Model Predictive Control technology as the most typical method of Advanced Process Control has been developed rapidly and applied widely in process industry. Weather MPC controllers operate well or not affect the enterprise benefit directly, therefore it is significant to carry out economic performance assessment and optimization. As a pragmatic method, LQG benchmark introduces manipulated variables together with controlled variables into quadratic dynamic index, which is more suitable for industrial application than Minimum Variance Control benchmark. However, traditional LQG benchmark is regressed from an unbalanced distribution of discrete points set, which causes unnecessary calculation and poor regression in more important parts of points set. To avoid this, the relationship between manipulated variables variance and controlled variables variance was calculated, an equal-grid LQG benchmark was introduced and added in two layers MPC optimization problem as an equation constraint, then the optimal MPC setting values or reference trajectory could be obtained. The economic performance assessment and simulation results for delayed coking furnace show the effect of proposed methods.
展开▼
