e-SparX

Machine learning (ML) is becoming a crucial enabler of the energy transition. Unfortunately, limited data and model sharing, irreproducible workflows, and a lack of standardized model benchmarking slow down the innovation in this particularly relevant area of research. Despite the existence of several platforms for sharing data and ML code, it is still very challenging to find and reuse relevant ML artifacts (data, scripts, model architectures, or weights) in the energy systems field. 

e-SparX is a novel tool that facilitates sharing and finding ML artifacts for fast and easy reuse, specifically designed for the needs of energy ML researchers. We build upon existing tools (e.g., the Open Energy Platform, GitHub, MLflow) and offer a user-friendly package to publish ML artifacts with minimal effort. e-SparX features a web-based GUI that organizes and visualizes ML artifacts in interconnected ML pipeline graphs, enabling the finding and benchmarking of ML projects from the energy community. The e-SparX prototype has been developed based on a requirements analysis and is fully functional and open-source.