Time-Series Prediction of Iron and Silicon Content in Aluminium Electrolysis Based on Machine Learning

Linsheng Chen,Xiaojing Sheng,Yongming Wu,Yingbo Liu,Tiansong Liu

doi:10.1109/access.2021.3050548

Abstract

In analyzing dynamic characteristic of time-series data, classic prediction models rely heavily on static historical data, and tacit knowledge is difficult to be mined effectively. Therefore, a hybrid prediction model GS-GMDH is proposed based on growing neural gas (GNG) and the group method of data handling (GMDH). Firstly, a dynamic prediction mechanism, based on an incremental learning algorithm and time-series prediction, is established by GS-GMDH, by which the singularity is recognized and the prediction efficiency is improved. Secondly, to compare the performance of the proposed method, the multi-step ahead predictions with time-series data onto iron and silicon content are employed, and the new model is compared with classic machine models. Finally, the results show that the hybrid prediction model (GS-GMDH) proposed in this paper ensure an accurate and efficient prediction of time-series data for iron and silicon content.

Highlights

The rapid development for the aluminium electrolysis, from which waste-water and discarded aluminium are generated, has caused the destruction for the ecosystem [1]
The first is the phase of singularity recognition, the GNG-based singularity recognition algorithm (GS) incremental learning model is used to monitor the dynamic characteristic of time-series data (2, 3 and 4-step ahead predictions are considered), which the monitoring results meets the condition is the premise for triggering prediction mechanism (The establishment of preset conditions should refer to the actual needs of the factory); The second is the prediction phase, five ML models are employed to predict the contents of iron and silicon, namely MULTILAYER FEED-FORWARD NEURAL NETWORK (MLFFNN), ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM (ANFIS) (ANFIS-GP, ANFIS-SC and ANFIS-FCM), ANFIS-genetic algorithm (GA), ANFIS-Particle swarm optimization (PSO) and group method of data handling (GMDH); The final evaluation indicators are used to select the best performance model
WORK A hybrid prediction model GS-GMDH is developed based on growing neural gas (GNG) and GMDH, which includes the singularity recognition and prediction phases

Summary

INTRODUCTION

The rapid development for the aluminium electrolysis, from which waste-water and discarded aluminium are generated, has caused the destruction for the ecosystem [1]. L. Chen et al.: Time-Series Prediction of Iron and Silicon Content in Aluminium Electrolysis Based on Machine Learning. An incremental learning clustering GNG-L is proposed by Wu et al [33], which can monitor the dynamic characteristic of real-time data, and on the basis of its research, the GNG-based singularity recognition algorithm (GS) is developed in this paper. The prediction of Fe and Si content in the electrolysis process is proposed as a new research method in this paper, the GS-GMDH model is used to predict the time-series data of iron and silicon content. GS-GMDH can efficiently and accurately predict the content of iron and silicon in aluminum electrolysis with the less and noisy data.

CASE STUDY

THE GS ALGORITHM BASED ON GNG

PERFORMANCE EVALUATION MEASURES AND EXPERIMENTS

THE SINGULARITY RECOGNITION PHASE-GS MODEL

Findings

CONCLUSION AND LATER WORK