Methods and systems for real-time multiphase flow prediction using sensor fusion and physics-based hybrid AI model(s)

A system, comprising:

• a well with a wellbore conveying a flow of a multiphase fluid, the flow being controlled, at least in part, by a set of well operation parameters;
• a first pressure sensor and a first temperature sensor disposed in the wellbore at a first and a second location and configured to measure a first pressure data and a first temperature data;
• a multiphase flow meter (MPFM) configured to measure a first flow rate of a first phase of the multiphase fluid at a third location;
• and
• a computer comprising one or more computer processors configured to:
• obtain the first temperature data and the first pressure data of the flow of the multiphase fluid from the first pressure sensor and the first temperature sensor;
• determine, with an artificial intelligence (AI) model, a first predicted flow rate of the first phase of the multiphase fluid at the third location given an input comprising the first temperature data and the first pressure data;
• set the measured first flow rate as a trusted first flow rate, where the trusted first flow rate represents an actual first flow rate of the first phase;
• make a first determination of whether the measured first flow rate is incongruent with the first predicted flow rate;
• make a second determination of whether the MPFM is in fault;
• set the first predicted flow rate as the trusted first flow rate in response to a determination that the measured first flow rate is incongruent with the first predicted flow rate and that the MPFM is in fault based on the first determination and the second determination;
• determine, with an optimizer applied to the AI model, an optimal set of well operation parameters based on, at least, the trusted first flow rate;
• and
• adjust, automatically, the set of well operation parameters to the optimal set of well operation parameters,
• wherein the set of well operation parameters comprises a choke setting for a choke valve of the wellbore such that the optimizer applied to the AI model determines an optimal choke setting,
• wherein the one or more computer processors are further configured to automatically command the choke valve based on the optimal choke setting in order to optimize the flow of the multiphase fluid, and
• wherein the optimizer applied to the AI model is a limited-memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) algorithm.