U.S. patent application number 15/927453 was filed with the patent office on 2018-09-27 for system and method for monitoring disposal of wastewater in one or more disposal wells.
The applicant listed for this patent is Genscape Intangible Holding, Inc.. Invention is credited to Deirdre Alphenaar, Jason Fuchs, Forrest Webb.
Application Number | 20180274334 15/927453 |
Document ID | / |
Family ID | 63582215 |
Filed Date | 2018-09-27 |
United States Patent
Application |
20180274334 |
Kind Code |
A1 |
Fuchs; Jason ; et
al. |
September 27, 2018 |
SYSTEM AND METHOD FOR MONITORING DISPOSAL OF WASTEWATER IN ONE OR
MORE DISPOSAL WELLS
Abstract
A system and method for monitoring disposal of wastewater in a
disposal well includes: an event monitor sensor configured to
identify a wastewater disposal event; and a second sensor
configured to collect data about one or more characteristics of the
wastewater during the wastewater disposal event. The data from the
second sensor at the disposal well is analyzed to determine a
classification of the wastewater, which is then reported to an
operator or another interested party.
Inventors: |
Fuchs; Jason; (Louisville,
KY) ; Webb; Forrest; (Evanston, IL) ;
Alphenaar; Deirdre; (Prospect, KY) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Genscape Intangible Holding, Inc. |
Louisville |
KY |
US |
|
|
Family ID: |
63582215 |
Appl. No.: |
15/927453 |
Filed: |
March 21, 2018 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
62477088 |
Mar 27, 2017 |
|
|
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
C02F 2209/06 20130101;
G06K 9/00771 20130101; C02F 2209/40 20130101; C02F 2209/05
20130101; C02F 2209/008 20130101; E21B 47/003 20200501; C02F 1/006
20130101; G06K 9/00711 20130101; C02F 1/008 20130101; E21B 41/0057
20130101; C02F 2209/10 20130101; G06K 2009/00738 20130101; E21B
47/10 20130101 |
International
Class: |
E21B 41/00 20060101
E21B041/00; E21B 47/00 20060101 E21B047/00; G06K 9/00 20060101
G06K009/00 |
Claims
1. A system for monitoring disposal of wastewater in a disposal
well, comprising: an event monitor sensor configured to identify a
wastewater disposal event; a second sensor configured to collect
data about one or more characteristics of the wastewater during the
wastewater disposal event; and a water analysis module operable to
receive collected data from the second sensor, and further operable
to determine a classification of the wastewater from the collected
data.
2. The system as recited in claim 1, wherein the event monitor
sensor is a camera.
3. The system as recited in claim 1, wherein the event monitor
sensor is a current sensor.
4. The system as recited in claim 3, wherein the current sensor
monitors the power consumption of one or more pumps that are
associated with the disposal well.
5. The system as recited in claim 1, wherein the second sensor is
selected from the group consisting of: a total suspended solids
(TSS) sensor; a pH sensor; and a conductivity sensor.
6. The system as recited in claim 1, wherein the second sensor
comprises: a total suspended solids (TSS) sensor; a pH sensor; and
a conductivity sensor.
7. The system as recited in claim 1, wherein the classification is
one of the following: produced water; flow-back water; pit water;
or sediment waste.
8. A method for monitoring disposal of wastewater in a disposal
well, comprising the steps: identifying a wastewater disposal event
at the disposal well; monitoring a volume of waste being disposed
at the disposal well; collecting data indicative of one or more
characteristics of the wastewater; determining a classification of
the wastewater based on the collected data; and reporting the
classification of the wastewater to an operator or another
interested party.
9. The method as recited in claim 8, wherein the step of
identifying the wastewater disposal event at the disposal well
includes the positioning of an event monitor sensor configured to
identify the wastewater disposal event.
10. The method as recited in claim 9, wherein the event monitor
sensor is a camera.
11. The method as recited in claim 9, wherein the event monitor
sensor is a current sensor.
12. The method as recited in claim 11, wherein the current sensor
monitors the power consumption of one or more pumps that are
associated with the disposal well.
13. The method as recited in claim 9, wherein the step of
collecting data indicative of one or more characteristics of the
wastewater is accomplished through use of a second sensor.
14. The method as recited in claim 13, wherein the second sensor is
selected from the group consisting of: a total suspended solids
(TSS) sensor; a pH sensor; and a conductivity sensor.
