MSE based drilling optimization using neural network simulaton

Patent metadata

Field	Value
Publication number	US-10221671-B1
Application number	US-201514799753-A
Country	US
Kind code	B1
Filing date	Jul 15, 2015
Priority date	Jul 25, 2014
Publication date	Mar 5, 2019
Grant date	Mar 5, 2019

Abstract

The method disclosed receives a data stream from an MWD system and determines the response of a specific energy (SE) relationship and a rate of penetration (ROP) relationship respectively to variables controllable by the operator, in order to enable operation at a lowest SE, or a highest Rate-of-Penetration (ROP) to SE ratio. The method utilizes artificial neural networks trained by MWD data to deduce a depth-of-cut and torque based on relationships manifesting between the various data points collected, and an SE equation and a predicted ROP is evaluated over a series of probable operating points. The method continuously gathers and analyzes MWD data during the drilling operation and allows an operator to manage the controllable parameters such that operation at the lowest SE or highest ROP or ROP to SE ratio can be achieved during the drilling operation.

First claim

What is claimed is:

1. A system for providing an optimum performance evaluating (PE) parameter, an optimum weight-on-bit (WOB), and an optimum revolutions-per-minute (RPM) in a drilling operation comprising:
a drilling rig generating a well bore having a well bore length and supplying a drilling fluid to the well bore, the drilling rig comprising a drill bit and generating a weight-on-bit on the drill bit, a rate of rotation of the drill bit, a rotary torque acting on the drill bit, a rate of penetration of the drill bit, a standpipe pressure of the drilling fluid and a flow rate of the drilling fluid;
a measurement while drilling (MWD) system in data communication with the drilling rig and providing at least a length drilled (LD) signal, a WOB signal, a RPM signal, a rate of penetration (ROP) signal, a Torque (Tor) signal, a standpipe pressure (S) signal, and a drilling flow rate (F) signal, where the LD signal corresponds to the well bore length of the well bore, and where the WOB signal corresponds to the weight-on bit on the drill bit, and where the RPM signal corresponds to the rate of rotation of the drill bit, and where the ROP signal corresponds to the rate of penetration of the drill bit, and where the Tor signal corresponds to the rotary torque acting on the drill bit, and where the S signal corresponds to the standpipe pressure of the drilling fluid, and where the F signal corresponds to the flow rate of the drilling fluid, and the MWD system in data communication with an LD data channel, a WOB data channel, a RPM data channel, a Tor data channel, an S data channel, and an F data channel, and the MWD system providing the LD signal to the LD data channel, the WOB signal to the WOB data channel, the RPM signal to the RPM data channel, the ROP signal to the ROP data channel, the Tor signal to the Tor data channel, the S signal to the S data channel, and the F signal to the F data channel;
a data processor in data communication with the LD data channel, the WOB data channel, the RPM data channel, the ROP data channel, the Tor data channel, the S data channel, the F data channel, and the output channel, where the data processor is programmed for,
receiving the LD signal from the LD data channel, the WOB signal from the WOB data channel, the RPM signal from the RPM data channel, the ROP signal from the ROP data channel, the Tor signal from the Tor data channel, the S signal from the S data channel, and the F signal from the F data channel,
establishing a plurality of filled data rows, where each filled data row in the plurality of filled data rows is established by,
sampling the LD signal from the LD data channel and generating a LD data point, and sampling the WOB signal from the WOB data channel and generating a WOB data point, and sampling the RPM signal from the RPM data channel and generating a RPM data point, and sampling the ROP signal from the ROP data channel and generating a ROP data point, and sampling the Tor signal from the Tor data channel and generating a Tor data point, and sampling the S signal from the S data channel and generating an S data point, and sampling the F signal from the F data channel and generating an F data point and,
generating a filled data row, where the filled data row comprises the LD data point, the WOB data point, the RPM data point, the ROP data point, the Tor data point, the S data point, and the F data point,
training a first artificial neural network (ANN) to provide a value for depth of cut (DOC) using the plurality of filled data rows by providing a group of inputs and a target output to the first ANN, where the group of inputs comprises the WOB data point of a given filled data row, the LD data point of the given filled data row, a data point (H) representing the standpipe pressure (S) times the drilling flow rate (F) of the given filled data row where the where the H data point of the given filled data row is equal to the S data point of the given filled data row times the F data point of the given filled data row, and the RPM data point of the given filled data row, and where the target output comprises the ROP data point of the given filled data row divided by the RPM data point of the given filled data row, thereby generating a trained first ANN,
defining a plurality of (weight-on-bit value (WOB EQN ), a revolutions-per-minute value (RPM EQN ), H value (H EQN )) points, where each (WOB EQN , RPM EQN , H EQN ) point comprises a WOB EQN value, an RPM EQN value, and an H EQN value,
quantifying a least length drilled (LD EQN ) value,
generating a depth of cut equation (DOC EQN ) value for each (WOB EQN , RPM EQN , H EQN ) point in the plurality of (WOB EQN , RPM EQN , H EQN ) points by providing an input group to the trained first ANN, where the input group comprises the WOB EQN value of the each (WOB EQN , RPM EQN , H EQN ) point, the LD EQN value, the H EQN value of the each (WOB EQN , RPM EQN , H EQN ) point, and the RPM EQN value of the each (WOB EQN , RPM EQN , H EQN ) point, and
generating a trained first ANN output and generating the DOC EQN value, where the DOC EQN value comprises the trained first ANN output, thereby generating a plurality of DOC EQN values,

