Voltage Fault Diagnosis of Power Batteries based on Boxplots and Gini Impurity for Electric Vehicles

Power battery is a critical factor affecting the safety of electric vehicles (EVs). Fault diagnosis and prediction of power batteries are of great significance to ensure the safety of EVs. This paper proposes a voltage fault diagnosis model based on boxplots and Gini impurity. Considering cells voltages are not normal distribution at any time, we use the boxplots to analyze the monitoring voltage data and identify the abnormal cells with coarse granularity. To quantify the abnormality of each cell, the anomaly distance is defined based on boxplots. Considering each time has different degrees of influence on the final result of each cell, we use the Gini impurity weighting method to measure the contribution rate of each time. By this means the goal of further locating the faulty cells accurately can be achieved. And then we can easily identify those faulty cells by utilizing the Z-score method. Different from other previous researches, the validation and contrast experiments in this paper are carried out by using the actual vehicle operation data of the National Monitoring and Management Center for NEVs in Beijing. The results of experiments clearly show that the proposed model has high diagnostic efficiency relatively and the faulty cells in the battery system can be located accurately.