15. The method as recited in claim 13, wherein the second sensor
comprises: a total suspended solids (TSS) sensor; a pH sensor; and
a conductivity sensor.
16. The method as recited in claim 8, wherein the classification is
one of the following: produced water; flow-back water; pit water;
or sediment waste.
17. A method for monitoring disposal of wastewater in a disposal
well, comprising the steps: receiving data from an event monitor
sensor in order to identify a wastewater disposal event at the
disposal well; receiving data from a second sensor at the disposal
well about one or more characteristics of the wastewater during the
wastewater disposal event; analyzing the data from the second
sensor at the disposal well to determine a classification of the
wastewater; and reporting the classification of the wastewater to
an operator or another interested party.
18. The method as recited in claim 17, and further comprising a
step of computing statistics for the wastewater based on the data
from the second sensor during the wastewater disposal event.
19. The method as recited in claim 17, in which the step of
analyzing the data from the second sensor at the disposal well to
determine the classification of the wastewater includes:
establishing a water classification model based on the data
collected from the second sensor during prior wastewater disposal
events as compared to accurate information about the wastewater
during prior wastewater disposal events acquired from an external
source; and applying the water classification model to data from
the second sensor during the wastewater disposal event.
20. The method as recited in claim 19, in which the water
classification model is based on a logistic regression model.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] The present application claims priority to U.S. patent
application Ser. No. 62/477,088 filed on Mar. 27, 2017, which is
incorporated herein by reference.
BACKGROUND OF THE INVENTION
[0002] In oil-producing and gas-producing regions, hydrocarbon
exploration and production companies require water when drilling
wells using hydraulic fracturing. Furthermore, producers need a
location to dispose of both the used fracturing fluid (flow-back
water) and the water that is produced naturally alongside the
hydrocarbons (produced water).
[0003] In many cases, depending on state, regional, or federal
regulations, the flow-back and produced water is deposited into a
dedicated disposal well. Such a disposal well may also be referred
to as a saltwater disposal well (SWD) or a wastewater disposal
well. Disposal wells are often operated in remote areas, often
unmanned by any staff or management to oversee disposal events.
Subsequently, operators often lack insight into key details of the
daily operations of a disposal well. These details include, among
others, site security, recordkeeping, billing, and scheduling of
preventative maintenance.
[0004] In many instances, wastewater disposal events are
self-reported by the driver of a disposal truck. The operator has
little recourse to verify that information regarding a disposed
quantity of water, such as the type and volume of water, is
accurate or even correct. An operator may charge vastly different
rates for disposal of different fluid types and is incentivized to
ensure there is as little error as possible. Furthermore, paper
records are often the only records of disposal events, which may be
difficult to audit to verify that events were reported
accurately.
[0005] Additionally, operators may be actively engaged in buying or
selling services in a water disposal marketplace where operational
data is valuable. For instance, an operator may advertise its
current prices for its disposal services in an effort to attract
truck drivers who are hauling waste to use those services. Factors
that influence the price might include, but are not limited to,
disposal capacity, traffic volumes, well pressures, tank levels, or
volumes of different types of wastewater over time. If the operator
can automatically measure and communicate these types of data to
the wider market, it can realize certain operational
efficiencies.
SUMMARY OF THE INVENTION
[0006] The present invention is a system and method for monitoring
disposal of wastewater in one or more disposal wells.
[0007] In the system and method of the present invention, disposal
wells are outfitted with sensors to determine information related
to wastewater that is disposed in the well, and that information is
then delivered to the operator of the disposal well. Specifically,
in the system method of the present invention, disposal of
wastewater in disposal wells is monitored by analyzing wastewater
that is disposed in a monitored well. Furthermore, a rules-based
classification system is employed to automate detection and
classification of future disposal events. Furthermore, the system
and method of the present invention allows the wastewater disposal
information to be leveraged to predict and characterize energy
commodity extraction in a region.
