Patent Application
20230392498
The present disclosure relates generally to wellsite operations and, more particularly (although not necessarily exclusively), to emissions estimations at hydrocarbon operation locations.
Gas emissions can be strictly regulated by government entities. Governments on both the national and local level can expect companies to report gas emissions. There can be an expectation that these gas emissions are reported accurately and in a timely manner. Inaccurate or untimely emissions reports can lead to fines.
The Environmental Protection Agency (EPA) requires reports of greenhouse gas (GHG) data and other relevant information from large GHG emissions sites and fuel and industrial gas suppliers. This data can be used by businesses and others to track and compare facilities' GHG emissions, identify opportunities to cut emissions, minimize wasted energy, and save money. Local governments can use GHG data to find high-emitting facilities in their area, compare emissions between similar facilities, and develop emissions policies.
Certain aspects and examples of the present disclosure relate to using and collecting a set of data to train a machine-learning model to predict emissions estimations at a hydrocarbon operation location. In an example, an emission factor for at least one equipment component of a hydrocarbon operation location may be determined to generate the emissions estimations. Machine-learning can be a form of artificial intelligence that makes predictions from data. The emissions factor can be a representative value that attempts to relate the quantity of an emission released by an equipment component associated with a release of the emission. For example, historical leak data of emissions for a type of tank used at an oil production site can be used to train a particular machine-learning model. In the example, the particular machine-learning model can be used to predict the emissions factor for another tank of the same or similar type at a different oil production site.
Data used in an example solution described in the present disclosure can include high resolution data, real-time emission data and highly granular equipment level data. These forms of data can be used to train the machine-learning model to predict emission factors with reduced uncertainty compared to other methods. A trained machine-learning model can provide insights into factors that contribute to emissions. The trained machine-learning model can help asset managers make fast, smart decisions related to the emissions and achieve yearly emissions commitments.
An emission factor prediction can be accomplished in two stages: a data collection/integration stage; and a model building stage. Data can be collected from diverse sources starting with remote data sensors such as Light Detection and Ranging (LiDAR). LiDAR can be a remote sensing method that uses light in the form of a pulsed laser to examine the surface of the Earth. A LiDAR system can employ an emitted laser signal that reflects off an obstacle. A reflected laser signal can return to the LiDAR system. The LiDAR system can detect and characterize the obstacle based on the reflected laser signal. In some examples of the present disclosure, the obstacle can be a gas molecule emitted from the at least one equipment component of a hydrocarbon operation location. An operator's real-time equipment operational and infrastructure data can be another source of equipment level data. Weather characteristics can be extracted from local weather reports. Factors such as historical leak data can be extracted from Leak Detection and Repair (LDAR) reports.
These forms of data can be gathered and analyzed to prepare a master dataset. The master dataset can be used to train a machine-learning model. The trained machine-learning model can be applied to predict an emission factor for at least one equipment component. The emission factor can be used to calculate a total emission over a predetermined amount of time for the at least one equipment component.
Illustrative examples are given to introduce the reader to the general subject matter discussed herein and are not intended to limit the scope of the disclosed concepts. The following sections describe various additional features and examples with reference to the drawings in which like numerals indicate like elements, and directional descriptions are used to describe the illustrative aspects, but, like the illustrative aspects, should not be used to limit the present disclosure.
The BHA 104 may include a drill bit 114 operatively coupled to a tool string 116, which may be moved axially within a drilled wellbore 118 and can be attached to the drill-string 106. The tool string 116 may include one or more well bore tools 109 for determining conditions in the wellbore 118 or for performing other suitable operations. During operation, the drill bit 114 can penetrate the subterranean formation 102 to create the wellbore 118. The BHA 104 can control the drill bit 114 as the drill bit 114 advances into the subterranean formation 102. Fluid or “mud” from a mud tank 120 may be pumped downhole using a mudpump 122 that can be powered by an adjacent power source, such as a prime mover or motor 124. The mud may be pumped from the mud tank 120, through a stand pipe 126, which can feed the mud into the drill-string 106 and can convey the mud to the drill bit 114. The mud can exit one or more nozzles (not shown) arranged in the drill bit 114 and can thereby cool the drill bit 114. After exiting the drill bit 114, the mud can circulate back to the surface 110 via an annulus defined between the wellbore 118 and the drill-string 106. Cuttings and mud mixture that can be passed through a flow line 128 can be processed such that a cleaned mud is returned down hole through the stand pipe 126.
