基于一种改进稀疏动态慢特征分析的高速公路服务区空调空气处理单元故障检测研究  

Fault Detection of Air Conditioning Unit in Expressway Service Area Based on an Improved Sparse Dynamic Slow Feature Analysis

作　　者：宋圆圆 刘雪菲 

机构地区：[1]山东正晨科技股份有限公司,山东 济南

出　　处：《计算机科学与应用》2024年第5期94-107,共14页Computer Science and Application

摘　　要：高速公路服务区暖通空调空气处理系统表现出很强的双向动态特性,为了通过处理空气处理系统的双向动态特性和对提取的潜在变量施加稀疏性,本文提出一种改进的稀疏动态慢特征分析策略来检测空气处理系统的故障。在提出的稀疏动态慢特征分析中,采用自回归移动平均模型来揭示变量之间的自相关关系。然后应用多路数据分析,通过将扩充的三维数据集转换为展开的矩阵,计算出在多个批处理运行中的分批动态特性。进一步建立动态慢特征分析模型充分处理批运行中的时间动态特性。最后,融入特征稀疏表示技术,通过对负载向量进行稀疏约束,消除了无意义变量之间的耦合。在ASHRAE研究项目RP-1312实验数据集上进行的案例研究验证了所提出的故障检测方案的有效性。HVAC air handling systems in expressway service areas exhibit strong bidirectional dynamics. In order to detect the faults of the air handling system by dealing with the bidirectional dynamics of the air handling system and imposing the sparsity on the extracted latent variables, this paper proposes an improved sparse dynamic slow feature analysis strategy. In the proposed sparse dynamic slow feature analysis, an autoregressive moving average model is employed to reveal the autocorrelation among the variables. Multi-way data analysis is then applied to handle the batch dynamics over multiple batch runs by converting the augmented 3D dataset into an unfolded matrix. A dynamic slow feature analysis model is further established to fully deal with the time dynamic characteristics in batch operation. Finally, feature sparse representation technology is incorporated, and the coupling between meaningless variables is eliminated by sparsely constraining the load vector. A case study conducted on the ASHRAE research project RP-1312 experimental dataset verifies the effectiveness of the proposed fault detection scheme.

关 键 词：故障检测 空气处理系统 特征稀疏表示 慢特征分析 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]