Due to limitations of hardware, cost and so on, the application of proportional-integral-derivative (PID) control is more convenient than predictive control. However, predictive control usually has better performance than traditional PID control, thus it is important to combine the advantages of these two control algorithms. A novel PI controller optimized by predictive functional control (PFC) is proposed and tested on liquid level in the industrial coke fractionation tower in this paper. Since this kind of process always shows the integrating behavior, a P controller is first used for it to generate a self-balancing generalized process, then the PFC based PI control is designed for the generalized process. The resulting controller displays the performance of both PFC and PI control with easy implementation in practice. The performance of the proposed PI controller is compared with traditional PI controller in terms of regulatory/servo setpoint tracking, disturbance rejection and measurement noise issues, from which results show that the proposed PI controller provides better performance than traditional PI controller. (C) 2014 Elsevier Ltd. All rights reserved.