Abstract: The economic performance of plants can be effectively improved in terms of control system structure optimization through the selection of necessary conditions of optimality (NCOs), including constraint and gradient information, as the implementation targets of control systems. In response to the difficulties in direct online measurements and in timely and accurate offline estimations of NCOs, a novel approach to NCO modeling, prediction, and control based on a Gaussian process is proposed. To process real-time data, the proposed process can update the Gaussian process model dynamically online and the NCOs can be accurately predicted to enable the best decision of the optimal control rate for the next moment, thereby enabling process economic performance to achieve optimality. The proposed approaches are applied to an exemplary chemical process simulation, and the results show the effectiveness of the proposed method.