For the purpose of a parallel process for large scale systems, a balance of each processor tasks distributed in large-scale systems has to be achieved, that is, subsystems of the system should have some kind of balanced structures after decompositions, or the dimensions of subsystems are of similar size. Based on block triangle decomposition, the coefficient matrix A of linear system state equations can be of block diagonally dominant by using imbalanced compensating scheme; therefore, the ideal result of system model simplifications is obtained. It is proved that the given block diagonally dominant decomposition of large-scale systems is feasible by simulations, which presents a new system model simplifying method for the analysis and design of large-scale system decentralized controls.