In this paper, an innovative workflow to allocate commingled oil production is explained. Despite many attempts in past years (Hwang et al., 2000 and Zhao-Wen et al., 2016), allocating production in commingled wells, i.e. assessing the relative production from multiple pay zones in different wells, is still a big issue, in particular when the end member oils are not available or budget for PLT (Production Logging Tool) campaigns is limited. Our approach overcomes the problem of having end member oils through a wise combination of experimental geochemical analytical techniques and unsupervised state-of-the-art Machine Learning computational methods: in particular, data coming from high-resolution chromatographic methods have been processed in an original way, by identifying and exploiting all the smallest chemical differences among samples. Our approach was successfully tested in a blind way (i.e., without knowing a priori the exact composition of the oils in the mixtures), using both synthetic mixtures and real commingled oils. The synthetic mixtures were prepared using end member oils produced from different levels but chemically almost indistinguishable the one from the other. The real commingled samples were collected at wellhead. The case studies took into account very different situations in term of oil composition and geographical provenance. When possible, the results were compared with the results coming from industrial standard software (in case of availability of end members oils). The precision of the proposed method ranges from 3% (2 end member oils) up to 6% (4 end members oils) and for this reason our approach can be proposed as a cheap, advanced and accurate tool for Reservoir Management best practices.
During an oil commingled production, fluids coming from different productive levels/wells are mixed. Monitoring and allocating contributions of different fluids are key aspects to optimize production: in particular, allocation refers to the practice of splitting wellhead-measured quantities of commingled oils coming from the different sublevels/wells of a reservoir. A clear identification of the production streams in terms of individual contributions in a single well, multi-well and multi-plants production scenario (see fig. 1) can facilitate timely decisions and interventions during all phases of development and management of an oilfield.