[0008] In an exemplary system made in accordance with the present
invention, a well facility, which may include one or more disposal
wells, includes an event monitor sensor associated with each of the
one or more wells. The event monitor sensor comprises one or more
sensors to identify the presence of a volume of wastewater to be
disposed and/or of a wastewater disposal event. In some
embodiments, the event monitor sensor is a camera or similar
imaging device that collects images to determine the start and/or
completion of a wastewater disposal event. In some embodiments,
rather than use a camera or similar imaging device, the event
monitor sensor is a laser beam and photo-eye combination that is
tripped or broken when a truck passes through the path of the beam,
thus identifying the start and/or completion of a wastewater
disposal event. In some embodiments, the event monitor sensor is a
coil of wire embedded in the road, a pneumatic tube, or a vibration
sensor that can detect when a truck passes over it, each of which
can identify the start and/or completion of a wastewater disposal
event. Finally, in some embodiments, a pump associated with a well
is monitored by the event monitor sensor; for example, the event
monitor sensor may be a current sensor that is used to monitor the
power consumption of one or more pumps that are associated with the
well.
[0009] Irrespective of the type of sensor employed, the event
monitor sensor collects data, and the collected data is then
transmitted to the central processing facility, where the collected
data is stored in a database for subsequent use or analysis.
[0010] In an exemplary system made in accordance with the present
invention, the well facility further includes a second sensor (or
sensors) associated with the well. The second sensor measures one
or more characteristics of the wastewater that is being disposed in
the well. For example, the second sensor may be one or more of: a
total suspended solids (TSS) sensor; a sensor; and a conductivity
sensor. Irrespective of the type of sensor employed, the collected
data is then also transmitted to the central processing facility
and stored in a database.
[0011] The collected data is then analyzed using a water analysis
module, which makes use of a digital computer program, i.e.,
computer-readable instructions stored and executed by a computer,
to carry out the analysis. In one exemplary implementation, the
analysis carried out by the water analysis module commences with
the collection and cleaning of the data, which may be accomplished,
for example, by applying transforms to create uniform date/time
formats and/or removing duplicate rows. Statistics are then
computed for the disposed wastewater during the wastewater disposal
event. Finally, the collected (and cleaned) data is then analyzed
to determine a classification for the wastewater, for example, by
applying a water classification model.
[0012] The water classification model is a function hat maps an
input variable (i.e., collected data from the second sensor) to one
or more discrete classes (i.e., water classifications). In this
particular case, the objective is to distinguish between four
different classes of wastewater: (i) produced water; (ii) flow-back
water; (iii) pit water: and (iv) basic sediment and waste (BSW).
Common models that may be used, for example, are decision trees,
nearest neighbor classifiers, logistic regression models, and
support vector machines. In each case, the water classification
model is built and established by using a training set of water
information from an external source (or "truth data") and then
correlating that water information to collected (and cleaned) data
from the second sensor. No matter which type of model is used, the
objective is to create a model that accurately predicts the values
of the unknown or future values. For example, since collected data
from the second sensor may be from total suspended solids (TSS)
sensor, a pH sensor, conductivity sensor (or other sensor), data
about total suspended solids, pH, and/or conductivity may all be
inputs into a water classification model that delivers as its
output a classification of wastewater: (i) produced water; (ii)
flow-back water; (iii) pit water; or (iv) basic sediment and waste
(BSW). Once built and established, the water classification model
is applied to subsequently collected (and cleaned) data to
determine a classification for the wastewater during a particular
wastewater disposal event.
[0013] Finally, the classification, along with statistics for the
disposed wastewater during the wastewater disposal event, is
communicated to an operator or other interested parties. It is
contemplated and preferred that such communication to the operator
or other interested parties could be achieved through electronic
mail delivery and/or through export of the data to an
access-controlled Internet web site, which the operator or other
interested parties can access through a common Internet browser
program.
[0014] In addition to being utilized by an operator of a well
facility, the system and method of the present invention may be
further leveraged to predict and analyze energy commodity and/or
water consumption in a region.
DESCRIPTION OF THE DRAWINGS
[0015] FIG. 1 is a schematic representation of an exemplary system
made in accordance with the present invention;
[0016] FIG. 2 is a flow chart illustrating the steps of an analysis
carried out by the water analysis module in an exemplary
implementation of the present invention; and
[0017] FIG. 3 is a plot of a logistics function.
DETAILED DESCRIPTION OF THE INVENTION
[0018] The present invention is a system and method for monitoring
disposal of wastewater in one or more disposal wells.
