An Optimization Strategy For Supply Chain Operation Based On Model Predictive Control

In today's complex market environment, the changing customer demands and the fierce competition stress require agile responses of all supply chain members. The resulting decision-making process is not such a simple job to supply chain managers. A lot of operational factors, restrictions, and various dynamic factors need to be considered together to make the consignment arrangement and production schedule. Thus an advanced and practical optimization scheme for supply chain operation and a corresponding decision-making support tool are strongly needed to provide a guideline for supply chain managers.Model predictive control (MPC) has gained notable successes in traditional industrial application. Its robustness and rolling horizon optimization strategy makes itself suitable for the optimal control and decision-making process of complicated systems. However, supply chain system is much more complicated. The key difficulty of this research is how to induct the theory of MPC into the supply chain application successfully.In this paper, we research on the decision-making support issue in supply chain operation based on the model predictive control, and propose an integrated optimization scheme for supply chain on-line operation. The key points conclude:1) Most of the present supply chain models are only qualitative or partial. In this paper we research on a dynamic quantitative model of the whole supply chain system to obtain an integrative decision-making solution schema. Futhermore, we make a revision to the cost function and optimization objective to ensure that the optimal solution minimizes overall costs, while satisfies a certain customer demand satisfactory (CDS) level and constraints;2) We accomplish the model predictive control algorithm design and implementation for supply chain operation based on the integrated model and the linear cost function without penalties to the control variables variation. A GAMS simulation framework is established to prove the efficiency of this algorithm. Simulation results illustrate that the application of MPC strategy brings in great improvement on both profit and CDS level magnification with determinate demands as well as uncertain demands. Also the algorithm shows great robustness.3) In order to confront demand uncertainty, we induct the Value-at-Risk (VaR) inventory management strategy and add the forecast bias into the inventory calculation, thus CDS level can be handled without accurate demand forecasting, and also the stock risk can be reduced. Futhermore, we bring the VaR inventory management strategy into the overall supply chain optimization, enhance the model predictive control strategy, and project a general dynamic solution schema for supply chain operation under uncertain demands based on the enhanced MPC. The GAMS simulation results illustrate that the enhanced strategy is proved to be more effective on profit and customer satisfaction improvement under uncertainty, with profit magnification increases of up to 59.1% in scenario 2(section 5.4.2). Then the paper compares the robustness of MPC and enhanced MPC, and shows the latter behaves a persistent ascending trend as uncertainty increases bit by bit.