电池储能系统状态估计综述

Review of state estimation for battery energy storage systems

为了提高多场景应用的技术经济性,本文对电池储能系统状态估计进行了综述。首先,分析了电池性能衰减的机理,介绍了目前常用的物理建模和数据建模方法,进而对荷电状态(state of charge,SOC)和健康状态(state of health,SOH)进行了定义与关联性分析,并对电池及其系统的状态估计方法进行了汇总;其次,为了获取更多精确的电池运行数据,重点介绍了能够刻画电池内部演化机理的原位/非原位表征技术,进而分析了嵌入式电池管理系统(battery management system,BMS)实际应用的主流开发路线;第三,提出了基于联邦学习的电池储能系统状态估计方法,基于轻量化模型在本地进行电池储能系统SOC的估计以保证控制实时性,基于大数据驱动策略在云中心进行其SOH估计以保证容量可信度,由此实现云边的交互与协同;最后,对电池储能系统未来可能的发展方向和研究重点进行了预测。研究结果表明:活性锂损失是锂离子电池容量衰退的主要原因,高温、低温、过充放等滥用也会加速电池性能衰减;数据驱动在电池系统级建模与状态评估方面具有较大优势;利用原位/非原位表征技术可以获取更多的电池内部状态数据,基于FPGA的BMS轻量化建模更易实现,基于联邦学习的状态评估方法能够提高电池储能系统的智慧化运维水平。

In order to improve the technical economy of multi-scenario application,the state estimation of battery energy storage systems was reviewed.Firstly,the mechanism of battery performance degradation was analyzed and the commonly used physical and data modeling methods were introduced.Then,the definition and correlation analysis of state of charge(SOC)and state of health(SOH)were carried out,and the state estimation methods of batteries and their systems were summarized.Secondly,in order to obtain more accurate battery operation data,the in/ex-situ characterization techniques that can characterize the internal evolution mechanism of batteries were mainly introduced.The mainstream development path of embedded battery management system(BMS)in practical applications was analyzed.Thirdly,a state estimation method of battery energy storage system based on federated learning was proposed.The lightweight model was used to estimate SOC locally to ensure real-time control,and the big data-driven strategy was used to estimate SOH in the cloud center to ensure capacity reliability,so as to achieve interaction and collaboration between cloud and edge.Finally,the possible development direction and research focus of battery energy storage systems in the future were predicted.The results show that the loss of active lithium is the main reason for the capacity decline of lithium-ion batteries.The abusive conditions including high temperatures,low temperatures,overcharging and discharging can also accelerate the performance degradation of batteries.Data driven has significant advantages in battery system level modeling and state assessment.More internal state data of batteries can be obtained by in/ex-situ characterization techniques.Lightweight modeling of BMS based on FPGA is easier to implement,and state evaluation methods based on federated learning can improve the intelligent operation and maintenance level of battery energy storage systems.