Neural Network for modelling a predictive control system- Using NARX

I am modelling a predictive control system for cooling unit used for data centers. I have used NARX for training the network. Network seems to behave well and performance data is good enough. I suppose that is the first step: System Identification.

Once the network is trained, for multi-step ahead prediction, I am using the new data and results are close enough. Now I am wondering how can we deploy the model for real time prediction. I have seen the examples of autonomous car which uses NN but still not getting how it uses real time data for training and deploy it for use.

Any help would be really appreciated. My many thanks,

Abhishek Guhe

Graduate Student UTA