Abstract: The traditional differential evolution (DE) could fall into local optimum easily when used for global optimization. To overcome this disadvantage, we herein present a new algorithm, namely the multicore parallel chaos simulated annealing differential evolution (PCSADE), for solving the optimization models of long-term operation of cascaded hydropower stations. This algorithm takes advantages of the chaos theory and the simulated annealing algorithm (SA) to enhance the DE algorithm and the Fork/Join parallel framework for higher computation efficiency. Particularly, it can make full use of the chaos theory's merits in handling strong randomness and ergodicity and the good local search ability of the SA algorithm to enhance its global search ability, and apply the chaos theory to the generation of initial populations and the dynamical adjustment to the DE algorithm's control parameters. And it replaces the selection operation of DE algorithm by a metropolis rule of the SA algorithm. Moreover, it can adopt the Fork/Join parallel framework for parallel computation of a complex simulation task. Application to the 14 hydropower stations on the Hongshui River show that PCSADE can make full use of the multicore resources of computers and improve the solution and efficiency significantly. Thus, our new algorithm is an effective and feasible method for long-term operation of hydropower stations.