Abstract:Taking network cost and nodal reliability as optimization object and restriction, a network topology optimization model is presented for the aim to achieve the best topologies of lifeline networks under earthquake. In order to speed up the optimization process, the element investment importance is introduced based on recursive decomposition algorithm. As this model is a typical combinatorial optimization problem, three approaches, genetic algorithm (GA), simulated annealing algorithm (SAA) and simulated annealing genetic algorithm (SAGA), are used to solve it. When GA is used, a generation including many genes is initially created with each gene representing a network. Then by using selection, crossover and mutation operators, a new generation is evolved. After a number of iterations or when some criteria are met, a near-global optimal solution could usually be found. SAA takes a network topology as its current solution and produce a new solution by perturbing. If the perturbation result is an improved solution, it is accepted and the current solution is updated accordingly. Otherwise, it can also be accepted at a probability. The perturbations and updates repeat until some criteria are met. Replacing the mutation operator in GA with perturbations and updates in SAA, SAGA is established to solve the optimization model. Moreover, two example networks are evaluated to compare the efficiency of these algorithms. The results indicate that SAA performs best.
