SW-LSTM: A Novel Oil and Gas Pipeline Leak Detection Method Integrating STFT and WOA Optimization Synergy

This paper proposes an improved SW-LSTM method to solve the challenge of extracting infrasound signal features in oil and gas pipeline leaks under strong noise interference. First, SW-LSTM applies Short-Time Fourier Transform (STFT) to generate a spectrogram and combines both time-domain and frequency-domain feature extraction methods to construct a high-dimensional feature vector. Principal Component Analysis (PCA) is used to reduce the dimensionality and normalize the high-dimensional feature vector to enhance the distinguishability of the features. Second, a leakage detection model based on Long Short-Term Memory (LSTM) networks is constructed. To improve the accuracy of leakage detection, the Whale Optimization Algorithm (WOA) is introduced to optimize the hyper-parameters of the leakage detection model LSTM. Finally, the proposed method was validated on the real-world dataset with three different leakage distances (3m, 8m, and 15m). The experimental results show that SW-LSTM achieves the best performance under all three conditions. Even when the leakage point is 15m away from the detection point, the proposed method reached 96.32% in the evaluation metric F1, indicating that the proposed method has promising application potential in oil and gas pipeline leakage detection.