ExplainerX: An Integrated and Explainable AI Framework for Nearly Zero-Energy Buildings

Climate change is a pressing global issue primarily driven by excessive energy consumption, with the building sector being a major contributor. In response to this challenge, Nearly Zero-Energy Buildings (NZEBs) have emerged as a promising solution. These buildings are designed to generate as much energy as they consume by leveraging energy-efficient technologies and renewable energy sources. However, achieving the NZEB goal relies on accurately predicting a building’s future energy usage. This precision is crucial for efficiently managing energy production, consumption, storage, and distribution within NZEBs. The current Artificial Intelligence (AI) solutions face significant shortcomings that extend beyond mere performance issues. These include critical aspects such as the lack of transparency in gathered data and predicted results, oversight in identifying model drift (especially in fast-paced sectors like energy), and the use of disparate tools in the model development process. This paper addresses these issues when developing AI solutions for NZEB projects by introducing ExplainerX, an integrated and explainable framework that streamlines all stages of building transparent and high-performance prediction models. ExplainerX ensures transparency in both data and results, accounts for data changes in dynamic settings, and promotes effective energy management in buildings. It leverages different components of the CRISP-DM methodology, providing explanations for each pipeline and tracing every decision made by data scientists and decision-makers. The framework is illustrated through a case study utilizing real datasets from the EU Improvement Project. ExplainerX source code is available on our GitHub repository.