| The popularity of electric vehicles can effectively alleviate environmental pollution and energy crisis,so electric vehicles are gradually becoming a very strategic new industry in China,and battery technology is the key to restrict the development of electric vehicles.The State of Health(SOH)is an important indicator of the degree of aging of lithium-ion batteries.An accurate estimation of SOH can improve the safety performance of electric vehicles.However,the aging of current lithium-ion batteries and the complex mechanisms and difficulties in extracting health characteristics make it difficult to set up an online SOH estimation system in real vehicles.Aiming at the problems of difficult extraction of electric vehicle features and the difficulty of accurate estimation of SOH,this paper designed a battery low-frequency impedance spectrum measurement system,extracting combined frequency impedance features by analysing aging data,and at the same time established an long short term memory(LSTM)model to estimate the battery SOH.The main conclusions are as follows.(1)Through the analysis of lithium-ion battery characteristics and aging process,it was found that the low frequency band of the Li-ion battery impedance spectrum contained a wealth of aging information,from which features characterising the trend of SOH changes could be extracted.Therefore,the low frequency impedance spectrum test system was designed.Firstly,the hardware design of waveform generation circuit,bandpass filter circuit and sampling circuit was carried out.Then,the software design was carried out to write the communication program,waveform generation program and ADC sampling program.Finally,the designed low-frequency impedance spectrum detection system was verified and the test error was 0.15mQ.(2)Extraction of combined frequency impedance health features of batteries.Firstly,four 18650 lithium batteries were selected for rapid aging cycle experiments,and the battery capacity and low frequency impedance spectrum data were collected.Then,the capacity data and low-frequency impedance data were analyzed to divide the battery aging process into two aging stages,and the maximum correlation frequency in the two aging stages was selected as the characteristics of the stage by Pearson correlation analysis respectively.Finally,the features of the first and second aging stages were combined to obtain the combined frequency impedance characteristics of the battery.(3)Based on the selected combined frequency impedance features,a long short term memory(LSTM)model for SOH estimation of the batteries was built by using the test data of batteries B1 and B2.The root mean square error of the estimation results was reduced by 46.1%compared to the LSTM model using single frequency impedance features as input,while the root mean square error of the LSTM model was reduced by 67%and 32%respectively compared to the back propagation neural network model and support vector machine model using combined frequency impedance features.Innovations in this paper:(1)A low-frequency impedance spectrum measurement method was designed,using an excitation voltage of 1mV to ensure the linearity of the cell characteristics during impedance measurement.A two-stage differential amplifier circuit was designed to complete the amplification of smaller response signals,while a fourth-order Butterworth bandpass filter and 50Hz power supply filter were implemented to reduce spurious interference in the measurement process and improved the measurement accuracy at a larger amplification.(2)A SOH estimation method based on combined frequency impedance features was implemented.The effective health features of different ageing stages of the battery were combined to form a combined frequency impedance feature,and the LSTM-based SOH estimation model was built using this feature,which reduced the cost and time of online feature acquisition and also significantly improved the SOH estimation accuracy. |
PDF Full Text Request