Research On Prediction Algorithm Of Aluminum Electrolytic Cell State

Aluminum electrolysis is a very complicated industrial process,and there will be a large amount of data generated in this production process.Through in-depth analysis and mining of these data,find potential information,analyze its relationship with the state of the tank,and analyze the state of the tank.Cluster analysis was performed.Then on the basis of clustering,the state of the tank was predicted.The prediction results facilitate professionals to discover the changes in the state of the tank in a timely manner,provide reference for subsequent decision-making,can make decisions in advance to reduce losses,and also improve intelligent control in aluminum.Development and application of the electrolytic industry.The main tasks are as follows:1.The process of collecting and normalizing the production data of the aluminum electrolytic cell is introduced,and then the nine features used are analyzed by correlation using the pearson correlation coefficient.Finally,the clustering method is used to cluster the state of the aluminum electrolytic cell.This paper introduces a DPCA-GMM algorithm.In order to solve the problem that the GMM(Gaussian Mixture Model)algorithm cannot adaptively cluster the number of K,it is proposed to first adopt The fast search and find of density peaks)algorithm determines its optimal clustering number K,and applies the value of K determined by the algorithm to the GMM algorithm.Experiments show that this improvement reduces the uncertainty caused by the random selection of the number of clusters K when the GMM algorithm is applied.By clustering the states of aluminum electrolytic cells,it can be used to distinguish between normal and abnormal states of aluminum electrolytic cells.2.Aiming at the problem of predicting the state of the aluminum electrolytic cell,a K-LSTM algorithm is proposed.This algorithm is based on the traditional LSTM algorithm,and aims at the problem of sample imbalance caused by the infrequentchange of the state of the aluminum electrolytic cell.In the oblivion gate with this problem,the method of setting weights is adopted to eliminate sample imbalance.Experiments prove that the algorithm can effectively predict the state of aluminum electrolytic cell.Compared with the traditional LSTM algorithm and EA-LSTM algorithm based on attention mechanism,the accuracy has been greatly improved.3.Using the DPCA-GMM algorithm and K-LSTM algorithm proposed in the article,an intelligent analysis system for the status of aluminum electrolytic cells is designed and implemented.This system can visually display the production data of aluminum electrolytic cells,including multidimensional analysis,correlation analysis,and the status of the cells.The result display of cluster analysis and the result display of tank state prediction make the management of aluminum electrolytic cell more convenient and intuitive.