Abstract
In modern gas turbines, signal monitoring as a basis for early warning and control has proven to be an effective tool for suppressing combustion oscillations. Conventional monitoring methods include combustion databases, time series analysis, and machine learning. However, these methods have the disadvantages of high computational complexity, limited accuracy, and insufficient versatility. To solve these problems, this paper proposes an online monitoring method based on extended state observers (ESOs). The method first uses an ESO to observe the combustion system and obtains the signal envelope. The monitoring information is subsequently obtained on the basis of the derivative information of the envelope. Considering the parameter tuning and noise immunity, the conventional high-order ESO is modified to a series of low-order ESOs. Combined with known system information, the ESO for observation can be further optimized to improve monitoring accuracy. Finally, the effectiveness and robustness of the method are verified via Rijke tube experiments.
