Challenges in real practice applications of data-driven intelligence to decarbonize our buildings and cities

  • Cipriano, Jordi (CIMNE)
  • Laguna, Gerard (CIMNE)
  • Mor, Gerard (CIMNE)

Please login to view abstract download link

Data-driven methods involve statistical analysis and machine learning techniques to extract insights and predictions from large data sets. For the last few years, these techniques have been widely applied in several fields of the building sector, addressing the assessment of the energy efficiency in buildings, optimizing the control of building systems or characterizing the energy performance of districts. These data-driven techniques have benefited from the fast-growing of artificial intelligence (AI), which relies heavily on them to learn from past data patterns and use that knowledge to make accurate predictions or decisions in new situations. Advanced methods such as transfer, deep, or reinforcement learning have been extensively applied to research works focused on highly monitored and automated buildings. Other techniques based on geo-computing are also starting to be used in the urban context. Although the research in data-driven methodologies is at a high readiness level, the massive adoption of these procedures and algorithms over current building stock is still in its infancy. The challenges to overcome, which hamper their rapid implementation in operational use, have nothing to do with a lack of knowledge in modelling the energy performance of the built environment correctly but rather the lack of high-quality data, interoperability issues among different data models and communication protocols, and the lack of standardized processing pipelines. Data must be collected, cleaned, and processed before being used in models. Additionally, data may be incomplete or inaccurate, which implies high data harmonization resources. Another challenge is the complexity of building systems and the lack of interoperability along the whole value chain. Data-driven solutions must be compatible and integrate seamlessly with these systems. Finally, energy efficiency solutions based on AI must engage users and provide clear feedback and guidance. However, many building owners and operators are not familiar with these technologies and are hesitant to implement new technologies without clear evidence of their effectiveness. This paper will show practical experiences in addressing some of these challenges. The pilot applications implemented in several EU-funded projects demonstrate the benefits and drawbacks and the limitations of data-driven techniques to address the climate and energy crisis in the urban environment, with a particular focus on buildings.