将含有风电场和抽水蓄能电站的电力系统随机动态经济调度问题描述为随机型存储器问题，以考虑风电场出力所具有的随机性。该模型含数学期望的计算，且实际问题状态空间、决策空间都是高维的，难以对其准确求解。因此，采用近似动态规划算法将值函数近似表示成分段线性函数的逼近形式，从而将随机存储器问题转化为一系列多阶段线性规划问题。通过扫描误差场景并相应求解所对应的线性规划问题，可实现对值函数进行训练，逐次修正各分段斜率值，直至值函数收敛后，再用来对预测场景下的线性规划问题进行求解，即得动态经济调度结果。该算法避免了求解准确解时面临的“维数灾”问题，具有较快的计算速度。以含风电场和抽水蓄能电站的小型测试系统和某省级实际电力系统为例，验证了所提模型与算法的可行性与有效性。“,”ABSTRACT:To consider the randomness of wind farm’s output, the stochastic dynamic economic dispatch problem of power systems containing wind farms and pumped storage power stations was described as an stochastic storage problem. Except for that the expectation should be calculated in this model, the actual state space and decision space of the model are high dimensional and are difficult to solve accurately. So based on approximate dynamic programming algorithm, the value function was interpreted as a piecewise linear function approximately and the stochastic storage problem was converted into series of multi-stage linear programming problems. By scanning error scenarios, the slopes of the piecewise linear functions were trained successively until they were converged. Then the value function can be applied to solve the linear programming problem corresponding to forecast scenario and the dynamic economic dispatch results were obtained. This algorithm avoids the curses of dimensionality and has fast computational speed. The results on a small test system and an actual provincial system with wind farms and pumped storage stations demonstrate the feasibility and effectiveness of the proposed method.