Battery ageing estimation with artificial intelligence
An accurate assessment of the battery SOH is pivotal for several aspects: correct estimation of the residual autonomy of electric vehicles, management of the energy flow in hybrid electric vehicles, prediction of the residual life of the battery and on-time replacement. This parameter cannot be directly measured and must be estimated.
Model based methods (i.e. Extended Kalman filters) and look-up tables are the current industrial solutions. They may suffer problems of inaccuracies due to difficult representation of the strongly nonlinear battery behavior. Data-driven approaches are not based on any model. They can guarantee good accuracy and little computational cost/memory occupation, provided that the training datasets include a sufficient set of operating conditions.
The adoption of Artificial Intelligence presents the following challenges: experimental characterization of the battery in laboratory environment with a 15kW tester (0-80V, 8 channels) to reproduce all possible operating conditions in terms of SOC, internal impedance and temperature on a climatic test chamber; accuracy of the SOH estimation ranging within ±5%; low computational cost and memory occupation of the algorithm to make it easily deployable on automotive electronic control units.
The project aims the following: development of aging estimation algorithms with an accuracy included within the range of ±10%; validation of the proposed method with an HIL setup: the HW is the control logic where the estimation logic is deployed; creation of a method that can be reproduced for different size and chemistry of the battery."
- PE7_3 Simulation engineering and modelling
- PE7_7 Signal processing
- PE6_11: Machine learning, statistical data processing and applications using signal processing (e.g. speech, image, video)
- Battery management systems
- Machine learning
- Lithium-ion batteries
- SoC and SoH estimation