How do i have to standardize new incoming data to a trained model?

Question

I have a trained model in python whose inputs are standarized arrays, following the formula (data_point-mean)/std. Now I wish to test the network with new data:

• Let's say (just an example to understand the situation) my input data is a 2D array, and the first dimension has 2 different values: 100 or 200. I want to test if the model is able to predict the intermediate values between the two previous ones: 101,102,103,..., 199 (one unit increments) with a fixed second dimension value.

With this in mind, how do I should standardize this new "artificial" data? Do I add them to the real unstandardized dataset, standardize all together and then remove the new data and reserve it for testing? Or do I have to standardize them separately without the original data?

Thanks.