基于征兆探测的空调系统故障预示及其运行可靠度研究

Decreasing of operation reliability will result in the decrease of actual energy efficiency in air conditioning systems. However, there is still no suitable evaluation method for air conditioning systems to evaluate its current operation reliability. And there is few achievements yet to study symptom detection and fault prognosis before some faults occur. In this research, suitable evaluation method for operation reliability will be studied for air conditioning systems. Efficient symptom detection and fault prognosis methods will be presented for the system before sudden faults occur or at earlier period of gradual faults. Incipient pre-tolerance method for symptoms based on operation reliability evaluation will be developed in this project. Firstly, the semi-practicality simulation and real physical experiment systems are used to analyze and obtain the correlation characteristic among controlled variables, faults and operation reliability. Operation reliability is defined for air conditioning systems and relative energy consumption and thermal comfort expectation are presented. The evaluation index is confirmed through combining the relative energy consumption and the deviation from thermal comfort expectation. Then the evaluation model for operation reliability is constructed in air conditioning systems. Secondly, a method based on variable time domain and variable spaces is presented to detect symptom and prognose future fault. The incipient symptoms are detected through monitoring the correlation among variables in micro-window. The future possible fault is prognosed through tracing and analyzing the process characteristics in macro-window. Multiple sudden faults are quickly diagnosed and isolated using designated component analysis in variable spaces. Finally, a composite tolerance method including incipient pre-tolerance based on optimizing operation reliability is realized in air conditioning systems. This research will expand the research contents of fault detection and diagnosis in the air conditioning field. It can promote the application technologies of high efficient operation and energy savings in air conditioning systems.

运行可靠性下降会导致空调系统的实际能效降低，但领域内尚缺少运行可靠性的评价方法，而对故障前的征兆探测与故障预示也鲜有报道。本研究探索空调系统运行可靠度的评价机制，研究故障发生前的征兆探测方法、故障预示技术，发展基于运行可靠度评价的空调系统征兆预容错技术。首先，利用半/全实物实验台分析参数、故障与可靠性的关联特征。提出"运行可靠度"的概念，通过引入相对能耗与热舒适期望，确定了综合相对能耗与热舒适偏离的评价指标，建立了空调系统运行可靠度的评价模型。然后，提出不同时域-不同空间的方法，通过对微观窗口参数相关性的放大监视，实现对早期征兆的探测；通过对宏观窗口过程特征的跟踪分析，实现故障的预示；通过不同空间的指定元分析，完成多个突发故障的快速诊断与隔离。最后，利用运行可靠度评价模型，实现了基于最佳运行可靠度的复合型容错控制。本研究将丰富空调系统故障诊断的研究内容，对系统高效运行及节能具有促进作用。

故障检测与诊断最早应用于航天、核能和自动化等领域，而在建筑空调领域中的研究起步较晚。经过近年的发展，空调系统故障检测与诊断研究已由设备环节逐渐发展到系统层面，国内外在该领域已取得一定成果，但是对运行可靠性和故障的征兆探测与预示却鲜有报道。本研究探索了空调系统运行可靠度的评价机制，研究了故障发生前的征兆探测方法以及故障预示技术，提出了突发故障的快速诊断和隔离方法，发展了基于运行可靠度评价的空调系统故障容错控制技术。首先，搭建了基于“舒适性-能耗”协同仿真的空调系统可靠度研究实验平台，分析了热舒适性和能耗与系统运行可靠度的关联，研究了传感器安装故障对空调系统运行可靠度的影响；其次，提出了控制完善度（Control Perfect Index, CPI）可靠度评估指标，建立了基于CPI指数的运行可靠度评估方法，并结合“舒适性-能耗”协同仿真构建了运行可靠度综合评价机制；然后，将实际控制器引入“舒适性-能耗”协同仿真平台，搭建了“硬件在环”实物控制器实验台，并开展了多类突发故障和渐变故障实验，提出了基于“减法聚类-双级BP神经网络”的突发故障快速隔离方法；之后，针对空调系统中的电气渐变故障和热力渐变故障，分别提出了基于“减法聚类-PCA”的电气故障征兆探测方法和基于“决策树-PCA”的热力故障征兆探测方法，模型中的SPE和T2统计量可以作为渐变故障的征兆，可以实现对微小故障的探测和故障征兆趋势的有效预示；最后，针对空调系统制冷剂泄漏故障，利用BP神经网络拓展了虚拟制冷剂充注量传感器模型，实现对微小故障的征兆探测，并以探测到的故障程度的和运行可靠度作为控制依据进行主动补偿，实现了基于运行可靠度评价的复合容错控制。本研究丰富了空调系统故障诊断的研究内容，解决了基于征兆探测的空调系统故障预示及其运行可靠度研究中的关键问题，对系统高效运行及节能具有促进作用，具有重要的实际意义和应用价值。