State-of-charge estimation based on theory of evidence and interval analysis with differential evolution optimization

In this paper, we propose a new method for estimating the state-of-charge (SoC) of a lithium battery. There are remaining drawbacks of the existing methods, such as inaccurate estimation, high computation, and the need for massive datasets or an expensive sensor. Hence, the technique that we use to resolve such problems is derived from the theory of evidence (Dempster-Shafer theory) in which the degree of belief, namely mass, based on prior knowledge has to be assigned to each source of a variable. The proposed method is based on bounded-error state estimation with forward-backward propagation. However, instead of describing each variable error that propagates in the propagation process by a single error bound or a single interval, we define it by sets of intervals. In addition, to determine the optimal masses assigned to the variables, differential evolution is applied. To evaluate the proposed method, we carried out the experiments that used two different current sensors and two different voltage sensors to represent those variables with varying levels of uncertainty. Experimental results show that the root-mean-square errors of the proposed method are slightly better than those of a Kalman-filtering-based approach. The optimum masses also show the best performance, compared with masses assigned randomly. The results suggest that the proposed method can correctly estimate the SoC of a lithium battery.