多变量融合多维度的车用动力电池剩余寿命在线预测

The power batteries are essential component of the electric vehicles which has been regarded as one important part of domestic strategic emerging industries. Accurate prediction of battery remaining useful life is the key factor to ensure the security and development of electric vehicles. The study object of this project is the power lithium-ion batteries of vehicles. The project aims to solve the problem of online prediction the remaining useful life (RUL) of vehicle power batteries. Firstly, an online estimation model of ohmic resistance based on Fractional-order impedance model and parameter identification is built. And an online estimation model of the maximum available capacity using the Equivalent Circuit modeling, control strategies and filtering algorithms is established. Secondly, the data of battery aging characteristics are generated employing multi-variable fusion in multi-dimension. And an aging model is established using Grey theory, which belongs to the data driven method. Finally, the aging model is utilized to build the state equation of particle filter (PF), and then the new method of Grey-Particle Filtering (GPF) is set up. Furthermore, various intelligence algorithms are introduced to modify and improve the resampling function in the PF process, and the RUL prediction strategies of Grey-Particle Filtering are generated with the switching model of multiple intelligent optimization models. These researches will provide novel insights into the online prediction of the RUL of vehicle power batteries. The project not only has important scientific research significance, but also a very high application value in engineering for vehicles.

电动汽车是我国战略性新兴产业的重要组成部分，动力电池是电动汽车的重要部件，因此准确预测电池剩余寿命是确保电动汽车安全运行和产业发展的关键因素。本申请项目以车用动力锂离子电池为研究对象，旨在探索解决目前动力电池剩余寿命（RUL）在线预测这一难题。项目首先拟用分数阶阻抗模型和系统参数辨识构建动力锂离子电池欧姆内阻在线估算模型，用等效电路模型、控制算法和滤波算法建立最大可用容量估算模型。其次建立多变量融合多维度的电池老化数据集，基于数据驱动构建一种灰色理论的老化模型。最后用该模型构建粒子滤波（PF）的状态方程，形成灰色粒子滤波 (GPF) 新算法；引入多种智能算法改进PF中的重采样函数，最终形成多智能优化模型切换机制的灰色粒子滤波RUL预测策略。为车用动力锂离子电池RUL的在线估计提供新思路。本项目的研究不但具有重要的科学研究意义，且具有极高的工程应用价值。

动力电池是电动汽车的重要部件，车用动力锂离子电池剩余寿命（Remaining Useful Life, RUL）的准确预测是确保电动汽车安全运行和产业发展的关键因素。本项目多变量融合多维度的车用动力电池剩余寿命在线预测在以下几方面取得预期进展: (1)用Thevenin模型建立锂离子电池模型，并使用遗忘因子的递推最小二乘算法(FFRLS)辨识电池模型参数、通过基于辨识的模型参数计算出的电池荷电状态和电流积分值关系，估算出电池容量。(2)建立基于I、U、T、欧姆内阻和容量等多变量融合多维度的电池老化数据集。(3)基于数据驱动引入分数阶拓展算子对灰色模型(Grey Model, GM)进行改进，得到分数阶灰色模型(Fractional Grey Model, FGM)模型；利用粒子群优化算法在更新迭代过程中获得FGM(1,1)最优阶数，用该模型构建粒子滤波的状态方程，形成分数阶灰色粒子滤波新算法(FGM-PF)。(4)在粒子滤波算法框架上，利用噪声自适应的无迹卡尔曼产生粒子滤波的重要性密度函数，构建一种自适应的无迹粒子滤波。结合FGM-PF构建一种自适应分数阶灰色无迹粒子滤波算法( FGM-AUPF)实现对电池RUL预测。(5)针对线性优化重采样粒子滤波( LORPF)算法应用于电池RUL预测过程中过于依赖老化模型、算法跟踪性差、预测精度不高的问题; 构建滑窗灰色预测模型(Sliding window Grey prediction Model, SGM)，的离线训练和在线预测框架。将SGM框架进行电池容量的长期预测作为LORPF算法的观测值，建立SGM-LORPF算法的电池RUL预测方法。(6)建立基于灰色粒子滤波的合理多种智能优化模型（主要包含FGM-PF、FGM-AUPF和SGM-LORPF）切换机制的RUL预测系统。实验表明这些算法具有较高的电池RUL预测精度，其适用于不同工况，不同老化特性的电池剩余寿命预测，为后续动力电池管理系统的RUL实际工程应用提供理论依据。本项目研究不但具有重要的科学研究意义，且具有极高的工程应用价值，相关研究成果已发在国内外重要期刊和国际学术会议上发表10多篇高质量论文（被SCI收录8篇，EI收录6篇）。