Active surge control of centrifugal compression systems

This thesis addresses modeling and active surge control of a simple compression system, using only the drive system of the compressor for actuation. Theoretical results are validated by experiments on a test rig that was built as a part of this work. Control laws are derived in two stages. First all system states, such as pressure mass ‡ow and impeller speed, are assumed as available signals. All the resulting control laws require feedback from mass ‡ow. However, transient measurement of this variable is not available. Observers are therefore developed for this purpose. The laboratory that was build and from which experimental data are gathered, consists of a compressor, plenum volume, control valve, in between ducting and drive system. The centrifugal compressor stage compresses air from ambient conditions, discharging into a duct connecting compressor and plenum. The plenum discharges over the valve. The drive system consists of a electric motor and drive. Controllers and observers are derived on the basis of the so-called variable speed Greitzer model. This model contains the compressor map, which in many cases involves some uncertainty. Uncertainty can be caused by varying ambient conditions and change of map due to wear and tear. For this reason it was investigated whether it was possible to express the compressor map by measurements. It is shown that this can be done if one has a su¢ cient length of duct upstream or downstream the compressor. Two control strategies are presented. One considers the impeller speed and the other considers applied torque as control variable. As previously mentioned, all control laws require feedback from mass ‡ow, and must be combined with an observer for implementation. Furthermore, torque control laws involve the compressor map. In view of modeling results, this map can be implemented by a model or a measurement in the case of su¢ cient duct lengths. Experimental validation of control laws indicate that torque control v vi SUMMARY is better than speed control for the setup in question, in terms of it ability to stabilize surge. The reason for this can be explained by the internal loop of the drive system, o¤ering a faster response for torque than for speed control. It is shown by experiments that torque control can extend mass ‡ow range of the machine by more than 15 to 20 percent relative to the point for which surge occurs for the open loop system. These numbers corresponds …