State Of Health Estimation And Life Prediction Of Lithium-Ion Batteries For Satellite In Orbit

In modern society,satellite is important in communication,navigation,and detection.The power supply and distribution subsystem is the key subsystem of satellite,so the power supply and distribution subsystem state of health and remaining useful life have received more attention.In the new generation of satellite platforms,the lithium battery has become the first choice in the power supply and distribution subsystem due to their advantages in energy density.The safe and efficient operation of lithium battery requires accurate state of health as the basis for the control of satellite power supply and distribution subsystem,which puts forward higher requirements for state of health estimation method and remaining useful life prediction method.However,the application of existing battery evaluation methods is limited due to the complex working conditions of satellite battery.Therefore,this paper firstly researches the satellite battery state of health estimation methods based on data and model respectively.Then,the multi-source state of health fusion method with cell data and the battery pack state of health estimation method with only battery pack data are researched respectively for battery pack.Finally,the battery degradation model and remaining useful life prediction method for satellite under variable conditions are researched.In the aspect of data-driven satellite battery state of health estimation methods,this paper extracts the first feature of interest of increment capacity curve,discharge capacity of interval voltage,and direct current resistance as the battery health indicator according to the characteristics of telemetry data of in-orbit satellite batteries,forming the following three estimation methods of battery state of health.(1)The method of battery state of health estimation based on improved increment capacity analysis: this method adopts smooth spline curve smoothing to reduce the error caused by low sampling precision of satellite telemetry data.Then for the discharge data with load,a method for calculating the increment capacity curve with constant current discharge data is established.In this way,the first feature of interest of increment capacity curve can be obtained from the telemetry data of in-orbit satellites,and the battery state of health can be estimated.(2)The capacity estimation method based on the discharge capacity of interval voltage: by correcting the influence of current and temperature on discharge quantity,this method obtains the battery state of health by analyzing the relationship between discharge quantity corresponding to specified open circuit voltage range and battery capacity under any working condition.(3)The state of health estimation method based on resistance: this method revises the direct current resistance by calculating the change of open circuit voltage in the process of current jump,and then obtains the battery state of health.The above methods use satellite telemetry data of different discharge positions to estimate the battery state of health,and their accuracy is verified to be better than that of the traditional methods.In the aspect of model-driven of satellite battery state of health estimation methods,the estimation methods of resistance and capacity are researched respectively.For capacity estimation,the capacity estimation method based on coupling temperature simplified single-particle model is proposed.In this method,the particle diffusion in the single-particle model is simplified to the diffusion of particles between two regions,and then the temperature distribution of lithium battery is combined to form a model to describe the battery discharge process,thereby reducing the calculation amount and the requirement of known parameters.Finally,a genetic algorithm is used to estimate the battery capacity by taking the difference between the model voltage output and the actual voltage as the objective function.This paper proves that the battery model has better accuracy than the equivalent circuit model under constant current and variable current conditions,and the computational complexity is also lower than other electrochemical models.The capacity estimation method based on this model also shows good accuracy.For resistance estimation,unscented Kalman filter is used,in which a simplified electrochemical battery model is used as the system equation of the filtering process.Through this method,we can achieve real-time estimation of internal resistance,which provides support for the health monitoring and evaluation of batteries.In the aspect of satellite battery pack state of health estimation,this paper studies the state of health estimation method for two cases: 1)measurement data of single battery is available,2)only measurement data of battery pack is available.When the measurement data of single battery is available,a fusion estimation method is proposed for battery state of health.This method uses Dempster-Shafer theory to obtain the reliability of each estimation result,and then uses Monte Carlo method for fusion,thereby the credibility of the estimation results is effectively improved.When only measurement data of battery pack is available,an estimation method of battery pack state of health based on genetic algorithm is proposed.This method builds function relation between capacity of cells and voltage of battery pack based on the assumption that the open circuit voltage curve is constant.The genetic algorithm is adopted to estimate the capacity of each cell in the battery pack,achieving an approximate estimation of the state of health of battery pack in the absence of single battery’s data.In the aspect of satellite battery remaining useful life prediction,this paper first proposes the linear superposition mode of cycle degradation,calendar degradation and capacity relaxation effect through mechanism analysis,in which cycle degradation is the main degradation factor of the battery.For the cycle degradation process of satellite battery under variable operation conditions,a multi-factor cycle degradation model of lithium battery has been established in this paper,which effectively describes the influence of different working conditions on capacity degradation.Then the particle filter algorithm is used to estimate the degradation model parameters of the on-orbit satellite battery and predict the remaining useful life.In this process,aiming at solving the invalid particles produced by particle filter,a partial mutated resampling particle filter algorithm is proposed in this paper.In this method,the model segmentation points are determined firstly,and then in the process of particle filtering,the particles in sparse regions are mutated to replace the invalid particles to alleviate the degradation and dilution of particles.It is found that this method can estimate the parameters of the battery degradation model and predict the remaining useful life by using the degradation data of variable working conditions.