Systems and methods for electromagnetic waterfront surveillance in a vicinity of an oil well
Systems, methods, and devices for performing real-time detecting and spatially-positioning a waterfront in an oil-producing reservoirs are disclosed. An example method of predicting movement of a waterfront in a reservoir may include generating a plurality of electrical signals having different frequencies with a surface electric source; injecting currents corresponding to the plurality of generated signals into the earth near a well extending into the reservoir with a surface dipole; sensing a vertical component of an electric field generated by each of the injected currents at a location in the reservoir with a sensor; detecting a location of the waterfront within the reservoir based on the received vertical components of the electric fields; and analyzing the detected vertical components of the electric fields taken on at least two different points in time with machine learning to predict a rate of movement of the waterfront within the reservoir.
1. A method of predicting movement of a waterfront within a reservoir during waterflooding of the reservoir, the method comprising:
generating a plurality of electrical signals with a surface electric source, each of the signals having a different corresponding frequency;
injecting currents corresponding to the plurality of generated signals into the earth within a predefined distance from a well extending into the reservoir with a surface dipole;
detecting vertical components of electric fields generated by the injected currents at locations in the reservoir, the vertical components of the electric fields corresponding to bulk formation resistivity changes present in the reservoir that represent the waterfront;
detecting a location of the waterfront within the reservoir based on the received vertical components of the electric fields; and
analyzing the detected vertical components of the electric fields taken on at least two different points in time using machine learning to predict a rate of movement of the waterfront within the reservoir, wherein the machine learning comprises a deep learning system;
generating a separate reservoir model for each point in time with a reservoir simulator;
converting the separate reservoir models into respective resistivity models;
generating simulated electromagnetic (EM) responses to a simulated waterfront contained within the resistivity models through EM data modeling;
forming EM data hypercubes with the simulated EM responses for each point in time; and
applying the deep learning system to the separate resistivity models and the EM data hypercubes, which form a machine learning knowledge-base.
2. The method of claim 1 , wherein injecting currents corresponding to the plurality of generated signals into the earth near a well extending into the reservoir with a surface dipole comprises injecting currents into the earth with a plurality of dipoles forming a surface-to- borehole acquisition system.
3. The method of claim 1 , wherein applying the deep learning system to the separate reservoir models and the EM data hypercubes comprises performing a series of regressions between the reservoir models and EM data hypercubes to predict the rate of movement of the waterfront within the reservoir.
4. The method of claim 1 , wherein the simulated EM responses are arranged in the EM data hypercube based on a position of the waterfront relative to the well, the frequency of the corresponding generated electrical signal, a position of a sensor that detects the vertical components of the electric fields in the reservoir, and a radial offset of a shallow surface electrode forming an electrode of the surface dipole.
5. The method of claim 1 , wherein injecting currents corresponding to the plurality of generated signals into the earth near a well extending into the reservoir with a surface dipole comprises injecting currents having a frequency in the range of 1 hertz (Hz) to 1000 Hz.
6. The method of claim 1 , wherein a dipole moment of the surface dipole is approximately 20,000 Ampere-meter (A·m).
7. The method of claim 1 , wherein analyzing the detected vertical components of the electric fields taken on at least two different points in time using machine learning to predict a rate of movement of the waterfront within the reservoir comprises predicting the rate of movement of the waterfront within the reservoir without an inversion.
8. A real-time monitoring system to detect and spatially-position a waterfront in a reservoir comprising:
a surface-to-borehole electromagnetic (“SBEM”) acquisition system comprising a plurality of electric dipoles formed on the surface, the SBEM acquisition system operable to inject currents having different frequencies into the earth and measure the resulting vertical electrical fields associated with the different frequency currents;
a data processing system operable to process the received measured vertical electric fields to enhance a signal-to-noise ratio of the measured vertical electric fields;
a machine learning data analysis system operable to detect patterns within the measured vertical electric fields and predict movement of the waterfront within the reservoir comprises a deep learning system operable to receive:
the measured vertical electric field measurements taken at different points in time, the vertical electric field measurements arranged in an EM data hypercube for point in time, and
a reservoir model for each point in time, and
wherein the machine learning data analysis system is operable to predict movement of the waterfront in the reservoir based on predicted patterns using regressions between the reservoir models and the EM data hypercubes; and
an alarm and control system operable to signal an alarm when a position of the waterfront is predicted to be a predetermined distance from the well.
