Study On Distribution Network Optimization Based On Ant Colony Optimization

With the argument of the pressure of market and diversification of demand of customs, many corporations have to reappraise their distribution strategy. The manufacture environment in which the corporations have turned the direction of increasing addtion value from manufacture to distribution and service. Thus, in the custom-driven environment, reasonable distribution network and strengthening the management of distribution network are important approch to improve customs' satisfication and buildup the competition power of corporation. But, we always come up against the Combination Optimization or non-deterministic polynominal problems (NP) in the Distribution Network Optimization (DNO), they are hardly come out by the traditional methods and many scholars turn to approximate approach or Simulating Biology Intelligent Algorithms. Ant Colony Optimization (ACO), which is introduced in this paper because of the ability of resisting combination explosion to NP-hard problem, now is used to solve the problem of DNO. The main results of this paper include the content below.Firstly, we apply ACO to solve the TSP and compared the accuracy of algorithm with Diagonal Algorithm, which is a traditional approximate approach, and compared the speed of algorithm with the traditional accurate approach. By data simulation and experiment, we need to pay very attention to choose the parameter and the application arrange of ACO, ie, the traditional methods are better when the n is little scale and ACO will be better when the n is very large scale.Secondly, ACO is used to solve the bi-level distribution network model in the DNO, put forward the calculate mode of ACO and the method of solution structure and concrete process of algorithm. At last, compared the result coming from method above with hybridize heuristic algorithm with conventional branch-and-bound method in literature [4]. The experiment represents, the method based on ACO in this paper can solve the NP hard problem in the DNO and can get better solution than conventional methods.Thirdly, the problem of reengineering the Distribution Network to reduce the cost includes time or expenses have been presented in this paper. Furthermore, We add the function of logistic moving and the information coordination function of manufacturing centers and put forward the triple-level model based on the bi-level for the sudden need or maladjustment of market need. Moreover, the two problem presented above are the same problem and can be solved by introduced the Steiner Tree. However, searching Steiner tree is important for optimizing the Distribution Networks, which is proved to be NP-complete. We simplify the model and make use of the improved ACO to search the Steiner Tree in the DNO. The results of experiment and comparison show that ACO and Steiner Tree are feasible to optimize the Distribution Networks.