Clustering algorithm based battery energy storage performance analysis method

Battery energy storage systems have been widely used in modern power systems. However, for a complex system with huge amount of batteries, the healthy, reliability and safety is still a big challenge. During the work of these systems, while one or several individual cells deteriorate or even broken, the entire battery pack will be affected. Fortunately, with the development of IOT technologies, an increasingly large amount of data can be collected from energy storage stations. Such data may be useful to understand battery status with certain big data tools. In this paper, k-means and DBSCAN clustering algorithm are introduced to identify and deteriorated batteries. Three parameters are proposed from the battery data, as an input model of clustering algorithms. The number of clusters and weight assignment are also adjusted considering battery’s special properties. The research used a lead-carbon energy storage system in establishing the method in the proof of concept work. As the result of clustering, an analysis of battery pack’s consistency was revealed and aged batteries were located. At the end, we verified the results by comparing the charging voltage curves of the batteries. This method may help the researcher to find deteriorated batteries in an easy way.