Topology Optimization For Long-Reach Passive Optical Network

Nowadays, China is experiencing an upgrading period of transitions from3G to4G as for the rapid development of communication networks. To facing the challenges of "last mile" and FTTx, as well as constructing the infrastructure of cloud computing, the State Council of China proposed the "Broadband China" strategy in August,2013which aimed to enhance strategic guidance and systematic deployment. Passive Optical Network (PON) has been reported as a promising scheme to support the "Broadband China" strategy and to promote the deployment of broadband network. However, traditional PONs have several limitations, such as small user scales, limited transmission distance and high operations and maintenance cost.Nest Generation PON (NG-PON) has been regarded as one of the most suitable optical network techniques to achieve the "Broadband China". The thesis focuses on Long-Reach PON (LRPON), which is the most promising technology of NG-PON. The main contributions of the thesis are summarized as follows:Firstly, we focus on comparing the characteristics of tree-and-branch and ring-and-spur topology of LRPON. Then the formulations of the two topologies for LRPON-Tree-Topology-Optimization (LTO-T) problem and LRPON-Ring-Topology-Optimization (LTO-R) problem are proposed, respectively. The idea of "flow" is applied to ensure the integrality and uniqueness of the network paths in the problem (LTO-T) formulation. In the problem (LTO-R) formulation, we solve the combined optimization of the placement and ring-structure problem of Optical Add-Drop Multiplexer (OADM) and the problem of OADM-ring deployment, in which a novel idea of "double flow control" is proposed.Secondly, the (LTO-T) model and (LTO-R) model are optimized using Gurobi Optimization Solver. The expriments demonstrate the feasibility of the two models by numerical scenarios and analyze the effects of deployment costs and network performances of (LTO-T) model and (LTO-R) model.Finally, wew propose a novel heuristic algorithm called Fast-Backward-Seeking (FBS) to reduce the high complexity of the problem (LTO-T) formulation. The FBS algorithm quickly optimizes the tree-and-branch topology deployment of LRPON and outperforms Gurobi in terms of efficiency. And there is only a small gap of less than10%between the results of FBS and (LTO-T) optimum which is acceptable for practical network deployment in real world.