[0019] In the system and method of the present invention, disposal
wells are outfitted with sensors to determine information related
to wastewater that is disposed in the well, and that information is
then delivered to the operator of the disposal well. Specifically,
in the system and method of the present invention, disposal of
wastewater in disposal wells is monitored by analyzing wastewater
that is disposed in a monitored well. Furthermore, a rifles-based
classification system is employed to automate detection and
classification of future disposal events. Furthermore, the system
and method of the present invention allows the wastewater disposal
information to be leveraged to predict and characterize energy
commodity extraction in a region.
[0020] FIG. 1 is a schematic representation of an exemplary system
made in accordance with the present invention. A well facility 10
is illustrated as including a single well 15 for simplicity, but a
well facility 10 may include multiple wells 15 within the facility.
The well facility 10 further includes an event monitor sensor 20
associated with the well 15. The event monitor sensor 20 comprises
one or more sensors to identify the presence of a volume of
wastewater to be disposed and/or of a wastewater disposal
event.
[0021] in some embodiments, the event monitor sensor 20 is a camera
or similar imaging device that collects images to determine the
start and/or completion of a wastewater disposal event. For
example, the wastewater disposal event may be defined as a disposal
truck entering and subsequently exiting the well facility 10. Truck
arrival and departure times may be determined by analyzing images
of the well facility 10 that are captured by the camera or similar
imaging device, which, of course, would be mounted or otherwise
positioned so that it has a sufficient view of the well facility
10, particularly the bays in which trucks unload wastewater. For
example, one camera that may be used for such image capture is the
Axis Q1775 Network Camera manufactured by Axis Communications AB of
Lund, Sweden.
[0022] The event monitor sensor 20 may collect images at a set time
interval, for instance, once every three minutes. Furthermore, as
shown in FIG. 1, the collected images are then preferably
transmitted to a central processing facility 60, where the
collected images are stored in a database 22 for subsequent use or
analysis.
[0023] Event information can then be extracted from the collected
images using one of several techniques. For example, in some
embodiments, a human could curate collected images to identify the
arrival and departure times of trucks. Alternatively, the collected
images could be analyzed via crowdsourcing (using platforms such as
Amazon Mechanical Turk or CrowdFlower) to allow for faster human
processing of the images.
[0024] In other embodiments, the collected images are analyzed
utilizing an image analysis module 24 at the central processing
facility 60, which makes use of a digital computer program, i.e.,
computer-readable instructions stored and executed by a computer,
to carry out the analysis. For example, the collected images may be
organized in chronological order, and the image analysis module 24
could then be used to detect changes in the images, such as the
arrival and/or departure of a truck. The image analysis module 24
may then also detect the start and/or completion of a wastewater
disposal event by identifying the first image (and time of the
image) that a truck is visible in an image, and subsequently
identify the last image (and time of the image) that the same truck
is visible. From such an image analysis, a wastewater disposal
event is identified.
[0025] As a further refinement, in some embodiments, collected
images may be further utilized to identify and track additional
information related to the customers of the well facility 10. For
instance, particular trucks, truck drivers, and trucking companies
may be identified based on analysis of the collected images from
the event monitor sensor 20 or other imaging device. For example,
in some embodiments, identifying marks from each truck may be
captured. This can be done by employing optical character
recognition (OCR) technology to read, for example, the license
plate, the waste hauling permit (WHP) number, the company name, or
other text from the truck itself. One exemplary imaging system is
the AutoVu.TM. automatic license plate recognition (ALPR) system
manufactured by Genetec, Inc. of Montreal, Quebec, Canada. In other
embodiments, visual information may be collected by cameras
positioned outside of the operator's property to capture details on
vehicle identifying marks or vehicular traffic in general. These
cameras are positioned so that the relevant information is
captured. They may be installed, for instance, above or near the
roadway outside the facility. Additionally, the cameras may be
positioned remotely, for instance, mounted to an aerial vehicle or
satellite. In any event, by independently identifying and matching
a wastewater disposal event to a certain truck, truck driver, or
trucking company, record-keeping and billing may be further
automated to ensure the operator of the well facility 10 is
properly compensated for all wastewater disposal events.
[0026] In some embodiments, rather than use a camera or similar
imaging device, the event monitor sensor 20 is a laser beam and
photo-eye combination that is tripped or broken when a truck passes
through the path of the beam, thus identifying the start and/or
completion of a wastewater disposal event.
