Study On Integrated Control Method For Centrifugal Compressor

Centrifugal compressors are widely used in petrochemical and chemical production processes. Saving energy and improving products'quality are the new goals of centrifugal compressor's control system. The integrated control system with advanced control theory and method for centrifugal compressor is the new challenge of the combination of control theory and its application.In this paper, centrifugal compressor's working principle and characteristic are studied in detail. Solutions are proposed to solve a series of problems which exist in centrifugal compressor control.(1) Based on Least Square Support Vector Machine Regression (LS-SVR), centrifugal compressor's modeling method is proposed to solve some problems in existent centrifugal compressor modeling methods. These problems chiefly include model precision being low, modeling process being complicated and some model parameters can not being measured directly. Model form is chosen as NNARMAX regression model. In order to reduce difficulty of modeling, centrifugal compressor's two inputs and two outputs model is transformed into two models according to control requirements. One is mass flow's two inputs and one output model, the other is discharge pressure's two inputs and one output model. Simulation results show that centrifugal compressor's LS-SVR model has a good dynamic characteristic. Compared to mechanism model and Greitzter model, its structure is simple and precision is high. Also, it can be easily modified.(2) Centrifugal compressor anti-surge nonlinear model predictive control strategy is proposed to solve the problem of centrifugal compressor's efficiency being low which is caused by control line being far away from surge line. Based on the prediction of predictive model, whether operating point crossing surge line or not can be judged in advance so the influence of actuator's adjusting time can be reduced, the distance between surge line and control line can be shorten and centrifugal compressor's efficiency can be increased. In order to avoid surge, anti-surge conditions are chosen as the limiting conditions of reference trajectory and rolling optimization. Simulation results show that the centrifugal compressor anti-surge nonlinear model predictive control strategy can increase centrifugal compressor's efficiency and avoid surge effectively at the same time.(3) Cooperative control strategy of centrifugal compressor's mass flow and discharge pressure is proposed to solve the problem of centrifugal compressor's operating point being unstable. This problem is caused by mass flow controller and discharge pressure controller influencing each other because of centrifugal compressor's characteristic not being considered. The error doesn't be fed in controller directly by adding a trajectory adjust module but is fed in the trajectory adjust module through which reference trajectories are calculated according to changing of operating point and adjusting time of two actuators. Then the reference trajectories are fed in controllers as set point through which let operating point move smoothly according to its reference trajectory. Reference time is chosen according to both the driver's and the throttle valve's adjusting time in order to make sure that the two actuators can eliminate errors synchronously. The experiment results show that the cooperative control strategy of centrifugal compressor's mass flow and discharge pressure can keep operating point moving smoothly and reduce the impact on the production process while set point is changed. Especially, when operating point is near to surge line, the cooperative control strategy can avoid surge effectively.(4) Wavelet de-noising double threshold optimization method based on ant colony algorithm is proposed to solve the problem of centrifugal compressor vibration fault detection's result being inaccurate which is caused by the measured signal containing noise. The reasonable optimization interval is determined according to the distribute characteristic of noise. The objective function is chosen according to the difference of auto-correlation coefficient between signal's wavelet coefficient and noise's wavelet coefficient. The optimal upper threshold and lower threshold are calculated by ant colony algorithm. Using the amplitude of key frequency as eigenvector and combining the wavelet de-noising technology, centrifugal compressor's vibration fault detection system is designed based on Back Propagation (BP) neural network. The vibration fault detection system can diagnose the vibration fault online. Simulation and application results show that the proposed de-noising method can choose the optimal double threshold and de-noising effectively. And the centrifugal compressor's vibration fault detection system can extract and identify the fault signature fast and precisely.(5) Centrifugal compressor's integrated control system is designed based on analyzing centrifugal compressor's working principle and characteristic. Industry PC, Programmable Logic Controller (PLC) and field bus I/O module are used. And Ethernet, industry Ethernet and field bus network technology are combined to build up a five level distributed control system. The centrifugal compressor's integrated control system makes sure that the production process is safe and centrifugal compressor's efficiency is increased. The application results show that the design proposal is effective.