Local self-optimizing control based on extremum seeking control

Self-optimizing Control (SOC) aims to find controlled variables with which setpoint regulation of the resultant feedback control loops can yield near-optimal operation under a range of disturbances. However, standard local SOC methods, e.g. the null-space SOC, require an offline analysis with large amounts of steady-state data, which can be computationally cumbersome. In this paper, we propose a new SOC procedure enabled by extremum seeking control (ESC) which will largely simplify the offline analysis process of null-space or extended null-space SOC methods. First, ESC is used to determine the optimal manipulated variable values under the nominal condition for the system. Next, by dithering the plant with periodic disturbances, the dither-demodulation technique in ESC is used to estimate the Jacobian and Hessian needed for obtaining the optimal measurement combination; then the null-space and extended null-space methods can be carried out in a computationally efficient fashion, for the scenarios with noise-free and noisy measurements, respectively. The proposed procedure are compared with the standard null-space and extended null-space SOC methods using a Modelica-based dynamic simulation model of an air-source heat pump (ASHP) system. The results show that a similar performance can be achieved with much simpler process of data acquisition and processing.