Modeling And State Of Charge Estimation Of Zinc-nickel Single Flow Battery

As an essential device to suppress the fluctuation of renewable energy power generation,energy storage battery has been rapidly developed with the rise of the concept of the low-carbon economy.The zinc-nickel single flow battery is a new type of energy storage battery.Compared with the all-vanadium flow battery using two liquid storage tanks,there is only one liquid storage tank;it does not need to use an ion-exchange membrane,which reduces the production cost.At the same time,compared with the traditional zinc-nickel battery,it solves the problems of zinc dendrite formation and uneven distribution of zinc,so it has the advantages of long cycle life,safety and reliability,and no environmental pollution.Zincnickel single flow battery has developed rapidly since they were first proposed;however,some critical problems are still to be solved when applied in large-scale power grids as energy storage devices.As one of the core technologies of the energy storage system,the battery management system(BMS)is of great significance to the zinc-nickel single flow battery,and the battery state of charge(SOC)estimation is one of the main functions of the battery management system.Its accuracy also affects many functions such as charge and discharge control,battery pack balance,and safety management in the battery management system.Since SOC is a state quantity inside the battery,sensors cannot directly measure it.Among the many proposed SOC estimation methods,the model-based SOC estimation method has been proved to be effective,but since its accuracy depends on the accuracy of the established battery model,for zinc-nickel single flow battery,how Establishing an accurate model has become a problem that must be solved for accurate SOC online estimation.Aiming at the core problem of accurately estimating battery SOC in BMS,this paper starts with how to establish an accurate zinc-nickel single flow battery model.It carries out the following research work in the following three aspects.The research results mainly include:(1)A second-order RC equivalent circuit model was established based on the charge-discharge voltage/current curve of zinc-nickel single flow battery.In order to evaluate the impact of different fitting data on the identification of model parameters,an experiment was designed to fit the model parameters using three different experimental data sets,long-term shelving,transient pulse discharge,and constant current discharge,by comparing the simulation results of different models.It was found that the model parameters of the zinc-nickel single flow battery changed with the increase of the storage time when the battery was stored for a long time.However,since the data set used for fitting is too tiny for instantaneous pulse discharge,overfitting will occur;the model parameter values obtained by fitting under constant current discharge are closer to the actual physical parameters of the battery.In order to further study the generalization of the model,a dynamic stress test(DST)experiment and a self-defined working condition experiment were carried out.The three errors of error(AE),root mean square error(RMSE),and absolute mean error(MAE)are only about half of those of the other two models.However,the model errors under custom conditions are relatively large,so the model structure must be further studied to improve model accuracy.An equivalent circuit model structure based on state switching is proposed by analyzing the modeling error.Compared with the second-order RC equivalent circuit model,it models the dynamic characteristics separately according to the working state of the battery.When the battery load current is not zero,the equivalent circuit model based on state switching has the same model parameters as the 2-order RC equivalent circuit model.However,when the battery is resting,the proposed model introduces a new electrical element to simulate the change in the battery’s terminal voltage while it is resting.The new model uses the Marine Predator Algorithm(MAP)to identify the increased model parameters and evaluate the performance of the proposed model under DST and custom conditions.The experimental results show that the proposed model has better performance when the load current changes rapidly.However,the performance evaluation indicators are similar to the second-order RC equivalent circuit model,but if the shelving time is slightly increased,the proposed model will show better performance.(2)To study the influence of different models on the SOC estimation of zincnickel single flow battery,a square root unscented Kalman filter(SRUKF)was proposed by combining the second-order RC equivalent circuit model and the equivalent circuit model based on state switching,respectively.The SOC of the battery was estimated,and its performance was evaluated under two operating conditions.The experimental results show that: when using the equivalent circuit model based on state switching,the SRUKF algorithm has higher accuracy;but because the simulation errors of the two models have impulse noise and gradually increase in the lower SOC range,the SOC estimation results divergent trend.For the two simulated error characteristics that cause this problem,an adaptive square root unscented Kalman filter with correlation entropy loss(MCC-ASRUKF)and a robust tracking square root unscented Kalman filter with multiple fading factors(MSTF-SRUKF)are proposed,respectively.Experiments verify)algorithm to estimate the battery SOC and the algorithm’s validity.The two algorithms outperform SRUKF in various evaluation indicators and converge to the real SOC value.(3)In the laboratory,due to the high measurement accuracy of the power battery test system,the current and voltage values measured by sensors can usually be considered high-precision data without noise and bias.In the natural working environment of zinc-nickel single flow battery,due to the influence of electromagnetic interference or the drastic change of the load current in the battery and other factors,the data obtained by the sensor measurement is difficult to avoid mixing with uncertain and non-stationary interference noise.The parameters may change due to environmental changes,and a SOC estimation method based on desensitized unscented Kalman filter(DUKF)is proposed.By introducing the concept of parameter sensitivity,the algorithm has a higher performance when the model is mismatched,or the sensor is biased.Robustness and its effectiveness are verified in two test conditions.The experimental results show that when the model parameters or sensors have large deviations,DUKF can still converge,and its performance is better than SRUKF.In order to further improve the accuracy of the SOC estimation algorithm when the models are mismatched,a SOC estimation algorithm combining the Kalman filter and bias compensation least squares method is proposed,compared with DUKF,the two SOC estimation methods have their advantages and disadvantages,and scope of application.