Research On Multi-objective Optimization Problem Of Demand-Driven CSPS System

With the advancement of information technology and the development of economic,corporate production is increasingly affected by fluctuations in customer demand.This means that production scheduling at the production center should be combined with inventory management at the sales center under conditions where customer demand is uncertain.Demand-driven production control becomes even more necessary.Therefore,this dissertation satisfies the customer’s needs and reduces the inventory cost as the optimization goal,study the multi-objective optimization problem of the demand-driven CSPS system,the main work is as follows:Existing CSPS system generally determines the behavior of the system through lookhead during production,in order to meet the customer’s demand and effectively reduce the inventory cost of the product bank,this dissertation proposes an inventory control method based on the existing research.According to the physical model and working mechanism,the dissertation models the system as a semi-Markov decision process,and uses the linear weighted policy iterative algorithm to solve the Pareto optimal solution set of the problem.The simulation results verify the effectiveness of adding inventory control.However,the linear iterative policy iterative algorithm is slow to solve,and only the optimal solution on the Pareto frontal convex hull can be obtained.In order to improve the efficiency and optimization effect,the dissertation uses two multi-objective evolutionary algorithms,CMOPSO and NSGAII,to solve the problem,and gives two encoding methods: binary-real number coding and continuous real number coding.The simulation results show that the multi-objective evolutionary algorithm can solve the problem effectively,and the CMOPSO algorithm with continuous real-coded can achieve better optimization results under different examples.In order to further improve the quality of the Pareto approximation solution set,a BCEIB-CMOPSO algorithm is designed.The evolutionary population of CMOPSO is divided into two populations by bi-criterion evolution framework.The algorithm applies the indicator-based criterion and the Pareto criterion in the individuals selection criterion of two populations respectively.At the same time,for the characteristics of the system a directed replication mutation operation is designed to improve the optimization performance of the multi-objective evolutionary algorithm in solving this problem.Finally,the effectiveness of the improved algorithm and the proposed mutation operation is verified by experiments.