In some examples, the well system 100 can include at least one remote data sensor 150. For example, the at least one remote data sensor 150 can be a drone with LiDAR capabilities. The at least one remote data sensor 150 can collect emissions data from emissions sources in a well system 100. As illustrated with respect to
In some examples, the well system 100 can include a computing device 140 that can be positioned belowground, aboveground, onsite, in a vehicle, offsite, etc. As illustrated with respect to
The computing device 140 can include a communication device 144. The communication device 144 can represent one or more of any components that facilitate a network connection. In the example illustrated in
In some examples, remote emissions data can be collected by one or more third party data sources 205. Remote emissions data can include data recorded by drones equipped with LiDAR technology. The one or more third party data sources 205 may also use other techniques to collect remote emissions data. For example, the other techniques can include satellite imaging, LiDAR sensors on aircraft, or any other techniques used by a third party to collect emissions data. The illustration shows three third party data sources 205. In some examples, there can be more or fewer than three third party data sources 205. In some examples, the third party data sources are not used, and the remote emissions data can be collected by first party data sources. For example, a drones equipped with LiDAR can be associated with the asset 203 and the remote emissions data can be collected by users at the asset 203.
In some examples, data from equipment emission/leak and local weather reports 207 can be collected. The data from local weather reports can include wind speed, wind direction, wind temperature, atmospheric pressure, temperature, ambient temperature, etc. The equipment emission or leak data can include data from Fugitive Emission LDAR inspection reports. Fugitive Emission LDAR reports can identify unintended emissions from equipment. The data from Fugitive Emission LDAR reports can include equipment component inspection frequency and equipment component health data. All of the types of data (e.g., equipment operational and infrastructure data 209, remote emissions data, and data from equipment emission/leak and local weather reports 207) can be integrated to form equipment emission data. In the illustration, equipment methane emission data 210 is shown as an example of equipment emission data. In other examples, the Equipment emission data can be associated with other types of gas emissions.
Graphic 204 illustrates a machine-learning training process. In some examples, the equipment methane emission data 210 from multiple sources may be consolidated into a central data repository 211. For example, the multiple sources can refer to multiple facilities or multiple assets. The data stored in the central data repository 211 can be further preprocessed to generate master data 212, which can be in a required format for machine-learning modeling. Examples of methods associated with preprocessing data can include attribute selection, data reduction, and data cleaning. The master data 212 can be used to train a machine-learning model, as denoted by the modeling 213 in the illustration. The machine-learning model can be optimized by undergoing several iterations of training. The iteration stage of the process can be denoted by the satisfactory decision block 214 in the illustration. If a certain criteria is not met, model optimization 215 occurs and the modeling 213 iteration continues. Once the certain criteria is met, the iterations can stop and the machine learning model can be finalized at block 216. In some examples of the present disclosure, the machine-learning model can include a deep convolutional neural network (DCNN).
Graphic 206 illustrates an application of the machine-learning model. The machine-learning model can be applied to at least one equipment component to determine an emission factor estimation 218 for at least one equipment component. Equipment 217 can be associated with a second hydrocarbon operation location. As an example, the equipment 217 can be tanks at a second hydrocarbon operation location, and the machine-learning model can be applied to predict a methane emission factor estimation 218 for the tanks at the second hydrocarbon operation location.
