An indirect prediction method of remaining life based on glowworm swarm optimization and extreme learning machine for lithium battery

Abstract

Aiming at the problem that the capacity of lithium battery is difficult to monitor on-line, an indirect health factor method based on time interval to equal discharging voltage difference is employed. Partial correlation coefficient analysis method is used to prove strong correlation between actual capacity and time interval to equal discharging voltage difference. Glowworm swarm optimization algorithm is employed to optimize extreme learning machine, meanwhile a model based on glowworm swarm optimization and extreme learning machine is built in order to realize the indirect prediction of remaining useful life for lithium battery. Data set of NASA B5 lithium battery is taken as a research object to evaluate the remaining useful life. Experimental results show that the algorithm can inherit the advantage of fast learning speed of extreme learning machine, and be of better accuracy and stability compared with extreme learning machine.