[0027] In some embodiments, the event monitor sensor 20 is a coil
of wire embedded in the road, a pneumatic tube, or a vibration
sensor that can detect when a truck passes over it, each of which
can identify the start and/or completion of a wastewater disposal
event.
[0028] In some embodiments, a pump associated with a well 15 is
monitored by the event monitor sensor 20. For example, the event
monitor sensor 20 may be a current sensor that is used to monitor
the power consumption of one or more pumps that are associated with
the well 15. Specifically, the event monitor sensor 20 (or current
sensor) may be used to determine when a pump turns on or off, or
how long a pump associated with the well 15 is in operation. Such
monitoring of current flowing to a pump is described in U.S. Patent
Publication No. 2016/0019482, which is entitled "Method and System
for Monitoring a Production Facility for a Renewable Fuel" and is
incorporated herein by reference. As described therein, current
sensors are placed on power cables associated with one or more
pumps; such placement is preferably non-invasive (e.g., around the
power cables) and does not interrupt operation. For example, one
preferred sensor for use in the system and method of the present
invention is a PAN-series current sensor manufactured by Panoramic
Power Ltd of Kfar Saba, Israel, one of which would be placed on a
power cable for each of the pumps of the well 15
[0029] In some embodiments, such a current sensor may be remotely
positioned, for instance, near electric power transmission lines
that are connected to and supplying power to the facility. In this
instance, the current sensors do not come in contact with the wires
through which they are measuring the current. Instead, the sensors
are arranged to remotely measure the magnetic and electric fields
produced by the conductors of the electric power transmission lines
and calculate the power moving through the conductors, as
described, for example, in U.S. Pat. No. 6,771,058 entitled
"Apparatus and Method for the Measurement and Monitoring of
Electrical Power Generation and Transmission" and U.S. Pat. No.
6,714,000 entitled "Apparatus and Method for Monitoring Power and
Current Flow," each of which is incorporated herein by
reference.
[0030] Again, the event monitor sensor 20 may collect data at a set
time interval, for instance, once every three minutes, and the
collected data is then preferably transmitted to the central
processing facility 60, where the collected data is stored in a
database 22 for subsequent use or analysis,
[0031] With respect to collected current data, to convert such
current data to operational status information, the current data is
analyzed using a current analysis module 26, which makes use of a
digital computer program, i.e., computer-readable instructions
stored and executed by a computer, to carry out the analysis. In
the current analysis module 26, the analog data is digitized based
on a given threshold. If the measured current is above the
threshold, the pump is considered to be "ON," whereas, if the
measured current is below the threshold, the pump is considered to
be "OFF." The time at which the pump transitions from one state to
another can then be extracted from the data. From this data stream,
information about when the pump turned on, when it turned off, and
how long it was in operation for a given period can be
generated.
[0032] In order to determine how much fluid has flown through the
pump during operation (i.e., during a wastewater disposal event),
it is necessary to create a model relating pump current to the
fluid flow rate, for a given set of fluid properties. Then, this
current-to-flow-rate mapping can be applied to future events. Such
creation of transforms which takes collected data and transforms
the collected data into operational statuses is also described in
U.S. Patent Publication No. 2016/0019482, which is entitled "Method
and System for Monitoring a Production Facility for a Renewable
Fuel" and is incorporated herein by reference.
[0033] Additionally, pumping events may be determined by making use
of operator-supplied data streams, such as those created by a
supervisory control and data acquisition (SCADA) system. As part of
this SCADA system, a tablet, laptop, personal computer, or other
input device may be used by the truck drivers or pump operators to
input characteristics about a load of wastewater. For instance, the
operator may input the arrival time, departure time, volume of
wastewater disposed of, wastewater classification, license plate
number, waste hauler permit number, or department of transportation
permit number of a truck. This input method is tied to the
operator's backend financial accounting system, where the data can
be stored in a database and retrieved as needed.
