| CPC E21B 44/00 (2013.01) [G05B 13/027 (2013.01); G06N 3/084 (2013.01)] | 14 Claims |
|
1. A method, comprising:
performing a plurality of historical well intervention events at a plurality of production wells using one or more well intervention providers and based on one or more well conditions,
wherein the plurality of historical well intervention events comprise one or more hydraulic stimulation operations, one or more flow back operations, one or more wellhead maintenance operations, one or more slickline operations, and one or more coiled tubing operations;
obtaining, by a well intervention manager and from the plurality of production wells over a well management network, training data comprising training well data and a plurality of scheduling criteria, wherein the training well data and the plurality of scheduling criteria correspond to the plurality of historical well intervention events;
training, by the well intervention manager, an artificial neural network based on the training data and a machine learning algorithm to produce a trained model;
generating, automatically by the well intervention manager and based on a predetermined scheduling criterion, a first well intervention plan for a well site,
wherein the first well intervention plan is generated using a first plurality of data inputs regarding one or more well intervention providers and one or more well conditions, and
wherein the well site comprises a wellbore, a production tree, and a wellhead assembly;
obtaining, by the well intervention manager, first real-time well data regarding the well site and second real-time well data from the plurality of production wells;
obtaining, by the well intervention manager and from a server, service provider data regarding a plurality of service entities,
wherein the service provider data describes one or more contract timing expirations and one or more service provider availabilities for a respective service entity among the plurality of service entities;
updating, automatically by the well intervention manager in real-time, the trained model using the first real-time well data and the second real-time well data, wherein the well intervention manager continues to train the trained model during performance of one or more well operations performed during the first well intervention plan;
adjusting, automatically by the well intervention manager and using the first real-time well data and the service provider data, the predetermined scheduling criterion to produce an adjusted scheduling criterion, wherein the adjusted scheduling criterion corresponds to a second plurality of data inputs that are different from the first plurality of data inputs;
generating, automatically by the well intervention manager, a predicted well intervention plan for the well site using the trained model and based on the first well intervention plan,
wherein the adjusted scheduling criterion, the first real-time well data, and the service provider data are inputs to the trained model to output the predicted well intervention plan;
transmitting, by the well intervention manager, a first command to a first control system comprising a programmable logic controller (PLC) at the well site; and
adjusting, by the first control system, a first well operation based on the predicted well intervention plan in response to receiving the first command from the well intervention manager,
wherein the first well operation comprises a hydraulic stimulation of a reservoir region connected to the wellbore at the well site, and
wherein the hydraulic stimulation of the reservoir region increases hydrocarbon recovery at the well site.
|