Identification of a Self-Optimizing Control Structure from Normal Operating Data

Abstract

In building systems, it is necessary to maintain comfort conditions while minimizing operating costs. One way to do this is to use a real-time optimizer (RTO) to adjust the setpoints of the controlled variables. Another way is to use a self-optimizing control (SOC) structure to find one or more new variables that, when controlled to the appropriate constant setpoints, drive the operating cost of the system to, or close to, its optimal point. A requirement of SOC is that an optimal operating point be used in the calculation of the self-optimizing variable(s) and the identification of its parameters. In this work, we propose a formulation of SOC that allows for the use of non-optimal data. The proposed method makes use of normal operating data to identify the parameters used to calculate the self-optimizing variables. Simulation of an HVAC system shows that the performance of the proposed method is similar to that of SOC based on optimal data, and also to the performance of an RTO alternative based on extremum-seeking control (ESC).