Intelligent control for rapidity and security of all-electric ships gas turbine under complex mutation load using optimized neural network

Due to the small grid capacity of the all-electric ship, the excessive rotor speed overshoot, slow response, and potential faults in ship gas turbines operated under high mutation loads pose a serious challenge. Therefore, this research proposes a novel intelligent control approach using genetic algorithm to optimize the state parameters and learning factors of Neural Network to achieve adaptive real-time adjustment and tracking. The precise top control system and the sub-control system of ship three-shaft gas turbine are constructed with multiple signal deviation limit by coupling physical parameters such as air inlet guide vane angle, fuel valve opening, rotor speed and temperature. Furthermore, a hybrid data-driven approach based on experimental data and mechanism principles is adopted to establish an accurate gas turbine model and controller model. Research shows that at the design point, the generate efficiency of 20 MW level gas turbine established is 32.3 %, and the overall error is within 3 % with a good accuracy. When subjected to a sudden increase in ship load from 60 % to 100 % of the rated power, the fuel rapidly increases to 2.51 kg/s and the IGV increases rapidly to 100 %. The rotor speed overshoot decreases by 4.25 %, the steady time decrease by 28.29 %, and surge margin of the Low-Pressure Compressor increase by 3.46 %. Similarly, during ship load shedding scenarios, through online parameter self-tuning, it reduces 14.709 % steady time, 37.2 % decrease maximum rotor speed overshoot, and increase 5.3 % surge margin.Importantly, both simulations and verifications suggest that this proposed approach greatly enhances the rapidity and security of gas turbine generation. This contribution provides valuable technical groundwork for future endeavors in rapid tracking and intelligent control within all-electric ship propulsion systems.