Novel Multiple Parameter Optimization for Improving Accuracy of Battery Ageing Model and Lifetime Prediction

Lithium-ion batteries are widely used in EVs and renewable energies. To determine its current state and predict remaining life, sophisticated evaluation of battery degradation is needed. A battery lifetime prediction method using discharge curve analysis (DCA) is one feasible option. DCA based ageing model determines battery degradation by optimization of fitting parameters subject to local minimum solution. However, with battery ageing and data errors, model optimization gives multiple local minimum solutions, and thus prediction accuracy reduces. Conventional technology reported prediction accuracy of 10% and prediction time of 10 hours. In this research, we have developed an automated parameter optimization algorithm of DCA ageing model which finds multiple local minimum solutions and select the best lifetime prediction function. Algorithm determines parameter’s useful search spaces and performs optimization for multiple ranges by dividing these search spaces to locate multiple local minimum solutions. The combination sets of these solutions according to calendar and cycle life are fitted to power functions and one with lowest normalized error is considered for lifetime prediction. It is implemented in Python and verified for various battery calendar and cycle test data. It was found to maintain 5% prediction accuracy and execution time within an hour.