Shengli Liao

Long-Term Generation Scheduling of Hydropower System Using Multi-Core Parallelization of Particle Swarm Optimization

A multi-core parallel Particle Swarm Optimization (MPPSO) algorithm is developed to improve computational efficiency for long-term optimal hydropower system operation, in response to rapidly increasing size and complexity of hydropower systems, especially in China. The MPPSO can be implemented in three steps with easily accessible multi-core hardware platforms. First, a multi-group parallel computing strategy is introduced to maintain the diversity of population for finding the global optima. Second, the fork/join framework based on divide-and-conquer strategy is adopted to distribute multiple populations to different CPU cores for parallel calculations to take full advantage of CPU performance. Third, the results generated in different CPUs are merged to achieve an improved acceleration effect on computational time cost and more accurate optimal scheduling solution. Results for a system of twelve hydropower stations in the Guizhou Power Grid in China demonstrate that the proposed algorithm makes full use of multi-core resources, and significantly improves the computational efficiency and accuracy of the optimal solution, in addition to its low parallelization cost and low implementation cost. These suggest that the proposed algorithm has great potential for future optimal operation of hydropower systems.

Some Practical Strategies and Methods for Large-Scale Hydropower System Operations in China

With the rapid increase of number and capacity of hydropower plants operated by single dispatching departments in China, more attention should focus on seeking more robust methods for reducing dimension curse, improving effectiveness and practicability of optimization results, and enhancing computational efficiency of large-scale complex hydropower system operations. In this paper, a general solution framework for large-scale complex hydropower system operations is presented from the real hydropower systems in China. The framework consists of intelligent strategies to reduce problem size during modeling, integrated optimization methods and search methods to vanquish dimensionality difficulties effectively and cope with complicated spatial-temporal constraints, as well as interactive interfaces to adjust optimal results. Two case studies are presented.