A neural network-based method for degradation estimation and adaptive fault-tolerant control of gas turbines*

The traditional fault-tolerant control method based on model switching cannot adapt to the control needs of gas turbines in realistic scenarios. Addressing this problem, this paper introduces a neural network-based adaptive model predictive control approach for gas turbine in degradation scenarios. Firstly, an online estimation method for gas turbine efficiency coefficient and flow coefficient based on extreme learning machine model is proposed. Secondly, a gas turbine rotation speed prediction model considering the degradation coefficients is established using the extreme learning machine. Based on the degradation estimation and speed prediction model, the model predictive control of the gas turbine is realized. Simulation experiments have proved that the extreme learning machine can achieve highly accurate degradation estimation and speed prediction. The proposed method has adaptive fault tolerance under different degradation degrees and variable operating conditions. Compared with traditional PID control, the proposed method has better control accuracy and can effectively improve the surge margin and economy. The proposed method also has more robust anti-interference ability. The proposed method has attracted attention for its superior control effect and practicality.