The Research On The Interval Predictive Control Model And Algorithm Of Multi-variable Systems

Many industrial practical systems are complex and multi-variable. Because of thecomplexity of multi-variable systems themselves and environment, the existence ofvarious constraints, and the requirements that people want to improve the performance ofsystems in recent years, the traditional optimal control methods are being challenged. Asan advanced optimal control technology, the predictive control algorithm has importanttheoretical and practical significance for the good control of complex multi-variablesystems. Based on the predictive control theory and the nonlinear optimization theory,and taking the multi-variable system as the research object, the topic studied theoptimization of the control model, the reduction of the algorithm’s complexity and therealization of some algorithm’s performance. The specific studies are as follows:(1) For the requirments that people want to improve the performance of the systemand realize the interval control of the cotrolled variables, the interval predictive controlmodel of the multi-variable system is established. Considering the handling of the controlmodel’s constraints, the coordinative control of the variables, and the optimization of theeconomic indicators, we use the interval predictive constructing method to establish theconstraint item, control item and economic item. On this basis, the restrictive relationshipof each item and each variable is analyzed, the multi-variable coordinating optimizationmethod and the variable’s principle of priority are used to give a model parameters’control strategy of the interval predictive control, providing basis for the realization of thesystem’s good control in the optimization solving process of the control model.(2) For the constrained feasibility of the control model, the judgment theorem andthe treatment program of the soft constraints under the infeasible case are proposed.Considering the case that the constraints which can not be simultaneously satisfied willlead to the emergence of the infeasible solution, we use the nonlinear programmingmethod to propose the feasibility judgment theorem of the consrained control model. Onthis basis, the relationship of each variable and each constraint is analized, the adjustmentamounts of constraints are obtained by solving a nonlinear programming problem, and the treatment program of the soft constraints under the infeasible case is given.Furthermore, considering the interval control requirement of the controlled variables, weconvert the constrained control model that has been given soft constraints’ adjustmentinto the interval predictive control model, providing feasible program for the assurance inthe constrained feasibility of the interval predictive control model.(3) For solving the interval predictive control model, a boundary feasible sequentialquadratic programming algorithm is proposed. Considering the computational speed, thefeasibility and the convergence of the algorithm, we respectively start from several issuesthat the selection of the initial point, the simplification of the constraint set, the solvingacceleration of the search direction, the establishment of the feasible descent direction,the establishment of the high-order correction direction, and the assurance of the Hessianmatrix’s positive definiteness, to research the design of the boundary feasible sequentialquadratic programming algorithm. Then, we obtain the optimal solution that meets thecertain accuracy by solving the interval predictive control model in short time and underless computation, providing basis for the fast and accurate control of the system.(4) Considerig the parameters’ automatical adjustment problem of the intervalpredictive control model, we use the nonlinear function constructing method to design theparameters’ adaptive adjustment strategy. Furthermore, in order to illustrate the designidea of the whole algorithm more clearly, the technology roadmap is given. Finally, asimulation platform is built, MATLAB is used to program. Experimental testing,comparative analysis and performance evaluation are given to verify the feasibility,validity and some other performance of the algorithm proposed in this paper.