To address the issue of insufficient sensitivity to pressure pulsation indicators during the automatic combustionadjustment process of GE 9E gas turbines，an automatic combustion adjustment scheme is proposed that takes flame intensity as thecore target parameter and integrates a micro-step method with a self-learning mechanism. Firstly，based on the monitoring principle of industrial-grade ultraviolet flame detectors for the radiation intensity of CH/C2 radical characteristic spectra(253.7 nm/206.2 nm)， the necessity of using flame intensity as a direct parameter for combustion stability in automatic combustionadjustment is elucidated:when the fuel system control command(FSRXSR)deviates from the optimal value，fluctuations in flame intensity are demonstrated to more accurately reflect the risk of flame detachment compared to pressure pulsations， thereby providing a reliable input basis for real-time control. In the proposed system，flame intensity is introduced as another combustion adjustment target parameter in addition to NOemissions，and it has been validated on the 9E unit at Huadian Zhejiang Longyou Thermal Power:when the fuel system control command(FSRXSR)is artificially increased to induce flame intensity instability，the system adjusts it back to 82.5%，at which point the flame intensity recovers to within the safe threshold range，the NOemissions are maintained at a level of less than 21 mg/Nm3，and the pressure pulsation remains below 30 mbar throughout the process. The research results indicate that the application of flame intensity，a mature monitoring parameter，to combustion adjustment effectively overcomes the limitations of traditional indicators and improves the control accuracy ofcombustion stability.