| 摘要 |
The systems and methods described herein allow for automatic identification experiments in a closed loop, where the old control strategy, already tuned and tested, is utilized. The strategy is modified to inject additional signal optimized for identification. The experimenting time may be reduced by performing only those system manipulations which explore model uncertainties important to potential degradation of controller performance by discrepancy between the system and the model. The disruptions are reduced by keeping the control loop closed, which eliminates waiting for steady state before applying steps to the inputs and reduces the risk of process limits crossing. The energy of additional signal can be set to meet the maximum allowable disruption requirements. The energy of additional signal is in a direct relation to the speed of identification related information gathering. It can be varied in time to follow the needs of system operators.
|
