基于随机森林的风机状态监测数据可视化研究

Data visualization research on monitoring data of wind turbines based on random forest

针对风电机组状态监测数据可视化存在的直观性不强和交互性差等缺陷,提出基于随机森林的可视化技术。即首先对监测数据进行基于RF的相似性度量,使数据在新特征空间的类可分性增强;然后采用主成分分析法进行特征变换与降维,将多维数据的关系信息变换到适合人类视觉认知的低维空间里;最后对数据在低维空间里采用散点图和平行坐标图进行可视化展示。实验结果表明,风机状态监测数据经过RF处理后,可视化效果良好,便于管理人员从整体上把握数据的集中特性、分布规律以及属性间的关系等信息,对提高风电机组的运行可靠性具有重要意义。

The current monitoring data visualization of wind turbines suffers from poor interactivity and less intuitive- ness. In this paper, a visualization technology based on random forest is proposed. Firstly, the random forest was used for similarity measurement on monitoring data, which enhanced the separability in the new feature space. Then, the principal component analysis （PCA） is adopted to reduce the dimension, through which the relationship of the multi- dimensional data information is transformed into low dimensional space for human visual perception. Finally, the data in low dimensional space using a scatter diagram and parallel coordinates figure is displayed. The experimental results shows that the condition monitoring data of wind turbines processed with the random forest, has a good visual effect and it＇ s easy for the wind turbines manager to figure out the data characteristics, distribution, development trend and the relationship between attributes on the whole grasp. It＇ s of great significance to improve the running reliability of wind turbines.