A Critical Review on Data Preprocessing Techniques for Building Operational Data Analysis

The wide adoption of Building Automation System (BAS) and Building Energy Management System (BEMS) has provided building professionals with large amounts of building operational data for knowledge discovery. Considering the intrinsic complexity in building operations and common faults in data collections, data preprocessing has been recognized as an indispensable step in building operational data analysis. It can be used to enhance data quality by removing outliers and missing values, ensure data compatibility with data mining algorithms, and improve the sensitivity and reliability in data analysis. This study provides a comprehensive review on data preprocessing techniques in analyzing massive building operational data. The paper firstly reviews techniques for conventional data preprocessing tasks, including missing value imputation, outlier detection, data scaling, reduction and transformation. Afterwards, the paper proposes promising techniques for advanced data preprocessing tasks, including data partitioning, feature engineering, data augmentation and transfer learning. Based on the critical review, future research directions and potential applications for building data analysis has been summarized. This paper can provide a general picture on data preprocessing methods for building operational data analysis. The insights obtained are valuable for the development of advanced data-driven solutions for smart building energy management.