Abstract: To enhance the operational efficiency and fault response capability of gas turbine DCS control systems, research was conducted on the optimization and fault early warning of DCS control systems for smart power plants. First, operational state features of gas turbines were extracted based on multidimensional sensor data; an adaptive model predictive control strategy was designed to dynamically compensate for DCS core loops, mitigating control deviations caused by load disturbances and fuel fluctuations. A fault early warning mechanism was established to detect early degradation features and quantify abnormal deviations, generating interpretable warning signals and identifying potential fault sources. Comparative experiments demonstrated that this method effectively improves the adaptive capability of control systems under complex operating conditions, verifying its feasibility and superiority in practical engineering scenarios of smart power plants.