Investigation on the quality assurance procedure and evaluation methodology of machine learning building energy model systems

Data-driven approach is important for a wide variety of building energy applications including prediction, management, optimization, and predictive control. Due to the large-scale deployment of smart meters in buildings and the technology advance in machine learning techniques, machine learning models are becoming an integral part of building energy management platforms. The development and deployment of machine learning models is of utter importance to the efficient operation of buildings. The study found that there are deficiencies in the quality evaluation methods of machine learning model systems. From a software engineering perspective, the deployment process can be divided into requirements, design and validation. This paper proposed a quality model for machine learning building energy systems which can be used in the quality assurance of development of the system.