Particle filter for state of charge and state of health estimation for lithium–iron phosphate batteries

Abstract The paper presents a new approach for state estimation of lithium–iron phosphate batteries. Lithium–iron phosphate/graphite batteries are very intricate in state of charge estimation since the open circuit voltage characteristic is flat and ambiguous. The characteristic is ambiguous because open circuit voltages are different if one charges or discharges the battery. These properties also hinder state of health estimation. Therefore conventional approaches like Kalman filtering which represents a state by only the mean and the variance of a Gaussian probability density function tend to fail. The particle filter presented here overcomes the problem by using Monte Carlo sampling methods which are able to represent any probability density function. The ambiguities can be modelled stochastically and complex models dealing with hysteresis can be avoided. The state of health estimation employs the same framework and takes the estimated state of charge as input for estimating the battery's state of health. The filter was developed for A123 lithium–iron phosphate batteries. For validation purposes user profiles for batteries in different ageing states like electric vehicles and off-grid power supply applications were generated at a battery testing system. The results show good accuracy.