[H-SC06] CCUS (Carbon Dioxide Capture, Utilization, and Storage) for Climate Mitigation

[HSC06-10] Detection of carbon dioxide seepage in the ocean using machine learning

Yuka Matsumoto¹, *Toru Sato¹ (1.University of Tokyo)

Keywords:Detection of CO2 underwater seepage , Dissolved inorganic carbon, Partial pressure of CO2, Dissolved oxygen, Machine learning

CCS is a technology that separates carbon dioxide (CO2) from gas emitted from factories and stores it by injecting the captured CO2 under a shielding layer. A large-scale demonstration experiment is being conducted in Tomakomai, and currently monitoring is being conducted to see if the CO2 stored under the seabed is seeping into the ocean. In monitoring, the Ministry of the Environment stipulates that a reinvestigation should be conducted if the upper 95% prediction interval of the curve relationship based on the power approximation of dissolved oxygen saturation and CO2 partial pressure is exceeded, but the frequent occurrence of reinvestigations due to false positives is an issue. Therefore, the purpose of this study is to build a model that can more accurately determine CO2 leakage in the ocean using machine learning and to consider the influence of each index and feature value on the model. To reduce false positives of CO2 seepage in the ocean in CCS, we built a machine learning model that can more accurately determine CO2 seepage in the ocean and also evaluated the importance of each feature value and index. As a result, we were able to build a machine learning model using random forest that exceeds the accuracy of the judgments using the current Ministry of the Environment index and the indexes of previous research. When using a value obtained by combining raw data as an index as a feature, or when using the raw value of the data before being combined as a feature based on the indexes of previous research, the accuracy was higher when using the combining raw data as an index as the feature, regardless of the algorithm.