Battery state of health estimation using the generalized regression neural network

Abstract

Batteries have been widely used in the field of electric vehicles. So prediction of the state of health (SOH) is important to the safe and efficient use of them. In this study, SOH is estimated by the generalized regression neural network (GRNN). GRNN is established by the radial basis neurons and linear neurons. The network has the advantages of approximation ability and the learning speed. In this test, there are 12 pieces of Li-ion batteries. Constant current charging and discharging are performed on them, until the capacity drops to below 80% of nominal. The SOH of the battery is estimated by the data that obtained from the operation. The data from the test shows that the recharging time by the constant current on the battery, the instantaneous voltage drops in discharge, and the output energy under a certain depth of discharge (DOD) are important to estimate the SOH of battery. The data from the 6 pieces of batteries are performed to train the GRNN. And the feasibility of this method is verified by the data from the other batteries. The test shows the difference of the SOH of the battery can be estimated accurately by this method, and it has great significance in the performance improvement of the battery management system.