Abstract: The aluminum electrolysis process is characterized by complex mechanisms and harsh production environments.Currently, cell control systems mainly rely on single-mode online voltage detection and limited offline measurements, resulting in incomplete and untimely information acquisition.This hinders the optimization of process parameters and makes it difficult to achieve intelligent management upgrades.As a key technology for enhancing information perception, soft sensor has demonstrated strong predictive capabilities and practical value, playing a significant role in improving information detection levels and promoting the intelligent transformation of the aluminum electrolysis industry.This paper reviews the development status of soft sensor technology and its typical applications in aluminum electrolysis processes.It first introduces the basic principles and modeling procedures of soft sensor, then focuses on three main modeling approaches:mechanism-based modeling, data-driven modeling, and hybrid modeling that integrates both.Furthermore, it outlines representative applications in aluminum electrolysis, including alumina concentration prediction, anode effect warning, and superheat estimation.Finally, future development directions of soft sensor in complex industrial environments are discussed.