ML-Based Decision Support System for Energy Efficiency Retrofits

With global energy consumption increasing and environmental problems worsening, the need for effective solutions to reduce energy consumption in buildings is urgent. Energy renovation plays a key role in achieving this objective, allowing the energy efficiency of buildings to be improved and their negative impact on the environment to be mitigated. However, stakeholders involved in the renovation process often struggle to select the most appropriate energy efficiency measures to be implemented in a building. This paper aims to promote energy efficiency retrofits by providing an ML-based decision support system for identifying the most suitable and impact-heavy renovation actions for a certain building. First, an analysis of the available data from energy efficiency projects that have been implemented in the past is carried out and the different energy efficiency measures are identified. Then, different machine learning algorithms are applied, and the algorithms with the highest degree of accuracy are identified and further optimized. The results estimate the most energy efficient measures, in addition to the CO2 savings achieved. These results show how different combinations of features and algorithms can influence the outputs, leading to the selection of the best combination of techniques, indicating that Random Forest and LightGBM algorithms outperform other models in predictive accuracy, making them suitable for large scale energy management applications. Moreover, the results emphasise the importance of machine learning in the field of energy efficiency and energy in general, since the potential of machine learning models to support stakeholders in decision-making procedures are highlighted, taking advantage of existing data, conducting systematic analysis of available information and providing useful insights and predictions of the impact of specific interventions.