Real-time fluid contamination prediction using bilinear programming

摘要

A wellbore fluid comprises unknown fractions of a formation fluid and a filtrate fluid, each of unknown composition. Method 200 comprises receiving a sequence of measurements of a wellbore fluid (step 210), obtaining constraints on a sequence of filtrate fractions (230), performing a constrained bilinear optimisation (240) that minimises total measurement error over the measurement sequence subject to obtained constraints, computing minimum and maximum optimal values for the filtrate fraction (250), and providing these to a well operator (260). Measurements may be spectroscopic measurements of the filtrate and/or formation fluid, density measurements or refractive index measurements. Spectra may lie in a linear polyhedral space. Constraints may be based on initial conditions, or be temporal, and may be linear or bilinear. Advantageously, predicting fluid formation fluid purity in real time allows the operator to guide the physical sampling process, eg by forecasting a time before filtrate fraction drops below a threshold.