Abstract:Operational optimization of the Hall-Héroult cell is essential for achieving high efficiency and cost-effectiveness in the aluminum electrolysis process. Due to the complicated mechanism and variable working conditions, manual operational decision-making is extensively used in practice. They challenge the reliable and optimal operation of aluminum electrolysis process. In this paper, we develop a data-knowledge-driven decision-making support system (DMSS) to achieve operational optimization for the aluminum electrolysis process. DMSS consists of a prediction model, a multi-objective optimizer, and a knowledge-guided decision-making module. Specifically, we propose a working-conditions-based attention with the exogeneous inputs auto-regressive neural network (WCA-NARX) to construct a data-driven heat balance indicator (HBI) prediction model, where the working condition-related variables serve as covariates to enhance predictability. In addition, the designed structure of introducing working condition information through an attention mechanism can decouple covariates from operational variables and autoregressive variables, facilitating subsequent operational optimization. Then, a novel knowledge-assigned reference vector evolutionary algorithm (KRVEA) is designed to solve the multi-objective optimization problem of the aluminum electrolysis process, in which Pareto front solutions can be solved in the preferred region. Finally, we utilize the knowledge base that stores historical optimization cases to make decisions regarding the selection of a practical-requirement-based control scheme from the Pareto set. Real-world industrial experiments demonstrate that DMSS can effectively enhance control performance and achieve superior results compared to other competitive methods. The source code is available at https://github.com/wjiecsu/WCA-NARX.

Application Research on Neural Network Predictive Control Technology in Aluminum Electolysis Process

Nonlinear Predictive Functional Control Based on Artificial Neural Network

The Research On The Neuro-Fuzzy Network Control Strategy For Electric Discharge Machining Process

Research of Fault Prediction of Anode Effect Based on Elman Neural Network

Progress in Aluminum Electrolysis Control and Future Direction for Smart Aluminum Electrolysis Plant

Intelligent Optimal Control in Aluminum Electrolytic

Research and application of hybrid intelligent optimal control method for operation of aluminium electrolytic processes

Aluminum Reduction Cell Situation's Forecast Based on Wavelet Neural Networks

Ti6Al4V laser alloying process monitoring by using adaptive neural network controller

Diagnosis and Decision Analysis System of High Temperature Molten Aluminum Cell based on Neural Network

A Method for Anode Effect Prediction in Aluminum Electrolysis Cells Based on Multi-scale Time Series Modeling

Prediction of technological process parameters in aluminum production based on artificial neural networks

Adaptive neural network controller for the molten steel level control of strip casting processes

Design Of Model Predictive Controller Based On Neural Systems

Intelligent Control of the Electrochemical Nitrate Removal Basing on Artificial Neural Network (ANN)

Analysis on Environmental Load for Electrolytic Aluminum Process and Study on Predication Model

Development of Data-Knowledge-driven Predictive Model and Multi-Objective Optimization for Intelligent Optimal Control of Aluminum Electrolysis Process

Study and Application of Digital Constant Current Control System for Aluminum Electrolysis

Expert Control Strategy Using Neural Networks for Electrolytic Zinc Process

Current efficiency of low temperature aluminum electrolysis studied by neural network

Development of fuzzy expert control technique for aluminum electrolysis