A framework based on big data for intelligent monitoring of battery packs

Existing literature focus on the prediction of states of batteries are scattered and are individually studied based on several battery aspects such as: 1) Chemical (ionic concentration measurement or diffusion coefficient evaluation), 2) Electrochemical (capacity), 3) Electrical (internal resistance), 4) Thermal (temperature), 5) Mechanical (stack/enclosure stress) and 6) In-situ/ex-situ (characterization methods to measure porosity and grain size). Unfortunately, these studies have been done by experts of different fields and are yet to be combined in a common platform to predict the states of batteries in a comprehensive way. In this paper, the aim of this research is to propose a framework so as to establish a big database (from sources of literature, by performing real-time experiments and uncertainty studies) for batteries at all operating conditions by incorporating all aforesaid aspects. Once the data base is established, a suitable artifical intelligence approach such as artificial neural network will be applied to train and build the model for state of health prediction and physical evaluation that subsequently have the prime advantage of accurately predicting the battery capacity at system level as well as cell level based on all existing design parameters (diffusion coefficient, grain size, temperature, internal resistance, etc.) from the big database. Data collection will be processed on brand new batteries by repeating cycles of charge and discharge modes under dynamic current profiles at different temperatures for accuracy. The proposed battery model can be then integrated to the battery management system in the electric vehicle without any additional integration complexity.