Journal
APPLIED THERMAL ENGINEERING
Volume 164, Issue -, Pages -Publisher
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.applthermaleng.2019.114442
Keywords
Waste heat recovery; Organic Rankine cycle; Power optimization; Real-time implementation; Internal combustion engine
Ask authors/readers for more resources
Organic Rankine cycle waste heat recovery technology has been gaining more and more attention in recent years. Real-time power optimization is crucial to the system performance. Even though individual real-time optimization methods exist, literature rarely investigate comparison of different real-time power optimization methods. This paper first time compares three real-time implementable power optimization methods for an organic Rankine cycle waste heat recovery system. Three optimization methods include Proportional-Integral-Derivative rule-based method, nonlinear model predictive control and dynamic programming. In the Proportional-Integral-Derivative rule-based method, rule-based method defines the optimal working fluid vapor temperature trajectory and Proportional-Integral-Derivative controller manipulates the pump speed to follow that trajectory. In the nonlinear model predictive control method, reference vapor temperature is defined close to the saturation temperature and then the model predictive control minimizes the vapor temperature tracking error by controlling the pump speed. In the Dynamic Programming method, the net power produced by the organic Rankine cycle system is defined in the cost function, which is maximized by controlling the pump speed. A Random Forest machine learning model is utilized to extract the rules for Dynamic Programming and then implemented in real-time. All the three methods are implemented in the same experimentally validated plant model for the comparison analysis. The comparison results show that the Dynamic Programming Random Forest method has similar performance with nonlinear model predictive control method and outperforms the Proportional-Integral-Derivative rule-based method by 9.9% in net power production. Dynamic Programming Random Forest method can be an alternative to the nonlinear model predictive control for its low computation cost and high net power production.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available