Time Series Clustering Analysis And Visualization Of Aluminum Electrolysis Cell

In the long-term electrolytic cell production process,a large number of control,measurement and test data have been accumulated.Through mining and analyzing these massive data,we can understand the overall development trend of the electrolytic cell and the change and classification of the cell condition from a macro perspective,providing analysis basis for the process technicians to carry out the production management of the electrolytic cell.This paper mainly uses time series clustering algorithm to mine,analyze and visually display the historical production data of aluminum electrolytic cell,including overall time series clustering and time series subsequence clustering,to find out the similarity and relevance between the production of different electrolytic cells,and to detect the abnormal production status,so as to realize multidimensional analysis and time series clustering of production data of aluminum electrolytic cell Analysis.1.Preprocess the original production data of a large number of aluminum electrolytic cells according to the hierarchical relationship of factories,workshops and work areas,including abnormal statistics,correlation analysis,principal component analysis,etc.,and display them intuitively through a variety of charts.2.DTW(dynamic time warping)algorithm is selected as the distance measurement of time series,hierarchical clustering algorithm is used to cluster the time series data of aluminum reduction cell as a whole,fuzzy clustering algorithm is used to cluster the time series subsequence and detect the anomaly.Finally,the clustering results are visualized.3.Before clustering time series subsequences,it is necessary to segment time series.In this paper,on the basis of traditional time series segmentation algorithm based on special points,a time series segmentation algorithm based on the boundary area of trend turning point is proposed,and its performance is compared with the traditional algorithm on the public dataset.4.This paper develops and implements a time series clustering analysis system for aluminum reduction cell,which integrates data preprocessing,feature analysis,clustering algorithm,data visualization and other functions.