Research On Model Predictive Control For The Global Operation Efficiency Optimization Of The System

The optimizing control towards the global operation efficiency of the system can realize the improvement of the total factor productivity;thus,it has become more and more popular in recent years.Take industrial processes as an example,the global operation efficiency refers to the total economic performance of the entire process operation.The traditional method is to use a two-layered(upper and lower)hierarchical optimizing control strategy,where the upper layer is the operational layer that optimizes optimal operating condition setpoints off-line,and the lower layer is model predictive control(MPC)layer that tracks the given setpoints accurately.However,this vertical two-layer structure can only realize sub-optimal global operation performance,and we need to consider optimization and control together online to obtain the optimal global operation performance of the system.This thesis investigates a novel optimizing control strategy to realize the optimal global operation performance of the system and proposes efficiency-oriented model predictive control(Efi MPC)strategy that has a nested twolayer(one inner layer and one outer layer).The research is supported in part by the National Natural Science Foundation of China under Grant 61973337,and in part by the Ministry of Education-China Mobile Research Fund major project: No.MCM2020—J-2.The main contributions of this thesis are as follows:1.In order to investigate the Efi MPC,the connection between the operational layer and the process control layer should be considered online,which puts forward high requirements toward the online optimization.Limited by the “No Free Lunch Theorem”,existing metaheuristics can only perform well on parts of the optimization problems and they cannot satisfy the requirements of the receding horizon optimization;thus,a new kind of metaheuristic algorithm,named optimal stochastic process optimizer(OSPO),is proposed in this thesis.OSPO algorithm is an adaptive metaheuristic algorithm,and it can control the exploration and exploitation properties adaptively online when solving different optimization problems to find the global optimal solution.The adaptive property of OSPO is realized with the help of the subjective probability density function(SPDF),preferential sampling and receding sampling strategy proposed in this thesis.2.Although a metaheuristic algorithm is suitable to be the solver of the Efi MPC,the most popular solvers used in the traditional nonlinear model predictive control(NMPC)are usually deterministic methods,and there still lacks a systematic comparison between deterministic methods and metaheuristic algorithms to test their performances as online solvers.A test platform,named N-GKLS,is proposed in this thesis,and its main functions are twofold: 1)With the help of the control Lyapunov function,N-GKLS addresses the four problems existed in traditional GKLS,and NGKLS can randomly generate NMPC control problems based on eight user-defined parameters which makes the comparisons between different solvers easier;2)With the help of the proposed optimal replace method,accumulated operational zone and auxiliary comparative table,the closed-loop performance,such as the global operation performance of the system,controlled by comparative optimization algorithms can be compared directly based on the generated NMPC problems.3.Existing MPC control strategies can only realize sub-optimal control of the global operation performance of the system;in order to realize the optimal global operation performance,this thesis investigates a novel optimizing control strategy,and the core innovative contribution is to propose efficiency-oriented model predictive control(Efi MPC).The specific innovations are threefold as follows: 1)define concepts of optimization margin and optimization efficiency,and convert the abstract notion of optimizing global operation efficiency into the direct optimizations toward optimization margin and optimization efficiency;2)the global process performance has been divided into two parts: the local process performance and a terminal truncation term,and the online optimization of the global operation performance can be realized based on these two concepts.The traditional vertical two-layer structure has been abandoned,and a new nested structure has been come up with to consider optimization and control together online.In this nested structure,the outer layer aims to optimize the local process performance,and the inner layer aims to optimize the terminal truncation;in this way,the local economic performance and the global economic performance have been connected directly.3)a periodic approximation(PA)technique has been proposed to estimate and optimize the global operation efficiency based on a finite horizon periodic operation,and the first type Efi MPC(Efi MPC1)algorithm is then constructed based on this PA technique.The proposed Efi MPC1 is suitable for solving problems where the controlled system has a periodic operation dynamic.4.This thesis further proposes two additional algorithms named the second type Efi MPC(Efi MPC2)algorithm and the third type Efi MPC(Efi MPC3)algorithm,where Efi MPC2 aims to solve problems that the controlled system has a given optimal steady state,and Efi MPC3 aims to solve problems that the controlled system has a given reference trajectory.1)The Efi MPC2 first proposes a multi-step second-order prediction concept to generalize the PA technique into the relaxed PA(r-PA),and then Efi MPC2 constructs a terminal region based on the given optimal steady state according to the zone optimizing control concept.Finally,Efi MPC2 uses a nested structure to realize the optimal global operation performance within a neighborhood of the optimal steady state.2)Unlike traditional tracking NMPC which directly optimizes the tracking performance towards the reference trajectory and the realization of the global operation performance is only an indirect optimization result,Efi MPC3 puts the tracking control performance and the global operation performance into a nested structure.In this way,Efi MPC3 directly optimizes the global operation performance,and Efi MPC3 realizes the optimal operation performance based on an acceptable tracking control performance.Finally,an intelligent machinery autonomous driving case is implemented to demonstrate the effectiveness of the proposed efficiency-oriented MPC.