9. The real-time monitoring system of claim 8 , wherein the SBEM acquisition system comprises:
a plurality of radials, wherein each radial comprises a plurality of shallow borehole electrodes spaced apart along the radial and wherein each of the shallow borehole electrodes forms a first dipole electrode;
a well casing forming a second dipole electrode; and
a surface electric source associated with each of the plurality of radials, each of the surface electric sources operable to generate the electrical signals to create the currents having different frequencies,
wherein each surface electric source is coupled to the well casing and is selectively coupled to one of the shallow borehole electrodes disposed along the radial associated with the surface electric source.
10. The real-time monitoring system of claim 8 , wherein the SBEM acquisition system further comprises a plurality of sensors disposed along a length of the well extending through the reservoir, each of the plurality of sensors operable to detect a vertical electric field corresponding to the injected currents.
11. The real-time monitoring system of claim 8 , wherein the SBEM acquisition system is operable to inject currents having frequencies within a range of 1 Hz to 1,000 Hz.
12. The real-time monitoring system of claim 8 , wherein a dipole moment of the plurality of electric dipole is approximately 20,000 A·m.
13. The real-time monitoring system of claim 8 , wherein the machine learning data system is operable to generate a reservoir model that includes a position of the waterfront in the reservoir.
14. A computer program product encoded on a non-transitory medium, the product comprising computer readable instructions for causing one or more processors to perform operations comprising:
generating a plurality of electrical signals with a surface electrical source, each of the signals having a different corresponding frequency;
injecting currents corresponding to the plurality of generated signals into the earth within a predefined distance from a well extending into the reservoir with a surface dipole;
detecting vertical components of electric fields generated by the injected currents at locations in the reservoir, the vertical components of the electric fields corresponding to bulk formation resistivity changes present in the reservoir that represent the waterfront;
detecting a location of the waterfront within the reservoir based on the received vertical components of the electric fields; and
analyzing the detected vertical components of the electric fields taken on at least two different points in time using machine learning to predict a rate of movement of the waterfront within the reservoir, wherein the machine learning comprises a deep learning system;
generating a separate reservoir model for each point in time with a reservoir simulator;
converting the separate reservoir models into respective resistivity models;
generating simulated electromagnetic (EM) responses to a simulated waterfront contained within the resistivity models through EM data modeling;
forming EM data hypercubes with the simulated EM responses for each point in time; and
applying the deep learning system to the separate resistivity models and the EM data hypercubes, which form a machine learning knowledge-base.
15. The computer program product encoded on a non-transitory medium of claim 14 , wherein injecting currents corresponding to the plurality of generated signals into the earth near a well extending into the reservoir with a surface dipole comprises injecting currents into the earth with a plurality of dipoles forming a surface-to-borehole acquisition system.
16. The computer program product encoded on a non-transitory medium of claim 14 , wherein applying the deep learning system to the separate reservoir models and the EM data hypercubes comprises performing a series of regressions between the reservoir models and EM data hypercubes to predict the rate of movement of the waterfront within the reservoir.
17. The computer program product encoded on a non-transitory medium of claim 14 , wherein the simulated EM responses are arranged in the EM data hypercube based on a position of the waterfront relative to the well, the frequency of the corresponding generated electrical signal, a position of a sensor that detects the vertical components of the electric fields in the reservoir, and a radial offset of a shallow surface electrode forming an electrode of the surface dipole.
18. The computer program product encoded on a non-transitory medium of claim 14 , wherein injecting currents corresponding to the plurality of generated signals into the earth near a well extending into the reservoir with a surface dipole comprises injecting currents having a frequency in the range of 1 Hz to 1000 Hz.
19. The computer program product encoded on a non-transitory medium of claim 14 , wherein a dipole moment of the surface dipole is approximately 20,000 A·m.