A novel CNN-based method for localization of temperature anomalies in RDTS system

In this study, the effect of convolutional neural network combined with cosine similarity approach on the localization of temperature anomalous events was investigated by designing Raman-based Distributed Temperature Sensing (RDTS) experiments. We first constructed convolutional neural network (CNN) models to extract temperature anomaly features. RDTS experiments were designed to obtain the dataset for training the CNN models. Subsequently, the cosine similarity between the anomalous features and the original signal was calculated to overcome the disadvantage of feature offset. Finally, the feature with the highest cosine similarity was used as the final temperature anomaly event feature. Moreover, another experiment was designed to verify the performance of the proposed method. Based on optical fiber data with a resolution of 1 m, we cut the original signal into data segments with a length of 100 sampling points and used them as input to the CNN model. The main results show that the proposed method has an accuracy of 98.37 %, a precision of 93.53 %, a recall of 75.54 %, and an overall evaluation index F1 score of 0.84 (0.81 for the Copula-based outlier detection method, and 0.75 for the quartile method). The proposed method shows excellent generalization performance on datasets with different signal-to-noise ratios, providing a novel and effective solution for locating temperature abnormal events in RDTS systems.