determining a plurality of PE parameters using the plurality of DOC EQN values and using a representative SE equation or a representative ROP equation or both, and selecting the optimum PE parameter, where the optimum PE parameter is a single PE parameter in the plurality of PE parameters, and an optimum WOB and an optimum RPM based on the optimum PE parameter,
providing the optimum WOB and the optimum RPM to an output channel; and

a display in data communication with the output channel and displaying the optimum WOB and the optimum RPM, thereby providing the PE parameter among other optimum controllable parameters in the drilling operation, where the first ANN comprises:
an ANN 1 input layer comprising an ANN 1 WOB input node, an ANN 1 LD input node, an ANN 1 H input node, and a ANN 1 RPM input node, and where the first ANN has an ANN 1 output layer comprising an ANN 1 output neuron, and where the first ANN has at least one ANN 1 hidden layer comprising ANN 1 artificial neurons, and where each ANN 1 artificial neuron and the ANN 1 output neuron has an initial set of weights and an initial bias, and where the data processor is programmed to train the first ANN using steps comprising:
setting the ANN 1 WOB input node equal to K 1 WOB A , where K 1 is a real number and where WOB A is the WOB data point of the given filled data row raised to the A th power;
setting the ANN 1 LD input node equal to K 2 LD B , where K 2 is a real number and where LD B is the LD data point of the given filled data row raised to the B th power;
setting the ANN 1 H input node equal to the K 3 H C , where K 3 is a real number and where H C is the H data point of the given filled data row raised to the C th power;
setting the ANN 1 RPM input node equal to the K 4 RPM D , where K 4 is a real number and where RPM D is the RPM data point of the given filled data row raised to the D th power;
determining an ANN 1 output value at the ANN 1 output neuron after setting the ANN 1 WOB input node equal to K 1 WOB A , setting the ANN 1 LD input node equal to K 2 LD B , setting the ANN 1 H input node equal to K 3 H C , and setting the ANN 1 RPM input node equal to K 4 RPM D ;
comparing the ANN 1 output value and the target output and determining an error; and
modifying the initial set of weights and the initial bias for each ANN 1 artificial neuron and the ANN 1 output neuron based on the error.

2. The system of claim 1 where the data processor is programmed to generate the DOC EQN value for the each (WOB EQN , RPM EQN , and H EQN ) point in the plurality of (WOB EQN , RPM EQN , H EQN ) points using steps comprising:
setting the ANN 1 WOB i

Assignees

• Zhang Wu
• Us Dept Energy

Inventors

• Zhang Wu

Classifications

• E21B44/00 Primary

  Automatic control systems specially adapted for drilling operations, i.e. self-operating systems which function to carry out or modify a drilling operation without intervention of a human operator, e.g. computer-controlled drilling systems; Systems specially adapted for monitoring a plurality of drilling variables or conditions · CPC title

• E21B2200/22

  Fuzzy logic, artificial intelligence, neural networks or the like · CPC title

• G06N3/084

  Backpropagation, e.g. using gradient descent · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/08

  Learning methods · CPC title