Joint event location and velocity model update in real-time for downhole microseismic monitoring: A deep learning approach

Abstract

Accurate event location in downhole microseismic monitoring depends largely on how accurate the velocity model is reconstructed. On the flipside, construction of an accurate velocity model is always an uphill task marred with numerous uncertainties. However, with proper inversion approach, both the model and the location of the events can be jointly evaluated. We propose, in this study, a deep learning approach for locating microseismic events and performing velocity model update, in real-time, efficiently and accurately. Both tasks were considered as a multi-dimensional and non-linear regression problem and a multi-layer two-dimensional (2D) convolutional neural network (CNN) was designed to perform the inversions. In training the neural network, low signal-to-noise ratio (SNR) synthetic microseismic data were used to mimic field data. Overall results indicate that the CNN is capable of learning the relationship between the microseismic waveform data and the events locations and update the velocity model, in real-time, to a high degree of precision compared to classical methods. The errors in the inversion results are less than a few percent. In addition, deep learning offers a number of benefits for automated and real-time microseismic event location and velocity model update, including minimal preprocessing, continuous improvement in performance as more training data is obtained, as well as low computational cost after the network has been trained.