| Under the background of the world energy revolution and China’s "3060" energy structure transformation,the renewable energy represented by solar energy is developing rapidly.Due to the weak construction of power grid in plateau region,users have a certain demand for photovoltaic energy storage system.However,energy storage System safety accidents occur frequently,so the safety detection design of Battery Management System(BMS)is becoming more and more important.State of Charge(SOC),State of Health(SOH)and Remaining Useful Life(RUL)of batteries are the core parameters of BMS,which are of great significance for the safe and stable operation of the energy storage system.In this paper,the battery state is taken as the research object,and the model-based SOC estimator,data-driven SOH and RUL estimator are studied respectively.The correlation between battery states and the deficiency of single state estimation were analyzed,and the single estimation model was improved.A dual time scale state joint estimation algorithm was developed.In addition,an experimental system for energy management of lithium battery pack is designed and its integrity is verified.The main research contents are as follows:(1)The correlation of battery parameters to battery operation and life attenuation was analyzed experimentally.The battery test platform was built and the ambient temperature and discharge rate were controlled.The characteristic parameter experiment and cyclic aging test were carried out on the 18650 ternary lithium battery.The effects of temperature and discharge rate on battery operation and capacity degradation were analyzed.(2)A second-order RC equivalent circuit model is built and SOC estimation based on the model is designed.The model parameters are identified online by Open Circuit Voltage(OCV)-SOC calibration experiment and recursive least square method with forgetting factor.The feasibility of the model was verified by Dynamic Stress Test(DST).Finally,the SOC estimator based on Extended Kalman Filter(EKF)and Unsented Kalman Filter(UKF)was designed using the battery model.UKF algorithm is better than EKF algorithm by comparing algorithm index.(3)A data-driven battery capacity attenuation model is designed,and a dual time scale state joint estimation model is designed to improve the single state estimation.The Long Short-Term Memory Recurrent Neural Network(LSTM-RNN)was constructed.First,a direct SOH estimator was established using external parameters to characterize battery health characteristics.Furthermore,constant current charging time,constant voltage charging time,average discharge voltage and constant voltage drop discharge time were extracted as Indirect Health Index(IHI)to establish the indirect predictor of RUL.The results showed that IHI had better ability to characterize the health characteristics of batteries.Finally,the correlation between battery states was analyzed,and a state joint estimation model with fast and slow time scales was designed.Experimental results show that this model is superior to the single estimation model.(4)Design a BMS with functions including data acquisition,operation state analysis and prediction,battery safety protection and early warning,energy management and data communication.STM32 and LTC6811 chips are used,and CAN communication is used.The system realizes the basic control of lithium battery pack and the detection of battery temperature,voltage,current,SOC,SOH and RUL and displays them in the upper computer.. |
PDF Full Text Request