Graphic 208 depicts a prediction of total methane emission 220 for at least one equipment component over a predetermined amount of time. A determination can be made by multiplying the methane emission factor estimation 218, a total number of equipment components associated with the methane emission factor estimation, and the predetermined amount of time. The predetermined amount of time is referred to in the graphic as total operational time 219. For example, the step 208 can be applied to predict the total methane emission for tanks over the course of a year at a hydrocarbon operation location. The total methane emission in this example is found by multiplying the methane emission factor estimation 218 for tanks from graphic 206 with the number of tanks at the location and with the total time of one year.
In some examples, an individual hydrocarbon operation location may include multiple types of equipment with multiple different methane emission factors. The values of total methane emission for each piece of equipment at the hydrocarbon operation location may be added together to generate the total methane emission value 220 for the hydrocarbon operation location. Similarly, graphic 208 can be applied to predict total emissions for other types of gases as well.
At block 302, the process 300 involves collecting a first set of equipment data from a first hydrocarbon operation location. The first set of equipment data can include operational and infrastructure data for a plurality of equipment components at a first hydrocarbon operation location. For example, the plurality of equipment components at the first hydrocarbon operation location can include a tank. The operational data of the tank can include pressure, volume, temperature, etc. The infrastructure data of the tank can include pressure rating, vacuum pressure rating, material, capacity, size, diameter, quantity, height, other operation conditions of the tank, etc. Equipment data from other types of equipment located at the first hydrocarbon operation location may also be collected.
At block 304, the process 300 involves collecting emissions data from the first hydrocarbon operation location. In some examples, a first set of emissions data can be collected using aerial techniques. Aerial techniques for collecting emissions data can involve at least one remote data sensor 150 fitted with at least one LiDAR sensor. Examples of remote data sensors can include drones, aircrafts, satellites, etc. Types of remote emissions data can include emissions rates, emissions plume heights, maximum emissions concentrations, emissions persistence, locations of at least one tank, etc. In some examples, the remote emissions data can be collected by third party contractors.
In some examples, another set of emissions data can include historical leak data for the plurality of equipment components at the first hydrocarbon operation location. The historical leak data can be collected from fugitive emission LDAR Inspection reports. For example, the plurality of equipment components at the first hydrocarbon operation can include a tank. The historical leak data for the tank can include inspection frequency and tank component health (e.g., strength of thief hatch, valve, etc.).
In some examples, emissions data can also include weather characteristics collected from weather reports for the first operation location. The weather characteristics can include wind speed, wind direction, wind chill, atmospheric pressure, weather temperature, ambient temperature, etc.
At block 306, the process 300 involves preprocessing the first set of equipment data and the emissions data. The first set of equipment data and the emissions data can be integrated and a central data repository can be prepared. For example, the first set of equipment data and the emissions data can be preprocessed by standardizing the units of the first set of equipment data and the emissions data. Master data generated from pre-processing the equipment and emissions data can be generated which can be in a format usable for machine-learning modeling.
At block 308, the process 300 involves training a machine-learning model to estimate an emission factor of the plurality of equipment components at the first hydrocarbon operation location. In some examples, the emission factor of the plurality of equipment components at the first hydrocarbon operation location can be a methane emissions factor. The master data can be used for training purposes. In some examples of the present disclosure, a supervised machine-learning model is used. In some examples of the present disclosure, the machine learning model includes a DCNN. The model can undergo iterations of training for optimization. In some examples of the present disclosure, once the estimate for the emissions factor of the at least one equipment component at the first hydrocarbon operation location meets a training criteria, training of the machine-learning model is completed and the machine-learning model is finalized. A separate machine-learning model can be trained for each type of equipment that produces emissions at the first hydrocarbon operation location or across several hydrocarbon operation locations.
At block 310, the process 300 involves collecting a second set of equipment data from a second hydrocarbon operation location. The second set of equipment data can include operational and infrastructure data for at least one equipment component at the second hydrocarbon operation location. In some examples, the equipment at the second hydrocarbon operation location may be similar to the equipment at the first hydrocarbon location or the equipment across several hydrocarbon operation locations.
