运用基于大数据处理架构的Naive Bayes分类方法提出了暂态电能质量评估方法, 将数据来源扩展至电网运行监测数据、 电力用户数据和公共信息数据等方面, 并将评估结果按严重程度分为暂态正常状态、 短时电压暂降状态、短时深度电压暂降状态、 短时电压失压状态. 基于MapReduce架构, 设计分布式Naive Bayes算法实现状态分类. 在分类器训练阶段, 对海量历史数据进行分布式学习, 周期性地生成评估规则库并部署到所有评估节点. 在状态评估阶段, 各评估节点基于流处理框架快速生成实时评估样本, 并根据当前规则库实时地得出评估结果. 试验结果表明, 所提出的基于大数据分析的暂态电能质量评估方法是可行, 在准确率和处理速度上都取得了较好的效果.
A transient power quality assessment method is proposed based on Naive Bayes classification in the architecture of big data processing. The data sources are extended to power grid monitoring data, power customer data and public data, and the assessment se-verities are classified into normal state, abnormal state, critical state, and failed state according to the results of Naive Bayes classifi-cation. A Naive Bayes classification method based on MapReduce to realize power quality assessment is designsed. In the classifier training phase, massive historical data are used as the distributed learning object, and assessment rules are generated periodically. In the state assessment phase, each assessment node updates the assessment rules generated by the training phase, generates real-time e-valuation of the samples from the stream processing framework, and evaluates the power quality state according to the current rule. Ex-periment results show that the transient power quality evaluation method based on big data analysis presented in this paper is feasible, and achieve good results both in classification accuracy and processing speed.