but if they are all occurring at night, then they would show up as regular daily seasonality pattern, so they shouldn't effect your ability to predict too much.

2) Is there any way to deal with such sparse data set?

What is the purpose of your forecast? Do you need half-hourly forecasts? If not, I would suggest that you aggregate up to a daily level. You will end up with 365 data points, which is a lot smaller than the original data set, but is still enough for fitting a time series model.

$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.