[0034] Referring again to FIG. 1, the well facility 10 further
includes a second sensor 30 (or sensors)associated with the well
15. The second sensor 30 measures one or more characteristics of
the wastewater that is being disposed in the well 15. For example,
in some embodiments, the second sensor 30 is a total suspended
solids (TSS) sensor, such as the Model 950 Suspended Solids Monitor
manufactured by Confab Instrumentation of Jackson, Calif. As
wastewater is disposed in the well 15, the TSS sensor monitors
total suspended solids in the wastewater. For another example, in
other embodiments, the second sensor 30 is a pH sensor, such as the
Model DPD1P1 Online Process pH Sensor manufactured by Hach Lange
GmbH of Dusseldorf, Germany. For yet another example, in other
embodiments, the second sensor 30 is a conductivity sensor, such as
the 3700 Series Analog Inductive Conductivity Sensor manufactured
by Hach Lange GmbH of Dusseldorf, Germany.
[0035] Referring again to FIG. 1, irrespective of the type of
sensor employed, the collected data is then preferably transmitted
to the central processing facility 60 and stored in a database 32.
The collected data is then analyzed using a water analysis module
34, which makes use of a digital computer program, i.e.,
computer-readable instructions stored and executed by a computer,
to carry out the analysis. For example, a remote backhaul device,
such as the Wavelet device manufactured by Ayyeka Technologies of
Jerusalem, Israel, may be utilized to sample and collect data from
the second sensor 30 and then communicate the collected data to the
water analysis module 34 at the central processing facility 60 for
analysis.
[0036] In this exemplary implementation, and referring now to FIG.
2, the analysis carried out by the water analysis module 34
commences with the collection and cleaning of the data, as
indicated by block 200. In this regard, data may be collected from
the second sensor 30 (or sensors) at the well 15 on a substantially
continuous basis, for example, at some fixed frequency, such as
every one minute. Alternatively, data collection may be initiated
when there is an indication of an occurrence of a wastewater
disposal event. The indication of the occurrence of the wastewater
disposal event may he based on the data from the event monitor
sensor 20. For example, as described above, the event monitor
sensor 20 may identify that a wastewater disposal event has begun
by image analysis and/or pump current monitoring, at which time the
second sensor 30 starts collecting data. Similarly, the event
monitor sensor 20 may provide a signal to indicate that a
wastewater disposal event has concluded, at which time the second
sensor 30 stops collecting data.
[0037] Referring still to FIG. 2, with respect to the cleaning of
the data, this may be accomplished by, for example, applying
transforms to create uniform date/time formats and/or removing
duplicate rows.
[0038] Referring still to FIG. 2, statistics are then computed for
the disposed wastewater during the wastewater disposal event, as
indicated by block 210. For example, a wastewater disposal event
may be defined as one truck entering the well facility 10,
unloading its wastewater, and then exiting the well facility 10.
Relevant data would be collected from the second sensor 30 (or
sensors) only when the pump utilized by that truck to unload its
wastewater is running. In other words, with respect to the
exemplary sensor types disclosed above, the total suspended solids,
pH, and/or conductivity measurements are all irrelevant when there
is no wastewater being disposed. Thus, the analysis focuses on the
time period of the wastewater disposal event. For example, as
discussed above, the time at which a pump turned on (pump start
time) and the time at which a pump turned off (pump stop time) can
both be determined by the event monitor sensor 20. Therefore, the
water analysis module 34 only needs to analyze data from the second
sensor 30 (or sensors) during the wastewater disposal event.
Statistics such as mean, standard deviation, maximum, and minimum
can then be found easily for each of the water quality metrics
within the time window defined by the wastewater disposal
event.
[0039] Referring still to FIG. 2, the collected (and cleaned) data
is then analyzed to determine a classification for the wastewater.
Such classification may be determined, for example, by applying a
water classification model, as indicated by block 220 in FIG.
2.
[0040] Referring again to FIG. 1, to establish a water
classification model, a water classification module 40 receives
water information from an external source 50. This water
information is "truth data" that is accurate information about the
wastewater that was pumped into the well 15 during a given
wastewater disposal event; for example, for initial training
establishment of the water classification model, such water
information may come from or be derived from operator records. In
some implementations, water information is split into a training
set and a test set instance, in some implementations, approximately
80% of the water information data is classified as the training
set, while 20% of the water information is classified as the test
set, although this can vary on a case-by-case basis. This allows
for the water classification model to be built using the training
set, and then subsequently applied to the test set to assess its
accuracy, as further described below.
