The Balancing Neumann-Neumann preconditioner is well-believed to be efficient to speed up the convergence. However, this method faces a difficulty of the coarse problem: with an increase in Degrees of Freedom (DOF), the size of the coarse problem increases accordingly and becomes difficult to solve eventually. To overcome this, an Incomplete Balancing Domain Decomposition (IBDD) preconditioner is developed in this work. The algorithm is implemented in parallel by the Hierarchical Domain Decomposition Method (HDDM), and it is feasible to solve large scale models of over 100 millions DOF.