Research On State Estimation Of Energy Storage Battery Management System

In recent decades,with the development of the ’Made in China 2025’ policy,the Chinese government has paid more attention to the field of new energy,and the related domestic industries such as energy storage battery have flourished.In order to ensure the safe and reliable operation of energy storage batteries,battery management system has become the hotspot both at home and abroad.State estimation is the core function of battery management system.The innovation of this dissertation is to combine the methods of parameter identification,unscented Kalman filter,empirical mode decomposition,long short-term memory recurrent neural network and propose a set of state estimation schemes for energy storage battery management system.The state estimation algorithms of state of charge,state of health and state of power are designed respectively,which realize the high-precision estimation.The research content of the dissertation is divided into the following three parts:(1)The problem of estimating the remaining capacity of battery in the charge-discharge cycle from fully charged to fully discharged is studied.A second-order RC equivalent circuit model is established,and the relationship between open circuit voltage and SOC is calibrated based on the data of constant current discharge experiment.For further parameter identification of the system,an improved recursive least squares algorithm is proposed,which enhances the correction effect of the newly obtained data.As the improved recursive least squares algorithm falls into the local extremum easily during the process of finding the optimal solution,the system parameter identification based on genetic algorithm is proposed.Combined with the results of parameter identification,an unscented Kalman filter is designed to estimate the state of charge.The simulation results verify the validity of the designed state of charge estimation algorithm.(2)The problem of estimating state of health during the life cycle from manufacture to failure is studied.A state of health estimation algorithm based on empirical mode decomposition and long short-term memory recurrent neural network is proposed.Multi-scale separation is carried out through empirical mode decomposition to separate the global degradation trend of capacity fading from the local rebirth characteristics,which solves the problem of severe fluctuations caused by the rebirth of battery capacity indicators.The strongly correlated signals are synthesized.After that,the LSTM-RNN networks of the global degradation trend and fluctuations are constructed respectively.Finally,the estimation results are integrated.The simulation results prove the effectiveness of the proposed state of health estimation algorithm.(3)The problem of estimating the limit ability of battery charge-discharge power under different states of charge is studied.Considering the limitation of battery model,SOC,current and power,a multi-constrained dynamic peak power estimation method is proposed.With the constraints of maximum charge-discharge current and power,the continuous peak power estimation algorithm is derived,which realizes the effective estimation of the dynamic peak power within the duration.The simulation results verify the effectiveness of the designed state of power estimation algorithm.