Model reduction of large-scale structures are improved and a weight coefficient reflecting the contribution proportion of a higher-order model is introduced in light of the shortcomings of conventional optimization algorithms, with the aim to solve the problem that conventional optimization algorithms do not serve the optimized distribution of large-scale structure observation stations. One hybrid optimization algorithms is proposed based on the effective independence method. The observation point distribution effects drawn from the modal strained energy coefficient method based on the effective independence weighted average are compared with that drawn from the effective independence methods for Guyan reduction based on modal strained coefficients and the respectively improved ones through a rocket simulation experiment. Results show that the algorithms effectively avoid the emergence of concentrated observation stations, and ensure the contribution of all modal strain energy and requirements that the better-arranged observation station has much more strained energy. Modal tests based on the method are carried on real GARTEUR plane, which show that the algorithms guaranteed the completeness and linear independence of monitoring mode.