NEVORA (Network-based Evaluation for On-line Reliability Assessment of batteries) is a diagnostic software platform that combines electrochemical impedance spectroscopy (EIS) with machine learning to enable real-time, interpretable monitoring of battery performance and health. Traditional battery management systems rely mainly on voltage and current measurements, which often struggle to accurately determine state of charge (SoC) and state of health (SoH) under dynamic or aging conditions. NEVORA addresses this limitation by transforming each impedance spectrum into a compact, physics-informed feature vector that captures frequency-dependent processes inside the battery, such as bulk transport, interfacial reactions, and degradation mechanisms.
Using these feature vectors, NEVORA trains multi-output machine learning models capable of simultaneously predicting key battery metrics such as SoC and cycle number (as a proxy for SoH) with very high accuracy. Beyond prediction, the platform emphasizes interpretability: feature-importance analysis links specific frequency ranges in the impedance spectrum to underlying physical and electrochemical processes, allowing users to understand why a given prediction is made rather than treating the model as a black box. This makes NEVORA both a powerful diagnostic tool and a scientific framework for studying battery behavior.
Designed for real-time operation and lightweight deployment, NEVORA is suitable for integration into next-generation battery management systems, particularly for advanced chemistries such as solid-state batteries where conventional methods fall short. By uniting impedance spectroscopy, machine learning, and physical insight into a single workflow, NEVORA provides a robust pathway toward smarter, faster, and more reliable battery monitoring and control.