Abnormal Identification of Steam Turbine Surge in Thermal Power Plants based on Fast Fourier Transform

In response to the abnormal surge problem of steam turbines in thermal power plants, traditional methods usually use time-domain analysis to identify anomalies. This method only observes signal changes from a single time perspective, resulting in a low F₁ value in the identification results. Therefore, a method for identifying steam turbine surge anomalies in thermal power plants based on fast Fourier transform is proposed. Using the fast Fourier transform algorithm to process the sound signal of the steam turbine, converting the time waveform signal into a time-frequency domain signal, in order to observe signal changes from multiple perspectives. Introduce the concept of sample entropy to analyze time-frequency domain signals and capture the feature information contained in the signals. Constructing a neural network immune recognition model, matching and detecting the acoustic signal features input, can output the recognition results of surge anomalies. The experimental results show that the recognition result F₁ value provided by this method is stable above 0.9, proving that it can accurately identify the abnormal surge of steam turbines in thermal power plants.