Abstract: With the advancement of the "Dual Carbon" strategic goals, thermal power plants face new technical challenges in fuel utilization and energy consumption control. Traditional fuel management modes, which rely on empirical rules, struggle to adapt to complex dynamic characteristics under multiple operating conditions. To address this, this paper proposes a data-driven method for fuel management and energy-saving control. By establishing a multi-source data fusion model that integrates information from fuel supply, boiler combustion, and energy consumption recovery, and by employing deep learning and dynamic optimization algorithms, adaptive regulation of fuel blending ratios and energy-saving control is achieved. Finally, a pilot test was conducted on a management platform at a thermal power plant. The results indicate that the proposed method can improve combustion thermal efficiency, reduce coal consumption, and effectively enhance the system's dynamic response characteristics.