Dynamic Modeling And Optimization Control Of CO₂Centrifugal Compress

Surge phenomenon is a normal illness in the running centrifugal compressor. It will reduce the compressor operating range and the unit’s safety. Currently, the most common anti-surge control methods in centrifugal compressors are fixed flow way and variable flow way. In The actual application process, it tends to use the method which will reduce the most system performance to control the surge phenomenon. Thus it is very necessary to seek a safe and efficient way to make surge phenomenon under control.In this thesis, the object of study is the Ningxia Petrochemical Company’s urea production plant which is a carbon dioxide centrifugal compressor. Combinedexperience of engineers, an identification scheme which is for the foundation of the mathematical model of the centrifugal compressor is raised up. The scheme is based on the theory of BP neural network and the conclusions are as follows.1. A turbine speed model of the centrifugal compressor which is based on the theory of BP neural network is established. Test data of the Valve which number is V1is collected from step response test method. In this method, the operating way is controlling the V1valve. The test data is analyzed under the theory of correlation analysis. The turbine speed of the centrifugal compressor is used as the inputs of the speed model and the model is tested under15groups of sample data. The result shows that the model credibility is89.75%and the average error is0.000465.2. A multi-input and multi-output model of the centrifugal compressor’s turbine speed is established under the theory of BP neural network. By changing the reflux valve which number is172, the state transition of the centrifugal compressor can be got. Using the correlation analysis, the inputs of the model can be found and the model is tested under15groups of sample data. The result shows that the model credibility is97.28%and the average error is0.00232.3. Under the platform of Simulink in the software of Matlab, the simulation controller is build up. By simulate the model, it can be found that the control system can be optimized under the condition of Feedforward decoupling.