At block 312, the process 300 involves applying a trained machine-learning model to the second set of equipment data from the second hydrocarbon operation location to estimate an emission factor of the at least one equipment component at the second hydrocarbon operation location. In some examples, the emissions factor of the at least one equipment component at the second hydrocarbon operation location is a methane emission factor.
At block 314, the process 300 involves generating a total emissions for a predetermined amount of time of the at least one equipment component at the second hydrocarbon operation location. For example, the at least one equipment component at the second hydrocarbon operation location can be a tank. The machine-learning model can be applied to estimate the emission factor of the tank. Then the total emissions of the tank for a predetermined amount of time can be generated by multiplying the emissions factor of the tank, the number of tanks, and the predetermined amount of time. Other sources of emissions can be present at the second hydrocarbon operation (e.g., motors, gas dehydrators, wellbores, etc.). A total emissions for the predetermined amount of time at the second hydrocarbon operation can be determined by adding together the total emissions of all of the sources.
As shown, the computing device 402 includes the processor 404 communicatively coupled to the memory 406 by the bus 408. The processor 404 can include one processor or multiple processors. Non-limiting examples of the processor 404 include a Field-Programmable Gate Array (FPGA), an application specific integrated circuit (ASIC), a microprocessor, or any combination of these. The processor 404 can execute instructions 410 stored in the memory 406 to perform operations. In some examples, the instructions 410 can include processor-specific instructions generated by a compiler or an interpreter from code written in any suitable computer-programming language, such as C, C++, C #, or Java.
The memory 406 can include one memory device or multiple memory devices. The memory 406 can be non-volatile and may include any type of memory device that retains stored information when powered off. Non-limiting examples of the memory 406 include electrically erasable and programmable read-only memory (EEPROM), flash memory, or any other type of non-volatile memory. At least some of the memory 406 can include a non-transitory computer-readable medium from which the processor 404 can read instructions 410. The non-transitory computer-readable medium can include electronic, optical, magnetic, or other storage devices capable of providing the processor 404 with the instructions 410 or other program code. Non-limiting examples of the non-transitory computer-readable medium include magnetic disk(s), memory chip(s), RAM, an ASIC, or any other medium from which a computer processor can read instructions 410.
Additionally, the memory 406 can further include equipment data 412, remote emissions data 418, historical leak data 420, and local weather data 422. The equipment data 412 can include operational data 414 and infrastructure data 416. The processor 404 can preprocess and integrate the equipment data 412, remote emissions data 418, historical leak data 420, and local weather data 422. The processor 404 can generate master data 424, which can be in a required format for a machine-learning model.
The memory 406 can further include an emission factor 426, a total operational time 428, and total emissions 430. The processor 404 can apply the machine-learning model to estimate the emission factor 426 for at least one equipment component. The processor 404 can determine the total emissions 430 of the at least one equipment component over a predetermined amount of time by multiplying the emission factor 426 and the total operational time 428.
In some examples, the computing device 402 can implement the process shown in
In some aspects, systems and methods for estimating emissions at a hydrocarbon operation location are provided according to one or more of the following examples:
As used below, any reference to a series of examples is to be understood as a reference to each of those examples disjunctively (e.g., “Examples 1-4” is to be understood as “Examples 1, 2, 3, or 4”).
Example 1 is a method, comprising: collecting a first set of equipment data from a first hydrocarbon operation location; collecting emissions data relating to the first hydrocarbon operation location; training at least one machine-learning model to estimate an emissions factor of at least one equipment component of the first hydrocarbon operation location using the first set of equipment data and the emissions data of the first hydrocarbon operation location; receiving a second set of equipment data from a second hydrocarbon operation location; and applying the at least one machine-learning model to the second set of equipment data to estimate an emissions factor of at least one equipment component of the second hydrocarbon operation location.
Example 2 is the method of example 1, further comprising generating total emissions of the second hydrocarbon operation location for a predetermined operation time using the emissions factor of the at least one equipment component of the second hydrocarbon operation location.
