Text helps filter training data to improve virtual flow meter accuracy
Text-Informed Training Data Filtering for Enhanced Virtual Flow Meter Performance
Machine Learning
Summary
In industries like oil and gas, machines that predict flow rates need good data to work well, but the data often has errors or strange readings. The authors created a method that uses both sensor data and engineers’ notes written in text to better spot and remove bad data. Their approach uses large language models to understand the text and combine it with sensor readings, which helps make the predictions more accurate. Tests on simulated oil and gas data showed this new method reduced prediction errors compared to just using sensor data alone.
What this means in practice
- •For industrial machine learning teams: Improve training dataset quality by integrating operational text notes with sensor data to enhance predictive models like virtual flow meters.
- •For energy operations teams: Use combined sensor and operator log filtering to better identify data anomalies, leading to more reliable flow rate estimates during oil and gas production.
Tested on simulated data.
Authors
Kristoffer Hjelmen Nesland, Bjarne Grimstad, Antorweep Chakravorty, Dan Sui
Abstract
In sectors such as energy, aquaculture and chemical processing, the effectiveness of both mechanistic and data-driven models is frequently limited by poor-quality training data, which obscures underlying physical laws and degrades predictive accuracy. Such quality issues commonly manifest as anomalies in the dataset, ranging from sensor faults to process upsets and incorrect operator inputs. While traditional rule-based filtering can remove a subset of the anomalies, it is unable to leverage contextual information stored as unstructured text, such as engineer comments and operational logs that explain under what conditions the data was collected. We propose a text-informed data filtering method that leverages large language models (LLMs) to integrate unstructured operational data with tabular sensor data. The method consists of two steps: (1) contextual anomaly detection by reasoning over both data modalities, and (2) generation of a binary mask that is applied to filter the training dataset for model development. To investigate the potential benefit, a steady-state soft-sensor regression model, namely a virtual flow meter that estimates well flow rates, is used as the downstream evaluation target. Experiments on a synthetic oil and gas dataset show improved model performance compared to traditional rule-based filtering. Ablation studies show that textual operator comments can provide filtering-relevant signals beyond sensor data alone, with the multimodal approach reducing downstream MAPE from 7.1% to 5.4% relative to sensor-only filtering. This work demonstrates that integrating unstructured operational knowledge with sensor data through LLMs can improve data quality for industrial machine learning applications, offering a scalable alternative to manual data curation.