In the past, Akita Prefecture was the leading petroleum-producing prefecture in Japan. Akita Prefectural Government established an organization to explore and produce this resource, to promote chemical industries, and to aim to develop them into a prominent regional industry.

The current authority responsible for petroleum and natural gas is called the “Clean Energy Industry Promotion Division,” but its origins trace back to the establishment of the “Mining Division” in 1949.

In recent years, the Clean Energy Industry Promotion Division is no longer engaged in petroleum development but focuses primarily on expanding the introduction of renewable energy, particularly offshore wind power generation.

This lecture introduces history of the division, as well as its efforts related to petroleum development.

Continuous monitoring of stored CO₂ is strongly required to achieve safe and effective CCS. We first introduce our recent studies focused on developing a continuous monitoring system based on portable active seismic source（PASS）and distributed acoustic sensing（DAS）. For the continuous analysis of this monitoring data, machine learning emerges as a valuable tool, enabling automated seismic interpretation. These monitoring technologies including machine learning should be optimized for each CO₂ storage site. Moreover, integrating the monitoring outcomes into CO₂ behavior modeling（i.e., reservoir simulation）is crucial to predict the future fate of stored CO₂. The geologic model for the reservoir simulation should be updated（or improved）by reflecting the continuous monitoring data. Our modeling study particularly emphasizes the integration of molecular-scale and pore-scale CO₂ behaviors into reservoir-scale simulations, aiming for an accurate description of CO₂ behavior within the reservoir.

To achieve “carbon neutrality by 2050,” there are many issues related to the monitoring operation in carbon capture and storage/carbon capture utilization and storage（CCS/CCUS）and EGS projects. Mechanization and automation of various monitoring tasks are essential to promote efficiency and cost reduction in the processing and analysis of huge amounts of seismic data and continuous monitoring data. Recent AI technologies, such as deep learning techniques, have shown their ability to compensate for the shortcomings of existing methods and human tasks in CCS/ CCUS and EGS. They are considered helpful tools for DX̶still, the potential risks of applying them need to be taken care and proper actions will have to be given to arising problems. While AI technology is a promising tool for streamlining and reducing the cost of processing and analysis of seismic data and continuous monitoring data, the responsibility for any consequences of using it lies with people. Both developers and users must take responsibility for potential risks when using it. Rather than completely replacing people, AI technologies are expected to gradually penetrate the market, helping to mechanize and improve efficiency so that anyone can complete the necessary tasks.

3D Seismic data interpretation, such as horizons and faults, in oil and gas exploration typically requires significant human resources and operational time. To address this, JAPEX and JOGMEC have developed AI models to streamline the interpretation process, reduce variability among individual interpreters, and minimize the risk of overlooking potential fields.

Two types of AI models have been built:（1）structure AI model and（2）Amplitude variation versus offset（AVO）AI model. The structure AI model predicts subsurface structures, including faults and channels. To improve the AI mode, a sufficient amount of training data is necessary. Over 1000 synthetic seismograms were generated based on artificial geological models for this purpose. The AVO AI model focuses on extracting AVO anomalies for oil and gas reservoirs. Similarly, large amounts of artificial geological models were created as with the first AI model.

Finally, a workflow was established to identify prospects during exploration. The structure AI model is executed initially to extract faults and channels from the 3D seismic data, followed by running the AVO AI model to screen high- potential reservoirs. Subsequently, further investigation would be conducted through seismic phase analysis and pre- stack inversion to support subsequent exploration efforts.

Micron-size particles in sediment samples such as microfossils and volcanic glasses are useful stratigraphic correlation markers for geological investigations. The Geological Survey of Japan has been developing technologies to improve the accuracy of automated classification for the particles and their high throughput workflow. As examples of this initiative, an automated microfossil classification-extraction system and a virtual slide scanner are introduced and describe the advantages and future challenges of the automated classification in this paper.

Japan Organization for Metals and Energy Security conducted a joint research project with the Commonwealth Scientific and Industrial Research Organization in Australia in 2021-2023 to automate cuttings descriptions using artificial intelligence - machine learning techniques. The project target was to investigate the possibility of minimizing the time for cuttings descriptions and establishing quantitative descriptions.

A machine-learning model was constructed to distinguish four lithologies（sandstone, mudstone, carbonate, and volcanics）in addition to the background from dried cuttings. The samples were placed in Petri dishes for observation under a stereomicroscope, and images were captured using a camera connected to the microscope. The captured 989 images were labeled with the lithologies using the open-source LabelMe annotation software. The labeled and captured images were paired to create a dataset. Subsequently, the dataset was divided into training, validation, and testing sets to construct the machine-learning model.

Eight machine-learning models were created and using four architectures（Unet, PSPNet, FPN, Linknet）and two backbone networks（EfficientNetB7 and ResNet152）for semantic segmentation approach. The combination of PSPNet and EfficientNetB7 showed the highest Intersection over Union（IOU）values. The IOU for each lithological type on the unseen test data were 81.4 % for carbonate, 53.4 % for mudstone, 67.4 % for sandstone, and 84.5 % for volcanics. IOU were higher for carbonate and volcanics and lower for sandstone and mudstone, because the rocks have consistent appearances in color and texture, making them easier to predict. However, siltstone, which is difficult to distinguish from sandstone and/or mudstone, and rocks with similar outlook, such as dark-colored volcanics and dark-colored mudstone, are more challenging for prediction.

In conclusion, it was possible to discriminate the four lithologies with a certain accuracy. While, other challenges, such as how to automate other items in cuttings description and practical application process on the rig, are recognized to automate cuttings description.