Intelligent Optimization Of Decision Sequences For Aluminum Electrolytic Cell

In the industrial production of aluminum electrolysis,the decision sequence formed by parameters such as the set voltage,the set amount of aluminum and the number of fluoride salt additions are of vital importance to the stability of the electrolytic cell.However,these decision sequences are often set by technologists based on experience,and there is no systematic and scientific decision sequence to set standards.In view of this problem,combined with the characteristics of aluminum electrolysis data as multiple time series,this paper transforms the problem of decision sequence formulation into the problem of similarity matching of electrolytic cell state,and obtains high-quality decision-making sequence from historical data by matching electrolytic cell state.Intercept the status sequence of the electrolytic cell to be determined as the sequence to be checked,put it into the historical data of the electrolytic cell of the same model for retrieval and matching,and obtain the similar interval.The subsequent decision output of the similar interval is used as the decision value of the cell to be determined.The main research works of this paper are as follows:1)The original production data of an aluminum plant were preprocessed and visualized.For this raw data,we cleaned it by processing outliers and missing values,and finally normalized the data using min-max method.The method of adaptive segmentation is used to obtain the sequence to be queried from the long aluminum electrolysis multivariate time series.At the same time,the voltage fluctuation coefficient per ton of aluminum is used as the judgment index of the quality of the decision sequence,laying a solid foundation for the follow-up research.2)Aiming at the problem that the traditional DTW method can not adapt to multivariate time series and can not deal with the inclusion relationship between sequences in time series alignment,double expand it,a generalized weighted variance DTW(G-WVDTW)method is proposed,which introduces weighted variance in local distance calculation to deal with the correlation between different parameters,break the boundary constraints and continuity constraints,and enable it to solve the inclusion problem between sequences,sequence matching can be performed more reasonable to obtain accurate decision values.On this basis,the selection criteria of decision sequence are given,and the optimization method of decision sequence of aluminum reduction cell based on G-WVDTW is finally formed.3)An efficient optimization method based on Top Down segmentation is proposed to solve the problem of low efficiency caused by too many data points calculated by G-WVDTW in 2).The adaptive segmentation of the long aluminum electrolysis time series is used as the query library,and the proposed adaptive constrained weighted variance DTW(AC-WVDTW)is used for rough retrieval to obtain the sequence segment set containing the best matching interval as the candidate set.In the rough retrieval process,the lower bound distance and global constraint technology are applied to speed up the retrieval and improve the retrieval efficiency.Finally,the G-WVDTW is used to obtain the best match from the candidate set,and the decision sequence is obtained according to the method in 2),and the efficiency is improve while ensuring accuracy.4)An intelligent optimization system for aluminum reduction cell decision sequence is designed and implemented.