在保障国家能源安全和实现“双碳”的双重目标驱动下，新能源产业快速发展。其中，锂离子电池被广泛应用于动力和储能系统。保证锂离子电池产品全生命周期内的安全性和高效性是当前亟待解决的问题，这依赖于智能化的电池管理系统对电池系统的监控和维护，而实现其各项功能的基础是准确的电池电荷状态参数。基于上述问题，本文提出了一种将数字孪生技术应用于电池全生命周期管理的解决方案，并从电池运行阶段的数据驱动建模和电荷状态估算入手，开展了电池管理软件、硬件和算法的设计与研究。本文主要研究内容如下：（1）分别对数字孪生技术在电池领域研究现状进行了综述。为了将电池数据有效的用于电池设计、运行和回收阶段并解决现有的本地算力和存储能力不足问题，提出了结合数字孪生技术的边-云协同结构电池管理解决方案。（2）基于模块化的硬件架构，设计并集成了一套电池数据采集和无线通讯网关，实现了电池管理系统的部分功能。作为数字孪生平台的数据源和终端，该网关将采集的电流、电压和温度数据发送至云服务器，并能够执行远程控制指令。在云端服务器基于GIN框架部署了后端服务平台，实现数据存储、电池模型更新和状态计算以及可视化功能。验证了数字孪生系统的数据采集、传输、存储、计算和控制指令回传的功能闭环.（3）搭建了电池性能测试台架，开展了电池性能测试试验和数据分析，建立了多温度梯度多工况下的电池运行数据集。将呈时间序列的电流、电压和温度数据作为CNN-LSTM网络的输入，建立数据驱动电池模型并将其部署于数字孪生电池平台。在稳定温度工况下验证该模型输出电压平均绝对误差低于7.1 mV，计算延时低于150 ms，具有较高的精度和实时性。（4）将上述模型和自适应无迹卡尔曼滤波算法结合，进行荷电状态估算。经在线实验验证，在各固定温度和变温工况中，荷电状态平均绝对误差约为0.72%，并且在有初始误差时能够快速收敛，表现出较好的鲁棒性。
Driven by the dual objectives of ensuring national energy security and achieving carbon peaking and carbon neutrality goals, the new energy industry is developing rapidly. Among them, lithium-ion batteries are widely used in power and energy storage systems. Ensuring the safety and efficiency of lithium-ion battery products throughout their life cycle is a controversial issue, which highly depends on the monitoring and maintenance of the battery system by an intelligent battery management system. And the foundation for achieving functions of the intelligent battery management system is accurate battery state parameters. Due to the above problems, this paper proposes a solution to apply digital twin technology to the whole life cycle management of batteries. And starting from the data-driven modelling and state of estimation, this paper carries out the design and research of software, hardware and algorithms. The main research components of this paper are as follows:(1)A thorough literature review of the state-of-art research on the topic of digital twin technology, battery modelling and battery state of charge estimation methods. A battery management solution incorporating digital twin technology is proposed in order to effectively use battery data for the battery design, operation and recycling phases and to solve the existing problem of insufficient local computing power and storage capacity.(2)Based on a modular hardware architecture, a set of battery data acquisition and wireless communication gateways are designed and integrated to implement some of the functions of the battery management system. As the data source and terminal of the digital twin platform, the gateway sends the acquired current, voltage and temperature data to a cloud server and is capable of executing remote control commands. A digital twin battery platform was built based on the GIN framework in Golang and deployed on a cloud server. The platform stores data from the battery production and operation process in a database, and it runs battery model updating and state estimation algorithms in the background. It can visualize the data in the front-end. The functional closed loop of data collection, transmission, storage, calculation and control command return of the digital twin system was verified. (3)A battery performance test bench was built. Battery performance tests and data analysis were carried out. And battery operation data sets under multiple operating conditions with different temperature gradients were established. Time series of current, voltage and temperature data were used as input to a CNN-LSTM network to build a data-driven battery model. And it was deployed on the digital twin battery platform. The model was validated to have high accuracy and real time performance with an average absolute error of less than 7.1 mV in output voltage under different temperature conditions and calculation delay less than 150 ms. (4)The above model is combined with the adaptive unscented Kalman filter algorithm to perform battery state of estimation. The average absolute error of the SoC is about 0.72% in each fixed temperature and variable temperature condition, and it can converge quickly when there is initial error, showing good robustness.
