Soft Sensor for Bottom-Hole Pressure Estimation in Petroleum Wells Using Long Short-Term Memory and Transfer Learning
A machine learning soft sensor was developed to estimate bottom-hole pressure (BHP) using wellhead and topside measurements, overcoming issues with physical gauges. A Long Short-Term Memory (LSTM) model outperformed other methods on real offshore data from Brazil's pre-salt basin, achieving low error rates. The approach is robust and adaptable through transfer learning across different operational environments.