PLPengfei Lv

Pengfei Lv

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Pengfei Lv has a total of 19 co-authors. Here are the top co-authors by number of shared publications:

  • Weiying Chen

    State Key Laboratory of Deep Petroleum Intelligent Exploration and Development, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China

    4 shared publications
  • Guoqiang Xue

    Key Laboratory of Deep Petroleum Intelligent Exploration and Development, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China

    2 shared publications
  • Guoqiang Xue

    1 shared publication
  • Jian Wang

    Key Laboratory of Earth and Planetary Physics, Institute of Geology and Geophysics Chinese Academy of Sciences Beijing China

    1 shared publication
  • Jinjing Shi

    Key Laboratory of Mineral Resources, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China

    1 shared publication
  • Shupeng Lu

    Chinese Academy of Sciences 2 State Key Laboratory of Deep Petroleum Intelligent Exploration and Development, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China and , Innovation Academy for Earth Sciences, Beijing, China.;;;;

    1 shared publication

Latest Publications

  • Open Access

    Lithology identification method based on Multi-mode adaptive prediction system: Algorithms and Applications

    by Pengfei Lv, Guoqiang Xue, Weiying Chen

    Lithology identification is crucial in mineral and energy resource exploration as it determines geological composition and guides exploration activities, improving resource location and evaluation efficiency. The advancement of artificial intelligence technology has promoted the application of machine learning-based multi-source geophysical data fusion methods in lithology identification. However, due to the differences in geophysical exploration techniques and data types across mining areas, single machine learning methods often struggle to adapt to diverse geological environments, lacking necessary universality and robustness, which severely restricts the practical application of intelligent identification technology in actual exploration. To address these limitations, this study introduces a Multi-mode Adaptive Prediction System (MAPS) for lithology identification. MAPS innovatively integrates three learning models (supervised, semi-supervised, and unsupervised learning), and can automatically select the most suitable learning mode based on prior information such as the quantity and quality of existing labeled samples and the completeness of geological background information, achieving rapid and accurate lithology identification. We verified MAPS's performance advantages through extensive comparative experiments: in supervised learning mode, compared to Support Vector Machine (SVM) and Naive Bayes classifier, accuracy improved by 0.7% and 3.5% respectively, with F1 scores increasing by 3.4% and 4.5%; in semi-supervised learning mode, compared to semi-supervised fuzzy C-means algorithm and self-learning algorithm, accuracy and F1 scores improved by a minimum of 33.67% and 0.15 respectively; in unsupervised mode, compared to traditional fuzzy C-means and Gaussian mixture models, MAPS demonstrated superior ability to mine and construct internal data structures, showing stronger feature learning capabilities. Furthermore, MAPS has shown excellent performance in the practical application of coal seam location prediction. The coal seam locations predicted by the system are highly consistent with actual drilling results, further validating MAPS's significant application potential in practical engineering. In conclusion, MAPS significantly improves the efficiency and accuracy of lithology identification, providing reliable technical support for mineral and energy resource exploration with broad application prospects.

  • Augmentation and Evaluation of Geological Tabular Data: Geo-TabGAN Model and Its Applications

    by Pengfei Lv, Weiying Chen, Xinyu Zou

    Data augmentation plays a crucial role in data-driven geoscience research by minimizing sampling costs and improving the generalization and predictive accuracy of models utilized in mineral exploration and oil and gas development. Although geoscience data are predominantly structured in tabular form, research focused on the augmentation of such structured data remains in its nascent stages. This study seeks to address two fundamental questions: 1) is it feasible to generate realistic synthetic data when only a limited amount of labeled data are available? and 2) what criteria can be established to evaluate synthetic data to ensure it contributes positively to model performance? To this end, we introduce the geological tabular data generative adversarial network (Geo-TabGAN) model and propose a comprehensive evaluation framework. Experimental results derived from core analysis data of the Bayan Obo deposit in Inner Mongolia indicate that the integration of synthetic data led to improvements exceeding 5% in the average accuracy, precision, recall, $F1$ score, and Matthews correlation coefficient (MCC) across three classifiers. This methodology significantly enhances the efficacy of big data analysis and predictive modeling within the geoscience domain.