[0041] In its simplest form, the water classification model is a
function that maps an input variable (i.e., collected data from the
second sensor 30) to one or more discrete classes (i.e., water
classifications). In this particular case, the objective is to
distinguish between four different classes of wastewater: (i)
produced water; (ii) flow-back water; (iii) pit water; and (iv)
basic sediment and waste (BSW). Common models that may be used, for
example, are decision trees, nearest neighbor classifiers, logistic
regression models, and support vector machines. In each case, the
water classification model is built and established by using the
training set of water information from an external source 50 (or
"truth data") and then correlating that water information to
collected (and cleaned) data from the second sensor 30. No matter
which type of model is used, the objective is to create a model
that accurately predicts the values of the unknown or future
values. For example, since collected data from the second sensor 30
may be from a total suspended solids (TSS) sensor, a pH sensor,
conductivity sensor (or other sensor), data about total suspended
solids, pH, and/or conductivity may all be inputs into a water
classification model that delivers as its output a classification
of wastewater: (i) produced water; (ii) flow-back water; (iii) pit
water; or (iv) basic sediment and waste (BSW).
[0042] For example, one specific method for classifying wastewater
is based on the use of a logistic regression model. A logistic
regression is a type of model that tries to predict the value of a
discrete binary variable, Y, given one or more independent
variables, X. It can answer questions such as: "Did a student pass
or fail this test?" or "Was this subject healthy or sick?"
Moreover, a logistic regression model can provide a probability
that a certain example fits into one class or the other. For
instance, the logistic regression model allows for statements such
as "there is a 51% chance that it will rain today," or "there is a
99% chance that it will snow tomorrow," rather than simply stating
that "it will rain" or "it will snow."
[0043] We can describe this model in a more formal way. The
outcome, or dependent variable, is y. We know y can take only one
of two values. In the case of the student passing a test, the value
is either "pass" or "fail." So, we can denote y taking on only
these two values by:
y.di-elect cons.{0, 1} (1)
[0044] Now, we want some function that can maximize the likelihood
or probability of y=1 when y really is 1, and y=0 when the opposite
is true. One function that achieves this is the logistic
function:
h .theta. ( x ) = g ( .theta. T x ) = 1 1 + e - .theta. T x ( 2 )
##EQU00001##
where .theta. represents a vector of weight parameters that are
applied to x. More simply, with z=.theta..sup.Tx, the equation can
be rewritten as:
g ( z ) = 1 1 + e - z ( 3 ) ##EQU00002##
[0045] As shown in FIG. 3, when plotted, equation (3) results in an
S-shaped curve, in which g(z) tends toward 1 as z.fwdarw..infin.
and tends toward 0 as z.fwdarw.-.infin..
[0046] Therefore, the objective is to adjust the parameter .theta.
so that when y=1, z=.theta..sup.Tx is high, and when y=0,
z=.theta..sup.Tx is low:
P(y=1|x; .theta.)=h.sub.g(x) (4)
P(y=0|x; .theta.)=1-h.sub.g(x) (5)
Or, more concisely:
P(y|x; .theta.)=(h.sub.g(x)).sup.y(1-h.sub.g(x)).sup.1-y (6)
[0047] Furthermore, the logistic regression model can be expanded
to provide probabilities in the case of more than two classes. In
this particular case, y is not binary and can assume more than two
states. The logistic regression model is simply run for each
possible state or class.
[0048] Again, in this particular case, the objective is to
distinguish between four different classes of wastewater: (i)
produced water; (ii) flow-back water; (iii) pit water; and (iv)
basic sediment and waste (BSW). The logistic regression model is
thus applied to predict whether a sample is produced water, or not;
whether it is flow-back water, or not; and so on. The results are
then aggregated.
