Efficient predictive control method for ORC waste heat recovery system based on recurrent neural network
•A novel predictive control method is proposed for the Organic Rankine Cycle.•The recurrent neural network with special linear transfer functions is used.•The optimal control problem is converted into a mixed-integer linear program.•The proposed method outperforms various control strategies in compa...
Gespeichert in:
| Veröffentlicht in: | Applied thermal engineering Jg. 257; S. 124352 |
|---|---|
| Hauptverfasser: | , , , |
| Format: | Journal Article |
| Sprache: | Englisch |
| Veröffentlicht: |
Elsevier Ltd
15.12.2024
|
| Schlagworte: | |
| ISSN: | 1359-4311 |
| Online-Zugang: | Volltext |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Zusammenfassung: | •A novel predictive control method is proposed for the Organic Rankine Cycle.•The recurrent neural network with special linear transfer functions is used.•The optimal control problem is converted into a mixed-integer linear program.•The proposed method outperforms various control strategies in comparison studies.
The model predictive control performs well in regulating operating parameters and ensuring system safety when the organic Rankine cycles are operating with variable heat sources. However, traditional nonlinear predictive control based on the organic Rankine cycle mechanism model involves significant computational complexity, making it challenging to quickly find a control solution. This limitation hinders its application in the organic Rankine cycle for rapid response control. To address this issue, a fast model predictive control method is proposed in this work. A recurrent neural network model is well-trained using the input–output data of the organic Rankine cycle process, and it is used as a surrogate model in the design of the model predictive controller for the control of organic Rankine cycle operating parameters. The formulated optimal control problem is then transformed into a mixed integer linear programming problem, which can obtain high-quality and fast solutions during the control process. Through comparison with recurrent neural network-based nonlinear predictive control and pseudo-sequential method-based fast nonlinear predictive control, the results show that the designed controller can effectively accomplish the control of organic Rankine cycle operating parameters with smaller overshoot. Moreover, its average control solution time is shorter by 89.59% and 93.27% respectively while the total net output power of the system during the control process is 0.54% and 1.3% higher than that of the other two controllers. It exhibits superior control performance, even under variable waste heat conditions. |
|---|---|
| ISSN: | 1359-4311 |
| DOI: | 10.1016/j.applthermaleng.2024.124352 |