Transient electromagnetic deep learning inversion method based on FASSA-CNN-LSTM
By using the FASSA-CNN-LSTM model and the optimizer to dynamically adjust the learning rate in transient electromagnetic inversion, the problem of complex and inefficient hyperparameter selection in the existing technology is solved, and efficient and accurate inversion results are achieved.
CN119738888APending Publication Date: 2025-04-01XINJIANG UNIVERSITY
0 Cites 3 Cited by
Patent Information
- Application Number
- CN202411801893.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-09
- Publication Date
- 2025-04-01
AI Technical Summary
Technical Problem
The existing transient electromagnetic inversion methods have problems such as complex hyperparameter selection, inefficient and difficult to put into use on a large scale.
Method used
Using the deep learning inversion method based on FASSA-CNN-LSTM, the layered model and FASSA-CNN-LSTM model are constructed, and the learning rate is dynamically adjusted by using the optimizer to optimize hyperparameters to improve the inversion efficiency.
Benefits of technology
It realizes efficient deep learning real-time inversion of transient electromagnetic data, improves inversion efficiency and accuracy, and can better restore the resistivity distribution of the formation.
✦ Generated by Eureka AI based on patent content.
Abstract
The invention provides a transient electromagnetic deep learning inversion method based on FASSA-CNN-LSTM, and the method comprises the steps: carrying out the inversion training in a non-observation time period, determining the underground resistivity and stratum thickness range according to the investigation data of an investigation region, constructing a layered model, and carrying out the calculation of the stratified model. Sampling time-decay voltage transient electromagnetic response data corresponding to the layered model are generated through forward modeling; building an FASSA-CNN-LSTM deep learning model for training, taking the electromagnetic response data as input data, dynamically adjusting the learning rate by adopting an optimizer, and finally obtaining corresponding depth-resistivity output data; and performing inversion on the acquired real transient electromagnetic response data by using the trained FASSA-CNN-LSTM inversion model in an observation time period to obtain an actual depth-resistivity geoelectric model. Compared with a traditional time sequence deep learning inversion model such as LSTM and GRU, when the transient electromagnetic long sequence is processed, the FASSA-CNN-LSTM inversion model can more effectively reserve important information in the sequence, the time needed by training is shortened, and therefore the transient electromagnetic inversion precision and the model training efficiency are improved.
Need to check novelty before this filing date? Find Prior Art