Abstract: Based on the analysis of factors which influenced the operational reliability of Air Conditioning (AC),aiming at the minimum difference of operational reliability of AC in similar regions,a multivariable high-dimensional clustering model of operational reliability on the basis of working environmental factors and user habit factors was established,and a dynamic optimal-selection cellular genetic clustering algorithm for solving the model was proposed.By introducing information entropy and optimal-selection-based strategy into canonical cellular genetic algorithm and fuzzy C-means,as well as using the new dynamic crossover and two-combination mutation operations,the algorithm integrated the merits of fuzzy C-means clustering method's fast convergence speed and cellular genetic algorithm's good diversity and strong capability of global search.Through testing six benchmark datasets,the proposed algorithm was proved to have better clustering accuracy and stability than FCM,GA-FCM,FCM-IDPSO.It was especially suitable for solving clustering problems of high dimensions.The proposed algorithm was applied to solve the model and evaluate the validity of granularity partition results,and then the optimal partition scheme was determined.It could effectively solve the relevant practical engineering problems.

Key words: operational reliability of air conditioning, clustering algorithm, cellular genetic algorithm, fuzzy C-means, granularity evaluating

摘要： 为了进行使用可靠性区域粒度划分研究,在分析空调使用可靠性影响因素的基础上,以其使用可靠性同类区域差异最小为目标,建立了使用可靠性基于工作环境和用户使用习惯两类影响因素的多变量高维聚类模型,提出求解该模型的一种动态优选元胞遗传模糊聚类算法。该算法在经典元胞遗传算法和模糊C-均值算法的基础上引入信息熵理论和优选策略,并采用动态交叉和两阶段变异算子,因此集成了模糊C-均值收敛速度快和元胞遗传算法在解决复杂问题时多样性好、全局搜索能力强的特点。通过6个标准测试数据集的测试结果,证明新算法相对于模糊C-均值、遗传模糊聚类算法和粒子群模糊聚类算法具有更高的聚类精度和稳定性,尤其适合处理高维复杂数据的聚类问题。最后运用该算法求解模型,并评价不同粒度层次下聚类结果的有效性,进而确定使用可靠性最优区域粒度划分方案,表明算法能有效解决相关的实际工程问题。

关键词: 空调使用可靠性, 聚类算法, 元胞遗传算法, 模糊C-均值, 粒度评价