The big question is whether agents in the first period anticipate the two choices of the central bank in the second period. If not and there is simply a surprise policy change in period 2, you can use the unstable Dynare version (6.x) together with perfect_foresight_with_expectation_errors_solver. See e.g.