基于密度权重支持向量数据描述的冷水机组故障检测

Chiller fault detection by density weighted support vector data description

虚警率(FAR)是评价冷水机组故障检测性能的关键指标,用户无法接受过高的FAR。为了降低支持向量数据描述(SVDD)在冷水机组故障检测时的FAR,将密度权重集成到SVDD中,提出了一种基于密度权重支持向量数据描述(DW-SVDD)的冷水机组故障检测方法,该方法考虑了样本数据在真实空间中的密度分布情况。使用ASHRAE RP-1043冷水机组实验数据对提出的方法进行验证,并将检测结果与传统SVDD的冷水机组故障检测方法进行比较。结果表明,提出的方法将FAR从10.5%降低到7%,同比下降超过了30%,同时对4个劣化等级下的7种典型冷水机组故障有着优良的检测性能。

False alarm rate（FAR） is a key indicator to evaluate performance of chiller fault detection methods, since customers cannot accept high FAR. In order to reduce FAR of support vector data description（SVDD）-based chiller fault detection, a density weighted support vector data description（DW-SVDD）-based chiller fault detection method was proposed by integration of density weight into SVDD with a consideration of density distribution of sample data in real space. The proposed method was validated with experimental data of RP-1043 ASHRAE and detection results were compared to those of traditional SVDD chiller fault detection methods. The results showed that the new method could reduce FAR from 10.5% to 7%, which was lowered about 30%, and had excellent detection performance for 7 typical chiller faults at 4 severity levels.