Planning And Optimization For Optical Network Based On Evolutionary Algorithms

With the development of network, the number of Internet users rises sharply, and the Network communication is developing toward big data and diversification. To meet the growing demand for bandwidth of users in any region, we not only need to study on the long-reach and wide coverage network topology planning scenarios, but also need to study on the network transmission optimization mechanism which is high bandwidth capacity, high resource utilization, and resource allocation flexible.This thesis describes the optical network planning and optimization on two aspects:topology and spectrum. It selects a representative long-reach passive optical network (PON) and spectrum-sliced elastic optical network (SLICE) as the objects of study.The main innovations are as follows:(i)In terms of the optical network topology planning, this thesis investigates the splitter cascade scenario in long-reach PON, and proposes the problem to establish the minimum cost optical network topology considering the physical constraints and management needs. Then this thesis establishes the mathematical model, and develops the nested GA (genetic algorithm) to solve the problem. Finally, this thesis analyzes the influence of user distribution and power budget on the topology and cost. The algorithm uses a nested way and a hierarchic optimization strategy, greatly narrowed the search for the solution space in the case of the huge solution space. And the introduction of "peer pressure" and "parent-child competition" mechanism in the algorithm, can solve the non-convergence problem. When building the optical network, the Prufer sequence is applied. It can not only guarantee a tree topology, but also meet the different requirements of each node category directly.(ii) In terms of the optical network spectrum optimization, this thesis investigates the routing and spectrum allocation for dynamic multicast traffic in SLICE network, and proposes the problem to establish minimum coding cost routing and allocate spectrum with the minimum fragmentation effects, while the QoS constraints and spectrum constraints are met. Then a mathematical model is established based on the routing process with network coding and spectrum allocation process with fragmentation effects. The aim is to seek the combination of bandwidth efficiency and flexibility by network coding, spectrum sliced and resource allocation. In the algorithm, this thesis uses PBIL (population based incremental learning) algorithm to solve the problem of network coding with minimal cost. PBIL algorithm combines the advantages of genetic algorithms and machine learning and can search the optimal solution quickly and accurately. A restart mechanism is added in the algorithm to prevent the convergence to the local optima. This thesis uses SPV (spectrum constrained path vector search) algorithm to add the spectrum effects in the routing process, so it can reject the route which don’t meet the spectrum conditions, and avoid its subsequent operation. This method can greatly enhance the computational efficiency of the algorithm. Finally, this thesis uses the fragmentation-based spectrum allocation algorithm. this thesis makes the fragmentation minimum in the current stage of the network business spectrum allocation, which is conducive to the subsequent business and reduce the defragmentation process.