| As the main transportation tool of modern road transport,the automobile has become an important means of daily travel and logistics transportation for people.As one of the mainstream directions of automobile development,electric vehicles have significant strategic significance in reducing greenhouse gas emissions,promoting the development of electric transportation,and achieving the "dual carbon" goals.However,the development of electric vehicles is constrained by many problems in the field of power lithium batteries.Among them,the estimation of the battery’s internal state plays a crucial role in the battery’s life,safety,and vehicle energy management,and it is currently a hot research topic.However,power lithium batteries exhibit strong timevarying nonlinear characteristics,and there are technical bottlenecks such as low estimation accuracy,poor robustness,poor environmental adaptability,and poor adaptability throughout the entire life cycle,making the real-time and accurate estimation of the internal state of power lithium batteries highly challenging.Therefore,this paper takes power lithium battery monomers and battery packs as the research objects,adheres to the orientation towards the forefront of science and technology and towards national major needs,and focuses on how to accurately estimate the internal state of power lithium batteries under complex operating environments such as extreme temperatures,aging,changes in charging and discharging conditions,user behavior,and inconsistency,to improve the performance and safety of batteries and promote the widespread application of battery technology in the automobile industry.To solve the problem that it is difficult to accurately identify the model parameters of power lithium batteries in complex use scenarios,an adaptive iterative parameter identification method combining weighted average online and offline is proposed.A power battery experimental platform is established,a test protocol for power battery cell and battery pack is designed,and a power battery test database is established based on the laboratory and big data platform.On this basis,the mathematical modeling methods and recursive equations of three different battery models are analyzed and deduced,and an online parameter identification method based on real-time measured current and voltage data is proposed.Combined with the traditional offline parameter identification,the power battery model is accurately identified in uncertain and complex environment.To solve the problem that it is difficult to accurately reckon the state of charge(SOC)of lithium batteries under time-varying temperature,a method of SOC estimation based on temperature and capacity compensation is presented.The map among SOC,open circuit voltage and temperature of power lithium batteries is clarified,the open circuit voltage characterization function with SOC and temperature as input is established,the mechanism of temperature affecting discharge capacity is analyzed,the fitting function of battery capacity under variable temperature is constructed,the discharge capacity change characteristics under the combined action of temperature and discharge rate are further discussed,and the capacity compensation model is constituted by using random forest algorithm.Based on the different battery models,accurate SOC estimation for power lithium batteries under uncertain environment temperature and complex operation conditions is achieved by using open circuit voltage correction,adaptive H infinite filter,and improved square root cubature Kalman filter.To solve the problem that it is difficult to accurately estimate the state of health(SOH)of power lithium batteries under random charging behavior,an efficient SOH estimation method for power lithium batteries under short-term and random charging behavior is proposed.The voltage variation characteristics of power lithium batteries during different aging stages are revealed in different voltage ranges and different SOC intervals.The voltage curve prediction method for complete charging process based on short-term and random charging data is investigated.A battery degradation model is built by using improved least squares support vector machine algorithm,and a voltage shape fitting and reconstruction method combining mechanical model and prediction model is constructed.Furthermore,an ensemble learning SOH estimation method based on induced ordered weighted averaging operator is presented,which achieves the timely and optimal weight allocation for each learner.To solve the problem that it is difficult to accurately estimate the internal states of power lithium batteries under aging conditions,a multistate co-estimation method based on fusion filter algorithm and machine learning algorithm is developed.The incremental capacity of power lithium batteries in different SOC intervals during different aging stages is analyzed,and the accurate estimation of SOC for power lithium batteries under time-varying temperature conditions in the whole life cycle is achieved by combining the adaptive H infinite filter and machine learning methods.The variation law of open circuit voltage in different aging stages is analyzed.The SOH is updated in time by integrating open circuit voltage and Ampere-hour integral method.The influence of multi-innovation update adaptive extended Kalman filter on SOC estimation is studied.Based on the constraints of power battery design current,voltage and power,SOC operation range and power continuous output time,a multi-constrained state of power(SOP)prediction manner is deduced.Overall,the estimation accuracy of SOC,SOH and SOP for battery cells has been improved.To solve the problem that it is difficult to accurately estimate the internal state for inconsistent power battery packs,a SOC estimation method based on the highest/lowest voltage single sample characterization and a SOH estimation method based on the transfer learning and mean and difference models are presented.The inconsistencies of voltage,SOC and SOH before and after power battery grouping are revealed,the SOC differences between battery cells are compared,the process of maximum and minimum SOC weight allocation and deviation update is deduced,and the iterative calculation method of battery pack SOC with weights and deviations is designed.The mean cell model and difference cell model of battery pack are constructed,an online SOH estimation method combining machine learning and transfer learning is proposed,and the effective transplantation method of SOH estimation model based on machine learning on different types of batteries with different aging degree is explored. |
PDF Full Text Request