| With the popularization of electric vehicles,lithium-powered batteries are favored in the new energy market because of their pollution-free,light quality,high energy density and wide application range.The residual capacity of power battery discharge is nonlinear mapped with the remaining parameters,and the temperature change has a significant effect on the residual capacity,so it is difficult for the general linear analysis method to accurately estimate the charge state(SOC,States of Charge)on line in practical use.In view of this problem,an RBF neural network model based on quantum particle swarm optimization algorithm is proposed,which sets up the temperature monitoring system designed and developed by Labview platform,and realizes the high-precision on-line SOC modeling and prediction analysis of lithium battery and lithium-ion battery pack at different temperatures,as well as the safety control of lithium battery charging and discharging process.In this paper,the internal and external structure and working principle of lithium-ion battery are described,and its specific parameters and factors affecting SOC estimation are analyzed.In this paper,the traditional RBF neural network model is used to predict the SOC of monomer lithium battery,and the experimental results show that the method converges quickly,but there is a problem that the error dispersionis large and the model is not stable enough.In order to solve the above problems,this paper uses particle swarm algorithm(PSO)to optimize RBF weights,optimizes the velocity and position of particles,and picks up the parameters such as iteration number and expansion coefficient.Experiments show that the stability of the method is improved,but the fitness function converges slowly and the training speed decreases obviously.Therefore,this paper proposes to optimize RBF network by using quantum particle swarm algorithm(QPSO),which only requires the optimization of particle position,shortens the learning time,and is suitable for the on-line prediction of single lithium battery SOC.In this model,the estimation accuracy is increased to 2% when the influence of temperature characteristics is considered,which basically meets the expected requirements.Secondly,in order to observe the temperature characteristics of the battery when charging and discharging,a temperature monitoring system of lithium battery based on LabVIEW platform is developed,and the sampling frequency of the synchronous battery detection system can be collected on line,and the multi-point temperature data of monomer battery and battery pack,both temperature correction,filtering,upper and lower limit alarm and fault data prediction function.The experiments show that the detection accuracy of the temperature monitoring system is up to ±0.2℃,which can effectively manage the safety of the batterysystem.The lithium battery temperature monitoring system also provides data support and test sample base construction for the SOC prediction of the battery pack.Finally,the unbalance of battery pack capacity and Temperature Division is analyzed,the multi-channel temperature monitoring system established by LabVIEW platform is combined with the sensor built into26650 battery pack,the temperature data in the actual working condition are identified from different time scales,and the SOC online prediction is carried out by using QPSO-RBF prediction model.By adjusting the number of iterations to the optimal and ensuring the accuracy,the training time is greatly shortened and the accuracy of the on-line estimation is up to 1.4%,which verifies that the QPSO-RBF neural network has good model test accuracy and robustness in the modeling and analysis of lithium battery pack.In summary,the improved neural network model proposed in this paper effectively improves the prediction accuracy of lithium battery and lithium battery pack SOC,and the prediction process stability is good,the temperature monitoring system of battery pack is practical and has a good application prospect. |
PDF Full Text Request