This paper proposes a decentralized alternating direction method of multipliers (ADMM) algorithm for solving the optimization problem of energy scheduling in microgrids. Different from the other ADMM-based distributed approaches which decomposes the monolithic problem spatially into smaller tractable subproblems, the proposed method adopts a temporal decomposition on the scenario tree inherited from multi-stage stochastic programming. Each node in the scenario tree serves its own optimization with local variables and constraints, and iteratively updates information with adjacent nodes. By implementing the proposed ADMM algorithm in a rolling fashion, simulation results have shown the fast convergence of the temporal distribution framework, and comparisons on optimal value and computation time with other optimization approaches reveals its advantages and effectiveness.