Delayed coking is an important process used to convert heave oils to light products. Energy consumption and fight oil yield are important factors for evaluating the delayed coking process. This paper analyzes the energy consumption and product yields of delayed coking units in China. The average energy consumption shows a decreasing trend in recent years. The energy consumption of different refineries varies greatly, with the average value of the highest energy consumption approximately twice that of the lowest energy consumption. The factors affecting both energy consumption and product yields were analyzed, and correlation models of energy consumption and product yields were established using a quadratic polynomial. The model coefficients were calculated through least square regression of collected industrial data of delayed coking units. Both models showed good calculation accuracy. The average absolute error of the energy consumption model was approximately 85 MJ/t, and that of the product yield model ranged from 1 wt% to 2.3 wt%. The model prediction showed that a large annual processing capacity and high load rate will result in a reduction in energy consumption. Delayed coking is an important process used to convert heave oils to light products. Energy consumption and fight oil yield are important factors for evaluating the delayed coking process. This paper analyzes the energy consumption and product yields of delayed coking units in China. consumption shows a decreasing trend in recent years. The energy affects different refineries vary greatly, with the average value of the highest energy consumption approximately twice that of the lowest energy consumption. The factors affecting both energy consumption and product yield were analyzed, and correlation models of energy consumption and product yields were established using a quadratic polynomial. Both model showed good calculation accuracy. The average absolute error of the energy consumption model was approximately 85 MJ / t, and that of the product yield model ranged from 1 wt% to 2.3 wt%. The model prediction showed that a large annual processing capacity and high load rate will result in a reduction in energy consumption.