Sahar Bakhshian, Katherine Romanak Bureau of Economic Geology The University of Texas at Austin
Soil gas monitoring is an important component of the environmental monitoring package to demonstrate storage permanence at a geologic CO₂ storage site. Soil gas monitoring generally consists of making point measurements of CO₂ concentration (or CO₂ surface flux) at small areas within a larger area of review to detect potential leakage from the storage site. However, CO₂ already exists naturally within soils, derived mostly from microbial and root respiration; thus, it is necessary to determine whether increased concentrations of soil CO₂ represent leakage or some other natural perturbation. In this study, we present a predictive anomaly detection framework, DeepSense, which is employed to soil gas concentration data streams acquired from sensors being used for environmental characterization at a prospective CO₂ storage site in Queensland, Australia. DeepSense takes advantage of deep-learning algorithms as its predictor module and uses a process-based soil gas method as the basis of its anomaly detector module. The proposed predictor framework leverages the power of convolutional neural network (CNN) algorithms for feature extraction and simultaneously captures the long-term temporal dependency through long short-term memory (LSTM) algorithms. To define a reliable threshold level for anomaly detection, we employ a processed-based methodology, which relies on gas ratios that define the physical process of soil microbial and root respiration in the vadose zone rather than CO₂ concentrations. The proposed process-based anomaly detection method is a cost-effective alternative to the conventional concentration-based soil gas methodologies which rely on long-term baseline surveys for defining the threshold level. The results indicate that the proposed framework performs well in diagnosing anomalous data in soil gas concentration data streams. The robustness and efficacy of the DeepSense were verified against data sets acquired from different monitoring stations of the storage site.