Research On Distribution Decisions Based On Bi-level Chance Constrained Programming Under Uncertainty

Distribution decisions are critical in enterprises,which determine the market distribution of products,affect many other operation decisions and directly affect the realization of corporate profit targets.A reasonable distribution decision must take into account the uncertainties and the relevance among decisions.With the goals of minimizing costs and considering the production decisions reaction simultaneously under the condition of uncertainty in production and random demand,distribution decisions problem based on bi-level chance constrained programming under uncertainty is proposed in the thesis.Firstly,a bi-level programming model with random variables is constructed which consists of upper distribution model and lower production model after detailed description of distribution.Then the mathematical model is transformed into a bi-level chance constrained programming model with theories of deterministic equivalence and chance constrained programming.Then two algorithms are designed to solve the bi-level chance constrained programming model for distribution decisions jointly.The general location of the global optimal solution is searched by improved genetic algorithm,and then as the initial feasible solution for the improved pattern search algorithm.The method can approximate optimal solution as much as possible.The calculation of lower production model is nested in the objective function value calculation of upper distribution model.And the random variables are processed by Monte Carlo stochastic simulation in both algorithms.Furthermore,the numerical results show that the time of calculation for improved genetic algorithm with global optimization is longer than the one for improved pattern search algorithm with local optimization.And the global optimization is found through the joint calculation of the two algorithm.Finally,the innovation and prospect are summarized.