Aiming at the complexity of the aluminum electrolysis production process, a soft measurement model of polar
distance is proposed based on grid-based shared nearest neighbor(GNN) clustering algorithm and least square support vector
machine(LS-SVM). In this model, GNN is used to separate a whole training data set into several clusters with different
centers, each subset is trained by LS-SVM and sub-models are developed to fit different hierarchical properties of the process.
New sample data that represent new operation information are introduced in the model, so the model can be updated on-line.
The simulation results show that the soft-sensing of polar distance based on GNN LS-SVM model can supply real-time and
accurate information for the operating optimization in the aluminum electrolysis production process.