基于Lebesgue采样的动力电池组状态联合估计与寿命预测研究

动力电池组准确状态估计及寿命预测是保障电动汽车高效、安全可靠运行的关键技术。动力电池电化学过程复杂、可测参数有限且特性耦合，成组使用后不一致性问题突出，导致动力电池组准确状态估计及寿命预测极具挑战，且现有研究主要基于黎曼采样框架设计状态估计及寿命预测方法，存在计算成本高、模型复杂且易受噪声干扰等问题。因此，项目以Lebesgue采样理论为指导，围绕动力电池组状态参数在时间尺度上的耦合机制，及Lebesgue采样框架下动力电池组寿命衰减的表征规律这两个关键科学问题，探索基于聚类分析的动力电池组“聚类-代表-差异”模型构建、自适应时间尺度下的动力电池组状态联合估计和基于Lebesgue采样与深度学习的动力电池组寿命预测方法，以期提高动力电池组动态建模精度、突破多状态协同估计及寿命预测的技术瓶颈，并为实现电动汽车动力电池组主动管理提供可靠的理论方法和技术手段。

Accurate state estimation and life prediction of the power battery pack are key technologies to ensure efficient, safe and reliable operation of electric vehicles. Due to the complex electrochemical process, limited measurable parameters, coupling characteristics and inconsistency after group used, accurate state estimation and life prediction of the power battery pack are extremely challenging. In the existing research, state estimation and life prediction are usually carried out in the Riemann sampling framework, which leads to problems such as high computational cost, complex model, and susceptibility to noise interference. Therefore, guided by the theory of Lebesgue sampling, the project focuses on the coupling mechanism of battery state parameters on the time scale and the degradation characterization law of the state of health under Lebesgue sampling framework, and investigates novel methods for the "cluster–representative–difference" model of power battery pack, state joint estimation with adaptive time scale, and life prediction based on Lebesgue sampling and deep learning method. Its expectation is to improve the dynamic modeling accuracy, the breakthrough for the technical bottleneck of multiple states joint estimation and life prediction, and finally provide reliable theories and technical solutions for active management of power battery pack in electric vehicles.