[0049] For example, the output from the application of one model on
a small data set is presented in Table A below:
TABLE-US-00001 TABLE A Sam- True Predicted ple Fluid Predicted
Predicted Pit Predicted Cor- No. Type BSW Flowback Water Saltwater
rect? 1 Pit Water 29% 0% 12% 60% No 2 BSW 6% 0% 43% 51% No 3 Pit
Water 0% 0% 98% 2% Yes 4 Pit Water 0% 0% 98% 2% Yes 5 BSW 75% 0% 0%
25% Yes 6 Flowback 0% 100% 0% 0% Yes 7 BSW 56% 0% 9% 36% Yes 8 BSW
62% 0% 31% 7% Yes 9 Pit Water 0% 0% 98% 2% Yes 10 Pit Water 30% 0%
63% 7% Yes 11 Flowback 0% 50% 50% 0% No 12 Pit Water 0% 50% 50% 0%
No 13 Saltwater 39% 0% 21% 40% Yes 14 Saltwater 13% 0% 0% 87% Yes
15 Saltwater 6% 0% 0% 94% Yes 16 Saltwater 43% 0% 3% 54% Yes 17
Saltwater 21% 0% 14% 64% Yes 18 Saltwater 19% 0% 10% 70% Yes
[0050] Each row in Table A denotes one sample of wastewater. The
column labeled "True Fluid Type" is the actual classification of
that sample as given by the operator. Each of the next four columns
is a probability that the sample is in the identified one of the
four classes. For example, Sample No. 3 was pit water. Based on the
pH, conductivity, and total suspended solids content of that
sample, the model predicted that it had a 98% chance of being pit
water, Thus, if a future sample had the same properties as Sample
No. 3, the model would predict that the sample was pit water with
98% certainty
[0051] Again, the water classification model is built and
established by using the water information from an external source
(or "truth data") and then correlating that water information to
the collected (and cleaned) data from the second sensor 30. Once
initially established, the model may be applied once to the test
set of water information in order to assess its accuracy. The water
classification model is then stored in a memory component that is
part of or associated with the water classification module 40.
[0052] Referring again to FIG. 2, as mentioned above, once built
and established, the water classification model is applied to
subsequently collected (and cleaned) data to determine a
classification for the wastewater, as indicated by block 220.
[0053] Finally, the classification, along with statistics for the
disposed wastewater during the wastewater disposal event, is
communicated to an operator or other interested parties, as
indicated by block 230. It is contemplated and preferred that such
communication to the operator or other interested parties could be
achieved through electronic mail delivery and/or through export of
the data to an access-controlled Internet web site, which the
operator or other interested parties can access through a common
Internet browser program. For example, an operator of the well
facility 10 may be provided with a wastewater classification for a
volume of wastewater. The operator may then verify with a driver
and/or customer who dumped the wastewater determine that all
records are accurate, that the customer was charged for the correct
type of wastewater disposal, and/or to otherwise ensure that the
customer is representing the contents of the wastewater truthfully.
The information may also be delivered to a back-end accounting
system that allows the operator to, for instance, automatically
send invoices, pay expenses, and comply with state, regional,
and/or federal recordkeeping requirements.
[0054] In addition to being used internally by the operator of a
disposal well, information derived from the system may be
automatically communicated to interested water marketplace
participants. For instance, the disposal well operator may choose
to communicate information about current well status, current
prices for various types of wastewater, current pump utilization
rates, current disposal capacity, current traffic volumes,current
well pressures, current tank levels, and the like.
[0055] In addition to being utilized by an operator of a well
facility 10, the system and method of the present invention may be
further leveraged to predict and analyze energy commodity and/or
water consumption in a region. For example, the classification and
other statistics for the disposed wastewater during the wastewater
disposal event may be provided to an aggregate system that collects
wastewater classifications from multiple wells and/or from multiple
facilities. By identifying the amounts and types of wastewater that
is being disposed of in a region, the aggregate system may infer
and/or predict water needs for the region. For another example, the
aggregate system may be able to infer hydrocarbon extraction site
characterizations, such as age of wells or mines, by possessing
knowledge of the composition of wastewater in a region. For yet
another example, the aggregate system may infer information related
to volume of hydrocarbon extraction in a region based on the
accurate sensor information from a plurality of wastewater wells in
a region. For still yet another example, the aggregate system may
be able to forecast hydrocarbon production by possessing knowledge
of water-to-oil or water-to-gas ratios in a region, as described in
U.S. Patent Publication No. 2016/0063402 entitled "Oilfield Water
Management," which is incorporated herein by reference.
[0056] One of ordinary skill in the art will recognize that
additional embodiments and implementations are also possible
without departing from the teachings of the present invention. This
detailed description, and particularly the specific details of the
exemplary embodiments and implementations disclosed therein, is
given primarily for clarity of understanding, and no unnecessary
limitations are to be understood therefrom, for modifications will
become obvious to those skilled in the art upon reading this
disclosure and may be made without departing from the spirit or
scope of the invention.
* * * * *