Abstract: Regarding the widespread problem of "large machines pulling small carts" and the disconnection from production rhythm in the steel smelting dust removal system, which leads to energy waste, this paper proposes and implements an intelligent control system based on the deep perception and analysis of PLC data from production equipment. The system collects real-time PLC operating parameters of key equipment such as AOD furnaces, LF furnaces, and continuous casting machines (such as oxygen blowing flow, inclination angle, gun position, arc state, and pulling speed), and constructs a multi-source data stream reflecting the instantaneous dust emission intensity. In the central server, using data fusion and machine learning algorithms, a dynamic prediction model from "production process status" to "dust removal demand air volume" is established. The model outputs instructions and is sent to the distributed control system (DCS) in the form of commands, and by precisely adjusting the opening degree of the electric valves of the dust removal branch pipes and the speed of the main fan"s frequency converter, the dust removal air volume is provided on demand and precisely. The application of this system in the steelmaking workshop of Guangxi Beigang New Materials shows that the system successfully transformed the dust removal system from the "experience-based operation" mode with constant air volume to the "intelligent following" mode following the production rhythm. While ensuring that the dust capture efficiency meets the ultra-low emission requirements, the comprehensive power consumption of the dust removal system has decreased by more than 18%, verifying the great potential and engineering feasibility of the data-driven strategy in the field of industrial energy conservation.