Research on a microseismic signal picking algorithm based on GTOA clustering

Abstract. Clustering is one of the challenging problems in machine learning. Adopting clustering methods for the picking of microseismic signals has emerged as a new approach. However, due to the low separation of signals generated by microseismic events in real environments and the presence of noise data, existing clustering methods often struggle to achieve satisfactory results. To address these challenges, this paper proposes a clustering method based on the Group Theory Optimization Algorithm (GTOA), combined with the Akaike Information Criterion(AIC) to form an innovative microseismic signal picking algorithm. The GTOA clustering method overcomes the shortcomings of traditional mean clustering algorithms, which are easily influenced by the initial positions of centroids, thus improving the quality of clustering results and enhancing the accuracy of microseismic signal picking. Experiments evaluate the clustering effects using three performance metrics, statistically analyzing the computational results of the GTOA clustering algorithm against six other clustering methods based on evolutionary algorithms across four datasets. The comparative results indicate that the GTOA clustering method outperforms other algorithms in terms of solution quality and exhibits good robustness. Finally, using the GTOA clustering method as the foundation for the designed microseismic signal picking algorithm, experiments comparing it with traditional signal picking methods on low signal-to-noise ratio time series data show that the GTOA-based microseismic signal picking algorithm designed in this paper achieves good picking accuracy.