  • A deep-learning approach for augmenting tabular geoscience data

    by Pengfei Lv, Guoqiang Xue, Weiying Chen, Ya Yu, Shupeng Lu, Wanting Song

    In recent years, artificial intelligence has been increasingly applied in geosciences; however, the scarcity of labeled data limits its effectiveness. Existing data augmentation methods using conditional generative adversarial networks (cGANs) have succeeded in fields such as finance and medicine but often struggle to account for the multiscale features and strong correlations inherent in geoscience data, making direct adaptation challenging. This study introduces an improved cGANs — improved conditional geoscience GAN (ICG-GAN) — specifically designed for augmenting tabular geoscience data. ICG-GAN uses continuous features as conditional inputs to preserve their physical properties and correlations, using a classification-based voting mechanism to predict and impute discrete features. This approach effectively overcomes the limitations of traditional cGANs when learning from sparse samples. To systematically assess the model’s performance, we developed a comprehensive, multidimensional evaluation framework. Using core analysis data as a case study, experimental results show that ICG-GAN more accurately replicates the original data distribution and achieves significant performance improvements compared with the leading CTGAN model across six evaluation metrics. This study presents a novel method for augmenting tabular geoscience data, with broad applications in resource exploration, geologic mapping, and environmental monitoring, while providing insights for enhancing the performance and robustness of data-driven geoscientific models.


  • Open Access

    Electrical Structure between the Main and Eastern Deposits of the Bayan Obo Mine: Results from Time-Domain CSEM Methods

    by Weiying Chen, Wanting Song, Pengfei Lv, Jinjing Shi, Yulian Zhu

    Bayan Obo is a well-known polymetallic deposit containing significant quantities of rare earth elements, niobium, thorium, and iron. However, the epoch in which mineralization occurred and the mineralization process are still debated due to the complex nature of its mineralization and geological evolution. Inadequate geophysical exploration has further contributed to this lack of clarity surrounding critical issues, such as the deep link between the main orebody and the eastern orebody, the form and distribution of the extensive dolomite, and the geologic structures in the area. Therefore, we implemented the time-domain controlled-source electromagnetic method (CSEM) to acquire electrical structures at depths down to 2.5 km between the Main and Eastern mines. According to the inverted resistivity structure, in conjunction with existing geological and drilling data, we classified the main lithologies and faults based on their resistivity characteristics. Overall, the mineralized carbonatite reflects high to moderately high resistivity. The mineralized carbonatite dips overall from north to south, with a maximum extension depth not exceeding 1.5 km, and its range of occurrence is controlled by nearly east–west-striking faults distributed along the bounding line between the roof and floor rocks. The Main and Eastern mines are connected at depth, but the morphology and position of the ore bodies have significantly changed due to multiple phases of tectonic activity. The electrical structure does not reveal any obvious syncline structures, further refuting the traditional view that the Bayan syncline controls ore formation.


  • Determining the location of the Bayan Obo rare earth elements mineralization body by the transfer learning method

    by Guoqiang Xue, Pengfei Lv, Weiying Chen, Xiaochun Li, Ya Xu, XinWu Wu, Jian Wang, Yonggang Zhao, Xianhua Li

    Bayan Obo is the largest rare earth element (REE) deposit in the world. The occurrence of REE is closely related to the dolomite in this area. Dolomite serves as the mother rock of REE mineralization and the ore body. How to accurately locate and characterize dolomite is the key to determining the distribution of REE and estimating its reserves. A large amount of geophysical work has been conducted in this area, including a dense seismic array, various electromagnetic methods, gravity and aeromagnetic surveys, as well as numerous petrophysical property measurements. To fully leverage the results obtained by these geophysical methods and develop an understanding of the physical property structure, a multisource geophysical data fusion technology is developed. First, various physical property profiles obtained from inversion on the same profile are converted into images with identical resolution and dimension. Then, an image adaptive feature extraction technique based on transfer learning is used to extract features of different scales from multisource images. Subsequently, the fusion image is reconstructed based on the local nearest neighbor weighted average feature fusion rule to obtain the final fusion result. This aids in identifying the spatial appearance pattern of the target for detection. Given the physical characteristics of the mineralized dolomite, which has high density, high resistivity, and high magnetic susceptibility, its location and shape can be defined in the fusion image. The results indicate that the occurrence depth of dolomite can extend up to 1500 m and the dolomite has a southward tilt as one of its primary structural characteristics. The predicted range of dolomite distribution is consistent with the formation range revealed by drilling, making it a reliable basis for predicting the distribution of rare earth ore bodies.