基于时间序列聚类的动力电池配组方法研究

Power battery grouping is one of the key issues to the fields of electric vehicles, etc.A successful grouping of batteries can improve the quality of the batter pack, reduce the cost of business services, and reduce environmental pollutions. The project proposes a time series clustering based battery grouping method, which selects the battery charge/discharge characteristics as the time series. The featured shapes which are sub-sequences of the original time series are extracted to characterize the original time sequence; battery grouping is then accomplished by a proper clustering algorithm based on these featured shapes, using an appropriate distance measure. The main research contents of this project: (1) featured shapes unsupervised learning methods for time series; (2) clustering method based on the featured shapes; (3) battery grouping technology based on time series clustering. By studying this project, we are expected to reveal the key external characteristics of batteries to the consistency of them; to propose an efficient high-performance time series clustering algorithm based on featured shapes, which can reduce the influence of noises existing in the time series; and to improve the performance while ensuring the efficiency for power battery grouping.

动力电池配组技术是电动汽车等领域的关键技术之一，成功的动力电池配组可以提高提高二次电池质量、降低企业的服务成本、减少环境污染。本项目提出一种基于电池多种充放电特性时间序列聚类的动力电池配组方法：通过对充放电特性时间序列进行非监督学习，抽取出能表征原始时间序列的若干特征子形状；进而基于这些特征子形状，基于合适的距离度量，采用合适的聚类算法完成动力电池的配组。主要研究内容：（1）时间序列的特征子形状非监督学习方法；（2）基于特征子序列集合的时间序列聚类方法；（3）基于时间序列聚类的动力电池配组技术。通过本项目的研究，有望揭示表征动力电池一致性的关键外在特性；提出一种高效高性能的基于特征子形状集合的时间序列聚类算法，有效降低时间序列中噪声对聚类效果的影响；在保证动力电池配组效率的同时，提高动力电池配组成功率。

动力电池配组技术是电动汽车等领域的关键技术之一，成功的动力电池配组可以提高电池组质量、降低企业的服务成本、减少环境污染。主要研究内容：（1）时间序列的特征子形状非监督学习方法；（2）基于特征子序列集合的时间序列聚类方法；（3）基于时间序列聚类的动力电池配组技术。项目设计并实现了多路动力电池在线数据采集与监测系统，提出了基于小波降噪和数据标准化的充放电曲线预处理方法，采用欧式距离和动态时间弯曲距离对特征进行相似性度量，研究了静态特征、峰度和偏度等自定义特征、特征子形状等特征抽取方法，提出了图割法、特征分布密度模型法、近邻传播法、密度峰快速查找聚类法、自组织神经网络法和自编码神经网络法等动力电池配组方法，有效提高了动力电池配组效率、降低了电池组内不一致性。相关研究成果对揭示动力电池一致性关键特征、基于时间序列聚类的动力电池配组及发展我国自主知识产权的电动汽车等具有重要的理论意义和实用价值。