| In order to alleviate energy and environmental problems,the development of new energy vehicles,especially battery electric vehicle(BEV),has attracted widespread attention.As a key component of BEV,the battery performance degradation or failure will affect the normal operation of the vehicle and even cause major safety accidents.Therefore,it is of great practical significance to assess the state of health of on-board batteries.This paper preprocesses actual vehicle operating data,extracts sensitive features to characterize the state of health of battery pack,and uses multi-source information fusion method to construct vehicle battery state of health assessment model.Verification of actual vehicle operation data shows that the method proposed in this paper can improve the accuracy of state of health assessment,timely detect abnormal state of battery pack,and reduce potential safety hazards in the normal operation of BEVs.The work of this paper includes the following contents:(1)The preprocessing method of real-vehicle operating data of BEVs is studied.The data types that have no analytical value among the acquired data types are removed.The missing data in the acquired data is divided into long-term missing data and short-term missing data.Deletion and weighted moving average interpolation are adopted to make up the missing data respectively.The box-plot method is used to detect and eliminate abnormal values in the abnormal data in the acquired data.Through the preprocessing of the data,the dimensionality of the data is reduced and the quality of the data set is improved.(2)Aiming at the problem of redundant information in multi-dimensional features,the feature extraction method representing the state of health of battery pack is studied.The common correlation measurement methods are compared and analyzed,and the maximum information coefficient(MIC)with low complexity and high robustness is finally selected as the feature selection method.Through four representative functions,the advantage of the MIC in the correlation measurement of multiple functions is verified.According to the maximum information coefficient,the process of feature selection is established,and the characteristics of battery pack are screened to obtain sensitive features that accurately characterize the state of battery pack.(3)Aiming at the problem that one single feature parameter cannot fully assess the state of battery pack,a method for assessing the state of health based on Gaussian mixture model(GMM)and multi-source information fusion based on Bayesian inference distance is studied.The Gaussian mixture model is established by using the data of the fault-free state,and the distance between the test data and the health state model is characterized by the feature fusion index based on Bayesian inference as the quantitative index for assessment.Through comparative analysis with the evaluation results of a single feature parameter,all feature parameter fusion,random extraction feature fusion,and the fusion based on support vector data description,it is verified that the multifeature parameter fusion index proposed in this paper has a better ability of state of health assessment.Aiming at the problem that the test data may have abnormal data,leading to the wrong state assessment results,the establishment of abnormal data judgment rule provides an idea for solving the problem that the data input model may be abnormal data,leading to the triggering of false alarm. |
