One of the most important objectives in reactive power planning (RPP) is maintaining required voltage stability margin under normal and single-contingency situations. For this, an optimization problem must be formulated considering N-1 contingencies constrained for a set of predetermined contingencies. Traditionally, all these contingencies must simultaneously be addressed in the optimization problem. This can dramatically increase the problem dimension and decrease the convergence ability of its solution for large scale applications. This paper uses an L-index based contingency filtering strategy to select a small number of contingencies to be addressed in RPP problem. For this, the predetermined set of all contingencies is divided into some subsets wherein each subset of contingencies affects an especial part of power system. Then, only the worst contingency from each subset is selected and addressed in RPP problem so that the allocated VAr sources can cover all of contingencies. This method reduces the problem dimension and improves convergence ability of its solution for large scale applications, whilst the obtained results are similar to the ones taken from enumeration method of N-1 contingencies. The implemented simulations on the IEEE 39 and 118-bus test systems verify the capability of the proposed method.