Summary: Geophysical logging is widely used in lithofacies identification, reservoir parameter prediction, and geological modeling. However, it is common to have well-log sections with low-quality and/or missing segments. Repeating the well-log measurements is not only expensive but might also be impossible depending on the condition of the borehole walls. In these situations, reliable and accurate well-log prediction is, therefore, necessary in different stages of the geomodeling workflow. In this study, we propose a time series regression model to predict missing well-log data, incorporating facies information as an additional geological input and using discrete wavelet transform (DWT) to denoise the input data set. The main contributions of this work are threefold: (i) We jointly use facies information with well logs as the input data set; (ii) we use DWT to denoise the input data and consequently improve the signal-to-noise ratio of the input data; and (iii) we regard the depth domain as the time domain and use a time series regression algorithm for log reconstruction modeling. We show a real application example in two distinct scenarios. In the first, we predict missing well-log intervals. In the second, we predict complete well logs. The experimental results show the ability of the proposed prediction model to recover missing well-log data with high accuracy levels.

Seismic facies classification is crucial in reservoir evaluation and guiding oil and gas exploration and development. While machine learning and deep learning models have shown promising results in this field, they often lack interpretability, hindering the understanding of their decision-making process. Additionally, the non-uniqueness of classification results and the need for uncertainty evaluation pose challenges. This study proposes an interpretable and high-precision workflow by combining the time series ensemble learning algorithm HIVE-COTE 2.0 (HC2) and the Shapley Additive exPlanations (SHAP) interpretation method. HC2, which excels in processing time-domain seismic data, leverages multiple base classifiers in diverse domains and employs the Cross-validation Accuracy Weighted Probabilistic Ensemble (CAWPE) technique to aggregate predictions, enhancing accuracy and generalization. Notably, the probability values from HC2 enable uncertainty quantification, enabling geologists and engineers to consider the model’s uncertainty in decision-making, thereby reducing risks. SHAP facilitates global and local interpretation, enabling the measurement of seismic attribute and elastic parameter importance and optimizing feature combinations for improved classification performance. Experimental results using real data demonstrate the superiority of the proposed workflow over classical machine learning and time series models, specifically in terms of single-well experiment accuracy and lateral continuity of seismic profiles.

Summary: 3D simulation of sedimentary facies using seismic data is vital for reservoir evaluation and estimation of oil and gas reserves. As a result of the high nonstationarity of sedimentary facies and the highly nonlinear characteristics of seismic attributes, the mapping relationship with sedimentary facies has certain ambiguities and uncertainties that affect the modeling results of sedimentary facies. Multipoint geostatistics (MPS) has proven to be an effective technique for modeling subsurface geological bodies. However, the conventional MPS only deals with stationary applications, and its capability of revealing the distribution of sedimentary facies is thereby limited. In addition, the task of seismic attributes selection, which significantly affects the performance of the modeling method in simulating the distribution of sedimentary facies, is difficult because the relationship between sedimentary facies and seismic attributes is complex. This article presents a nonstationary modeling method for simulating the distribution of sedimentary facies, which is featured by the multiscale spatial feature of patterns. In particular, the spatial location of the patterns is introduced as auxiliary information in the classification and simulation processes. The method incorporates multiscale results during the modeling procedure. Patterns from the multicategory training images (TIs) are classified by the optimized workflow. The seismic attribute selection is achieved by using fuzzy-rough sets. The proposed simulation method is verified by two typical TIs, followed by applications to predict the actual distribution of sedimentary facies. Compared with the filter-based pattern simulation (FILTERSIM) approach, the proposed simulation method is applicable for revealing detailed subsurface models, especially under complex geological conditions and limited information.