伴随水电规模的扩大,水电站群优化调度的计算量不断增加,需要探求新的方法。在分析离散微分动态规划(discrete differentiation andd ynamic programming,DDDP)算法的基础上,提出了基于分治模式的梯级水电站长期优化调度的细粒度并行离散微分动态规划(parallel discrete differentiation and dynamic programming,PDDDP)方法,并以澜沧江梯级的6个电站系统长期优化调度问题为应用实例,在多核计算环境下进行验证。结果表明,多核环境下的PDDDP方法简便易行,能充分利用闲置计算资源、大幅度提高优化调度的计算效率,是解决大规模复杂水电系统调度的高效和实用方法。
With the rapid development of the hydropower system in China, it is necessary to seek for new methods to improve the computing efficiency of hydro scheduling optimization. A fine-grained parallel discrete differentiation and dynamic programming (PDDDP) algorithm for long-term optimization of cascade hydropower station was proposed, which based on the analysis of normal discrete differentiation and dynamic programming algorithm (DDDP). The proposed algorithm was practically tested in the hydropower system in Lancangjiang River, which have 6 stations, and was implemented in multi-core calculation environments. Results demonstrate the PDDDP algorithm is easily implemented and greatly improves the computing efficiency due to making full use of parallel resources. It's a method with high efficiency and feasibility for solving optimal operation of large-scale cascade hydropower system.