| Large-scale development of Electric Vehicles(EVs)is an important approach to address the energy and environmental crisis and has become an industry and development direction that is strongly supported by governments all over the world.The performance of EV power batteries degrades with increasing usage,and when the performance drops to 80% of the original value,the battery can no longer meet the standard for EV usage.Retired power batteries cannot be used in EVs,but they have great potential for applications in power grid energy storage and communication base stations.To reuse retired power batteries,individual battery cells must be rearranged into groups.However,if the battery cell consistency is poor,the performance of the battery will be greatly reduced due to the further amplification of the consistency difference,resulting in a shorter lifespan for the individual battery cells and safety hazards for the battery group.Therefore,a comprehensive analysis of the characteristics of retired power batteries is required.In this thesis,retired ternary lithium batteries were selected as the research object,and the characteristics of retired power batteries were comprehensively analyzed.The impact of different charge-discharge rates on the battery’s internal resistance,capacity,and polarization properties were analyzed using retired 2698 0 ternary lithium batteries in charge-discharge experiments.In addition,the aging characteristics of retired power batteries were studied due to the accumula tion of side reactions during the battery’s operation,which results in irreversible effects on t he battery.Based on the polarization phenomenon,rate characteristics,and hysteresis phenomenon of retired power batteries,a comprehensive dynamic equivalen t circuit model of retired power batteries was established.The model parameters were identified using a hybrid pulse power characteristic experiment and verified for their accuracy in dynamic conditions.The comprehensive equivalent circuit model was comp ared with a model with hysteresis effects,and the experimental data confirmed that the comprehen sive equivalent circuit model had smaller errors indicating higher accuracy in dynamic conditions.The State of Energy(SOE)of the battery was estimated using a particle filter method.Nonlinear models inevitably produce noise during dynamic operating con ditions,which affects the accuracy of SOE estimation results.To overcome this issue,a particle filter method was used to estimate the SOE of the battery.Un scented Kalman Filter and Dual Particle Filter algorithms were used to estimate the SOE of the ba ttery,in combination with the particle filter method to improve the accuracy of SOE estimation.In dynamic conditions,the estimated results were validated,a nd it was shown that the SOE estimation using the unscented particle filter had higher accuracy.The Student Psychology Based Optimization(SPBO)algorithm was used to optimize a recurrent neural network,and the optimized recurrent neural network was used to estimate the State of Power(SOP)of retired power batteries.Considering that recurrent neural networks are prone to being trapped in local optimal states during training,the SPBO algorithm was used to optimize recurrent neural networks,which has good high-dimensional global optimization capabilities,to address the issue of inaccurate power estimation for retired power batteries.In dynamic conditions,the estimated results were validated,and it was shown that the recurrent neural network optimized with the SPBO algorithm had higher accuracy,effectively solving the issue of the network being trapped in local optimal states,enabling computations for large amounts of data and maintaining a high solution speed,improving the accuracy of retired power battery SOP estimation. |
PDF Full Text Request