A data-driven approach for self-optimizing control with equality constraints

A new data-driven self-optimizing control (SOC) strategy for selecting of controlled variables (CVs) with equality constraints is presented. The method is a follow-up on our earlier work on data-driven SOC technique for CV selection without equality constraints explicitly considered. The new approach does not require model linearization for a nonlinear process which usually results in large economic losses due to linearization errors. This new method, unlike the one developed in the previous work, does not require to solve equality constraints explicitly, hence, computationally more efficient. The procedure has advantage of tightening the optimization problem towards a right and feasible solution which required satisfying all the equality constraints especially in the event of large range of uncertainties and disturbances affecting the system. The proposed procedure is demonstrated on a 3-stream heat exchanger network (HEN) benchmark problem. The strategy results in economically optimal HEN operation with constant target temperatures and minimum energy usage.