As renewable energy generation continues to integrate into the grid, the demand for flexible operation in coal-fired power plants has significantly increased. However, increasing flexibility often results in decreased economic efficiency. In order to improve the operational economy of coal-fired power plants, a mechanism-driven, data-driven digital twin model for direct air-cooled units was established. Real-time data was utilized to identify key dynamic parameters in the twin model, and an optimal backpressure calculation method was proposed based on this model. This method leveraged a high-precision twin model to correct key parameters in the backpressure optimization process through historical data regression. Correlation analysis was employed to estimate the existing range of optimal fan speeds under different conditions, and an adaptive particle swarm optimization algorithm was used to optimize the net power fitness function. This enabled rapid optimization of optimal backpressure and corresponding fan speed based on the physical action laws of the system. The proposed optimal backpressure calculation method was applied to a 600 MW direct air-cooled unit. Research results show that the application of digital twin technology enhances model accuracy, reducing the average relative error of output results to below 2% compared to traditional fitting methods, with a Pearson correlation coefficient exceeding 0.9. It is concluded that the optimal backpressure and corresponding fan speed obtained through the above methods provide operational guidance for power plants, contributing to reduced fuel consumption and improved economic efficiency of the units. It is estimated that 6.82 tons of standard coal are saved daily.