Lithium Iron Phosphate Battery Electric Vehicle State-of-Charge Estimation Based on Evolutionary Gaussian Mixture Regression

Lithium batteries have the characteristics of high energy density and charge-discharge rate, but exhibit high chemical activity. State-of-charge (SOC) estimation is critical to the lithium battery electric vehicle (EV) operation safety. In this paper, a novel SOC estimation method is proposed based on Gaussian process regression. A mixture Gaussian process is used in this model to strengthen the reliability of data description and to increase the estimation accuracy. Optimal number of Gaussian processes is obtained by a revolutionary expectation maximum method. A nonlinear correlation feature selection method is introduced to improve the model efficiency. The effectiveness of the proposed method is verified by an EV field test. Compared with other data-based approaches, this method exhibits higher estimation accuracy and computational efficiency.

I. Introduction

With the characteristics of high power and energy density, lithium-ion batteries are becoming popular in electric vehicles (EVs). In order to prolong battery life and prevent battery damage, reliable and accurate observation of real-time battery status is required for a battery management system. As one of the key parameters, state of charge (SOC) is the indicator of the battery utilization rate. However, it cannot be measured directly and also difficult to be estimated due to series and parallel connections of batteries, which is a threat to the charge/discharge safety.