Example 3 is the method of examples 1-2, wherein the emissions data comprises emissions rates, emissions plume heights, maximum emissions concentrations, emissions persistence, and a location of the at least one equipment component of the first hydrocarbon operation location.
Example 4 is the method of examples 1-3, further comprising preprocessing the first set of equipment data and the emissions data by standardizing units of the first set of equipment data and the emissions data.
Example 5 is the method of examples 1-4, wherein training the at least one machine-learning model comprises training a plurality of machine-learning models, wherein each machine-learning model of the plurality of machine-learning models is associated with an individual equipment component of the first hydrocarbon operation location.
Example 6 is the method of examples 1-5, wherein the emissions factor comprises a methane emissions factor.
Example 7 is the method of examples 1-6, wherein the at least one machine-learning model comprises a deep convolutional neural network (DCNN).
Example 8 is the method of examples 1-7, wherein the emissions data comprises historical leak data extracted from leak detection and repair (LDAR) reports for the at least one equipment component of the first hydrocarbon operation location.
Example 9 is the method of examples 1-8, wherein the emissions data comprises weather characteristics extracted from local weather reports for the first hydrocarbon operation location.
Example 10 is a system comprising: a processor; and a memory that includes instructions executable by the processor for causing the processor to: collect a first set of equipment data from a first hydrocarbon operation location; collect emissions data relating to the first hydrocarbon operation location; and train at least one machine-learning model to estimate an emissions factor of at least one equipment component of the first hydrocarbon operation location using the first set of equipment data and emissions data of the first hydrocarbon operation location.
Example 11 is the system of example 10, wherein the emissions data comprises emissions rates, emissions plume heights, maximum emissions concentrations, emissions persistence, and a location of the at least one equipment component of the first hydrocarbon operation location.
Example 12 is the system of examples 10-11, wherein the memory further comprises instructions executable by the processor for causing the processor to: preprocess the first set of equipment data and the emissions data by standardizing units of the first set of equipment data and the emissions data.
Example 13 is the system of examples 10-12, wherein training the at least one machine-learning model comprises training a plurality of machine-learning models, wherein each machine-learning model of the plurality of machine-learning models is associated with an individual equipment component of the first hydrocarbon operation location.
Example 14 is the system of examples 10-13, wherein the emissions factor comprises a methane emissions factor.
Example 15 is the system of examples 10-14, wherein the at least one machine-learning model comprises a DCNN.
Example 16 is a non-transitory computer-readable medium comprising instructions that are executable by a processor for causing the processor to perform operations comprising: receiving a set of equipment data from a hydrocarbon operation location; applying at least one trained machine-learning model to the set of equipment data to estimate an emissions factor of at least one equipment component of the hydrocarbon operation location, the at least one trained machine-learning model trained using an initial set of equipment data and initial emissions data from an initial hydrocarbon operation location; and generating total emissions of the hydrocarbon operation location for a predetermined amount of time using the emissions factor of the at least one equipment component of the hydrocarbon operation location.
Example 17 is the non-transitory computer-readable medium of examples 16, wherein the initial emissions data comprises emission rates, emissions plume heights, maximum emissions concentrations, emissions persistence and a location for at least one equipment component of the initial hydrocarbon operation location.
Example 18 is the non-transitory computer-readable medium of examples 16-17, wherein the emissions factor comprises a methane emissions factor.
Example 19 is the non-transitory computer-readable medium of examples 16-18, wherein the at least one trained machine-learning model comprises a DCNN.
Example 20 is the non-transitory computer-readable medium of examples 16-19, wherein the initial emissions data comprises weather characteristics extracted from local weather reports for the initial hydrocarbon operation location.
The foregoing description of certain examples, including illustrated examples, has been presented only for the purpose of illustration and description and is not intended to be exhaustive or to limit the disclosure to the precise forms disclosed. Numerous modifications, adaptations, and uses thereof will be apparent to those skilled in the art without departing from the scope of the disclosure.
