I am working on black box model of chiller which is 5 input and 3 output model. For simplicity I have considered them as three set of 5 input and 1 output model. From the identification toolbox, I tried applying many techniques to identify the model using training data and got fit ranging from 70-80% with validation data. But the problem is most of the time the Auto-correlation plot of the output goes out of the confidence limits. Is it ok to work with this kind of model for Predictive control?
Dear Niraj,
Here are links and attached files that I hope they are in the topic.
-Identify Nonlinear Black-Box Models Using System Identification App ...
www.mathworks.com/.../identify-nonlinear-black-bo...
- Features - System Identification Toolbox - MathWorks
www.mathworks.com/products/sysid/features.html
- Black-Box Modeling - MATLAB & Simulink - MathWorks
https://www.mathworks.com/.../black-box-modeling....
- Theory and applications of HVAC control systems e A review of model ...
www.academia.edu/.../Theory_and_applications_of_...
- 2016 ACC Program | Wednesday July 6, 2016 - PaperPlaza
https://css.paperplaza.net/conferences/.../2016ACC_ContentListWeb_1.html
Nonlinear Model Predictive Control Using Polynomial Optimization Methods (I) ...
Best regards
Hello Rodrigo,
I am mostly getting this problem related to Output Error models where the Auto-correlation plot goes out of bound a lot. For some of the MISO models ARMAX structure works but for others it doesn't?
Thanks !
- Badges
- Science topic
- Similar topics
- Engineering
- Control Theory
- Control Systems
- Control Systems Engineering