System reliability allocation is an important ingredient in system reliability design, and it is also a decision-making issue of reliability engineering. To achieve the optimization of system reliability allocation, an optimization model for system reliability allocation, which takes the system cost as the objective function, is constructed through the general cost function. In order to overcome the shortcomings of particle swarm optimization (PSO) appearing in reliability allocation optimization, the premature and/or slow speed of convergence in later period, the sequential quadratic programming (SQP) was introduced to improve the PSO algorithm. The algorithm uses PSO as the global optimizer while the SQP is employed for accelerating the local search. Thus, the particles are able to search the whole space while searching for local optimization fast, which not only assures the convergence of the algorithm, but also increases the probability of obtaining the global optimum. Applied the algorithm to the problem of system reliability allocation, the simulation results show that it has excellent global search capability and provides rational optimization results compared to the existing approaches.