Non-derivative algorithm for optimization and control of steady-state systems

Abstract

A non-derivative algorithm for determining and maintaining the optimum steady-stale operating point of an industrial process is proposed. The method employs the curve fitting technique to construct an approximate linear model which can match the output at each iteration and match the required derivatives of reality at the optimum point. In order to update the process control vectors and ensure that the objective descends, a Newton step technique and an augmented term are then used. The algorithm's convergence and optimality conditions are investigated under mild assumptions. Simulation results show that the new algorithm requires less set-point changes, and exhibits a higher convergent rate than a previous non-derivative method.