Abstract: How to make use of efficient clean hydropower for peak-valley adjustment is a key technique to release peak regulation pressure under the circumstance of increasingly urgent demand for peak shaving. A short-term peak shaving model for cascade hydropower stations has been developed in this study considering the minimum of residual load variance. A hybrid progressive particle swarm optimization algorithm is also described herein, combining a progressive optimality algorithm (POA) in the outer layer for compressing solution space and reducing dimension with a particle swarm optimization (PSO) in the inner layer for searching continuously and randomly. To update feasible regions adaptively, a new dynamic corridor technique was used for handling complicated time-space coupling. Application of this model to four stations on the Yuan River shows that the residual load variance of Hunan Grid can be decreased by 93.6% while keeping a steady and smooth curve of residual load. The method is very effective and useful in short-term peak shaving for cascade hydropower stations.