一种油井流量预测处理方法、装置、设备及介质
By introducing a mechanistic loss function layer and a boundary constraint loss term into the oil well flow prediction model, and combining it with the ESP mechanism, the problem of high dependence on data volume in existing technologies is solved, and efficient small-sample learning and accurate prediction are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA UNIV OF PETROLEUM (BEIJING)
- Filing Date
- 2022-08-11
- Publication Date
- 2026-07-17
AI Technical Summary
Existing DNN-based oil well flow prediction models require a large amount of real sample data for training, resulting in low interpretability, high time cost, and low prediction efficiency.
An initial well flow prediction model with a mechanistic loss function layer is adopted. By designing a mechanistic loss function with smoothing function loss term, control equation loss term and boundary constraint loss term, combined with the ESP mechanism, the dependence on data volume is reduced and the model convergence speed is accelerated.
It improves the efficiency of oil well flow prediction under small sample data conditions, reduces model training time costs, effectively resists the impact of narrow data bandwidth, and improves prediction accuracy.
Smart Images
Figure CN115293434B_ABST