基于知识数据化表达的制冷空调系统故障诊断方法

制冷空调系统广泛用于建筑环境调节，是建筑能耗的重要组成部分，而系统故障运行会造成15%~20%的能耗增加。以深度学习为代表的数据驱动方法是故障诊断的热点技术。然而，数据驱动需要依赖大量标记数据从而限制了其应用。针对上述问题，提出一种基于知识数据化表达的故障诊断方法，通过将故障诊断先验知识以数据化的形式表达弥补真实标记数据不足的难题。首先，提出以随机缩放策略为信息扩增手段的知识数据化方法，并利用添加噪声达到生成样本与真实样本一致性更优的目的。然后，提出基于基准模型的目标系统偏离特性表征方法，将目标系统数据与生成数据的格式统一。最后，利用生成数据训练模型并在ASHRAE RP-1043数据集上验证，综合诊断正确率达82.67%，与经典的监督学习方法效果接近且完全无须标记数据，具有广泛应用前景。

Fault diagnosis method of refrigeration and air-conditioning system based on digitized knowledge representation

Refrigeration and air-conditioning systems are widely used in building environmental regulation and are an important part of building energy consumption, and system failures will increase energy consumption by 15%—20%. The data-driven method represented by deep learning has become the mainstream method for fault diagnosis. However, data-driven method needs to rely on a large amount of labeled data, which limits their applications. Focus on the above problems, a fault diagnosis method based on digitized knowledge representation is proposed, which makes up for the problem of insufficient real labeled data by prior knowledge of fault diagnosis in a data form. A method of knowledge digitization using random scaling strategy is proposed, and a noise addition strategy is used to achieve the goal of better consistency between the generated sample and the real sample. A characterization method of target system deviation characteristics based on benchmark model is proposed, which unifies the format of target system data and generated data. Using the generated data to train the model and verify it on the ASHRAE RP-1043 data set, the comprehensive diagnosis accuracy rate is 82.67%, which is close to the supervised learning method. Combined with that it does not need to labeled data at all, makes it has a wide range of application prospects.