Understanding How To Implement Rolling Horizon In PYOMO

[Dennis Gray]

Hi Everybody, 

I have a NLP that gets very large and it ends up being the case that I am unable to solve it just by calling Knitro, so I am hoping to try and implement a rolling horizon approach to help the model solve in a more timely manner. I am trying to figure out how I can implement a rolling horizon approach in PYOMO and was hoping someone might have an example of how they were able to do so, or provide some type of skeleton to help guide my own implementation. 

In a previous post on the PYOMO forum (Previous Rolling Horizon Question: Pyomo) it seems that the OP already had an idea as to what they were doing but I am not fully understanding it. For example, the OP sets their time horizon to be 12, but then to me it seems that they have 8760 solves in between? 

In any case, I have provided 2 links below - one as previously stated and another that details what the rolling horizon algorithm is. There is also an additional paper that uses a rolling horizon approach and is similar to my problem in the sense that I am trying to also develop a bioeconomic model of control. 

Any insights into how the rolling horizon algorithm can be implemented in Pyomo would be of help. 

Previous Rolling Horizon Question: Pyomo

OR Stack Exchange: Brief Intro To Rolling Horizon

Büyüktahtakın, İ. E., Kıbış, E. Y., Cobuloglu, H. I., Houseman, G. R., & Lampe, J. T. (2015). An age-structured bio-economic model of invasive species management: insights and strategies for optimal control. Biological invasions, 17(9), 2545-2563.

[jen p]

Hi Dennis, 

Did you manage to figure it out how to implement rolling horizon in pyomo? I am also trying to implement this and a bit stuck, it would be great to know if you managed to do it! 

Kind regards,

Jen

2020년 4월 16일 목요일 오후 3시 26분 30초 UTC+2에 Dennis Gray님이 작성: