The real-time measurement of key performance parameters in the pulverizing system represents a bottleneck forrefined operation in coal-fired power plants. Addressing the complex trade-off between pulverizing power consumption andboiler combustion efficiency，this paper proposes research and application of control optimization technology for coal mills inthermal power units based on a comprehensive energy efficiency index. Data are collected through performance tests coveringmultiple coal types and wide load ranges，enabling the construction of a Support Vector Regression(SVR)soft measurement model for accurate estimation of parameters such as coal powder fineness and concentration. An optimization model targetingcomprehensive energy efficiency of the pulverizing system is established，with Genetic Algorithm(GA)employed to solve for optimal adjustable parameters. Industrial application on a 600 MW supercritical unit confirms that the system identifiesoptimization strategies differing from conventional practices. By appropriately relaxing coal powder fineness requirements， pulverizing-specific power consumption isreduced by 4.09%，while the unburned carbon content in fly ash is significantly lowered，thereby achieving comprehensive energy efficiency improvements across the entire "pulverizing-combustion" chain.