Design And Research On Key Methods Of Integrated Battery Management System

It is the inevitable result of UAV cluster development that UAV battery management system can manage the overall energy of UAV cluster through cloud interconnection.However,completely cloud-based battery management cannot meet the real-time requirements of battery protection,and the stand-alone battery management system cannot provide comprehensive and accurate battery data analysis with limited storage and computing capacity after disconnection.Therefore,how to effectively mix the advantages of cloud storage computing and terminal realtime processing to form reliable,comprehensive and multi-scene battery life cycle information has become the key to improve the management level of clustered UAV in daily maintenance,flight control,task planning,spare parts guarantee and other aspects.The main research content of this project is as follows.Firstly,the requirements of UAV cluster and its stakeholders are analyzed,the technical requirements of integrated battery management system are determined,and the overall design scheme of the system is formulated.By building an experimental platform for the integrated battery management system based on end-cloud collaboration,the division and design of each functional module of the terminal battery management system and the cloud battery management system are realized respectively,laying a systematic foundation for the verification of key methods in the future.Secondly,a cloud-based estimation method for UAV cluster battery state is illustrated by taking battery state-of-charge as an example.Aiming at the problem of low accuracy and overfitting of battery state estimation,the improved long and short-term memory algorithm of function optimization and hyperparameter optimization was studied to estimate battery state-ofcharge,which made the convergence speed of estimation model faster,higher accuracy,alleviated the problem of overfitting,and had strong adaptability to battery data of different working conditions.Then,aiming at the problem of slow algorithm maturity due to insufficient data amount of UAV cluster battery state estimation method in the early stage of system operation,a multidimensional continuous time series data enhancement method based on generative adversarial network was used.Considering the limitation of quantity and quality of battery data obtained through flight experiment and data acquisition experimental platform,this project uses the improved generative admission-network algorithm with constraint conditions and embedded space to enhance and generate battery data,accelerate the maturity of estimation model and save experimental cost.Finally,a cloud estimation model compression method incorporating regularization,mask matrix and dropout is designed to compress the cloud battery state estimation model to an embedded battery management terminal,realizing the capability inheritance of the terminal battery management system from the cloud battery management system.Solve the problem that the battery string life cycle health management and real-time performance cannot be achieved when the cluster works intermittently.By making the decision of the terminal and cloud battery state estimation working together in different states,the end-cloud collaboration of integrated battery management system is realized.This project uses NASA battery data set and self-test battery data set to conduct simulation and experiment verification,proving the feasibility of integrated battery management architecture and the effectiveness of key methods in the architecture,which is of great significance to the further development of battery management system.