A rule-based data preprocessing framework for chiller rooms inspired by the analysis of engineering big data

The rapid development of building energy consumption monitoring platforms makes engineering data more diverse, which facilitates the goal of reducing emissions. It is increasingly acknowledged that data preprocessing deserves the same attention as intelligent algorithms. In this work, the data quality issue of the engineering big data from non-demonstration complexes in China are analyzed thoroughly, and the analysis is based on clustering-based algorithms. We can conclude that the data of the hourly power of equipment groups are quality and stable, which is suitable for the benchmark to check other data. The quality of the data about pipes is acceptable. The number of data types about cooling towers is less, and the quality is worse. Regarding other data, the quality is unstable, so researchers should deal with those case-by-case. According to the above analysis, we proposed a convenient, rule-based data preprocessing framework that utilizes the law of physics, ensuring the strong coupling of multi-variants. After the data preprocessing, these engineering data are more reliable and can be used to improve performance or train models. Additionally, the proposed framework is more suitable for preprocessing multi-variant engineering data.