基于云计算和量子粒子群算法的电力负荷曲线聚类算法研究

Research of power load curve clustering algorithm based on cloud computing and quantum particle swarm optimization

针对电力数据海量化、多维化的趋势,为了提高聚类算法的聚类质量,并解决传统聚类算法聚类海量高维数据时单机计算资源不足的瓶颈,提出了一种基于云计算的电力负荷曲线聚类的并行量子粒子群优化模糊C均值聚类算法。将量子粒子群群体智能算法引入到传统模糊C均值聚类算法中,利用QPSO较强的全局搜索能力,克服FCM算法易陷入局部最优以及其对初始聚类中心过于敏感的缺陷。最后,采用云计算的MapReduce编程框架以及HBase分布式数据库对算法进行并行化改进。经实验验证与FCM算法和AFCM算法相比聚类正确率提高了10%左右,且并行性能较好。

Encountering the trend of massive amd multidimensional data, in order to improve the quality of clustering algorithm and solve the computing resources bottleneck of traditional clustering algorithm when clustering massive amounts of high dimensional data, this paper proposes a Parallel Quantum-Behaved Particle Swarm Optimization Fuzzy C-Means clustering algorithm based on cloud computing for power load curve clustering. Quantum particle swarm intelligence algorithm （QPSO） is introduced into the traditional Fuzzy C-Means （FCM） clustering algorithm, QPSO‘s stronger global search ability is used to overcome FCM algorithm‘s weakness of falling into local optimum easily and being sensitive to initial clustering center. Cloud computing is adopted in the MapReduce programming framework and HBase distributed database to parallelization algorithm is improved. Many experiments verify that compared with traditional FCM algorithm and AFCM algorithm the clustering accuracy is increased by about 10%with better parallel performance.