Reservoir Parameter Prediction Using Geological Waveform Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current reservoir parameter prediction methods in the depth domain suffer from low accuracy, overfitting, and a lack of geological characteristic constraints, leading to inefficiencies and inaccuracies in seismic data processing.
Innovation Solution
A method involving the selection of dominant seismic attributes, waveform classification, and the construction of deep neural network models with LSTM-RNN architecture, constrained by geological characteristics, to predict reservoir parameters accurately, incorporating seismic and well logging data for optimization and fusion into spatial variation neural network models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single universal neural network model is used to predict reservoir parameters across different sedimentation environments, then the model structure is simple and easy to implement, but the prediction accuracy decreases and the model is prone to overfitting
Solution Approach 1:
The patent segments the prediction task by classifying seismic waveforms into different types based on geological characteristics (such as sedimentation environments). Instead of using a single universal model, multiple specialized neural network models are constructed, each tailored to specific waveform types. This segmentation allows each model to focus on particular geological conditions, improving prediction accuracy while maintaining reasonable complexity through modular design.
2Loss of time
If depth-domain data is directly used for reservoir parameter prediction without conversion, then the prediction process is efficient and time-saving, but the theoretical foundation is insufficient and basic theory is lacking
Solution Approach 1:
The patent performs preliminary classification of seismic waveforms into different types based on geological characteristics before the actual reservoir parameter prediction. This preliminary action establishes a solid theoretical foundation by organizing depth-domain data according to geological meaning, enabling subsequent predictions to be both efficient and theoretically sound. The classification step prepares the data structure needed for accurate depth-domain prediction without requiring time-consuming conversions.
3Ease of manufacture
If conventional inversion technology with depth-domain wavelet extraction is used, then the prediction process is simplified, but a proper theoretical model based on depth-domain data has not been established
Solution Approach 1:
The patent changes the approach from traditional wavelet extraction to a neural network-based direct prediction framework in the depth domain. By transforming the theoretical model from conventional inversion to deep learning-based prediction, the patent establishes a new theoretical foundation that is both simpler in implementation and more robust. The neural network models learn complex relationships directly from depth-domain seismic and well logging data, eliminating the need for complex wavelet extraction while maintaining theoretical soundness.
4Adaptability or versatility
If depth-domain seismic data is converted to time-domain data for prediction, then conventional prediction methods can be applied, but accumulative conversion errors occur and the process is time-consuming
Solution Approach 1:
Instead of converting depth-domain data to time-domain to apply conventional methods, the patent inverts the approach by developing specialized neural network models that work directly in the depth domain. This inversion eliminates the need for depth-time conversion, avoiding accumulative errors and time-consuming processing while maintaining compatibility with conventional prediction objectives. The models are trained to predict reservoir parameters directly from depth-domain seismic attributes and well logging data.
Data Source
AI summary
A method, an apparatus, a computer storage medium and a computer device for geological characteristic constraint-based reservoir parameter prediction are provided. The method includes: selecting dominant seismic attributes according to the relevance between different types of seismic attributes of a target stratum and reservoir parameters (S100); on the basis of the dominant seismic attributes, classifying seismic waveforms of the target stratum by a preset waveform classification network model and according to waveform features, so as to obtain a waveform classification result (S200); taking the waveform classification result as a constraint, and constructing different deep neural network models corresponding to different geological characteristics (S300); fusing different trained deep neural network models into a set of spatially varying neural network prediction model (S500); and predicting the reservoir parameters of the target stratum.