Detection and quantification of multiple methane leaks

Abstract

Systems and methods are described for detecting multiple methane leaks at a site is provided. The method includes receiving methane concentration data from a network of methane sensors positioned at various locations within the site. The data is analyzed to identify elevated methane levels, and a multi-leak detection algorithm is applied to distinguish between multiple leak sources. Spatial distributions of methane concentrations are generated to designate potential leak regions. The method further includes estimating the locations of the leaks by identifying regions with the highest concentrations, and quantifying the uncertainty of the estimated leak locations using statistical models, such as Markov-Chain Monte Carlo (MCMC) methods. The system assigns a certainty level to each leak source and can display the leak locations and uncertainty values on a graphical interface. This method enhances methane leak detection in real-time and improves accuracy in attributing leak sources within complex industrial sites.

Description

DETECTION AND QUANTIFICATION OF MULTIPLE METHANE LEAKSKashi f RashidCROSS REFERENCE PARAGRAPH

[0001] This application claims the benefit of U.S. Provisional Application No. 63 / 607,773, entitled "METHANE LEAK DETECTION" filed December 08, 2023, the disclosure of which is hereby incorporated herein by reference.BACKGROUND

[0002] The detection and management of methane leaks in oil and gas facilities is a critical challenge, both from an environmental and operational standpoint. Methane, a potent greenhouse gas, poses significant risks to the environment due to its high global warming potential. Even minor leaks over time can contribute substantially to atmospheric methane concentrations. Additionally, undetected or uncontrolled leaks can lead to safety hazards, operational inefficiencies, and regulatory non-compliance, making their timely identification and mitigation essential for facility operators.

[0003] Conventional methods for leak detection typically involve periodic manual inspections or the use of mobile sensors, such as infrared cameras or drones, to scan large areas. While these techniques have proven effective in certain scenarios, they are inherently limited by their infrequency and reliance on temporary data collection points. Such methods can result in missed leaks between inspections or lead to delayed detection, allowing small leaks to grow into larger, more dangerous emissions. Moreover, mobile and manual inspections can be labor- intensive, costly, and often unable to provide continuous real-time monitoring.

[0004] Recent advances in sensor technology have made it feasible to implement fixed- point methane sensors across oil and gas facilities, allowing for continuous monitoring of methane concentrations. These sensors, installed at strategic locations throughout a site, can provide real-time data on methane levels, making it possible to identify leaks as soon as they occur. However, the complexity of large facilities, which may contain hundreds of individual equipment components spread across vast areas, presents a significant challenge when trying to pinpoint the exact source of a leak.

[0005] Traditional approaches to analyzing sensor data often focus on single leak events and lack the capacity to effectively handle multiple concurrent leaks. In a multi-leak scenario, methane emissions from different sources can overlap depending on the wind direction and leak locations, leading to confusion about the location and magnitude of each leak. As a result, advanced methods are needed to disaggregate sensor readings and accurately identify both the number and locations of leaks within a facility.

[0006] As a result, a need exists for a way to detect, identify, and quantify multiple methane leaks.SUMMARY

[0007] Examples described herein include systems and methods for detecting, identifying, and attributing multiple methane leaks within a defined site, such as an oil or gas facility. Methane leaks pose significant environmental and operational risks, and detecting multiple leaks simultaneously is particularly challenging due to overlapping sources and the complexity of large industrial sites. The invention overcomes these challenges by utilizing a network of methane sensors that continuously monitor methane concentrations across the site, feeding real-time data to a computing device for analysis and leak detection.

[0008] The system includes methane sensors strategically placed at different locations within the site to provide comprehensive coverage. These sensors can be of various types, including infrared (“IR”) gas sensors, electrochemical sensors, catalytic bead sensors, and laserbased sensors. The sensors transmit methane concentration data to a central computing device, which analyzes the data using advanced algorithms. The system processes this raw data to distinguish between background methane fluctuations and actual leak events, ensuring that multiple leak sources can be accurately identified.

[0009] The method involves applying a multi-leak detection algorithm that generates spatial distributions of methane concentrations across the site. Generating spatial distributions includes mapping and analyzing methane concentration data across a defined site area to visualize where methane levels are elevated. The methane sensors placed at different locations throughout the site collect concentration readings, which are then processed by a computing device. Using these readings, the system creates a spatial map or grid that represents the distribution of methane concentrations across the site. Spatial distributions of methane concentration allows the system to identify regions where methane levels are significantly elevated, which is indicative of potential leaks. The system uses statistical models to distinguish between multiple concurrent leaks, even in dense areas with potentially overlapping emission sources. These regions are then marked as “incumbent” solution spaces where the likelihood of a leak is highest, allowing for precise identification and localization of leaks.

[0010] Once the potential leak regions are identified, the system further refines the analysis by estimating the exact distribution of the potential leaks within these regions. The computing device gathers possible leak locations and assigns them to regional groups to aid methane concentration data correlation with the spatial layout of equipment components, such aspipelines, valves, and storage tanks. This site segmentation allows the system to focus on areas with potential leak sources, and furthermore, to consider the larger group that each belongs to as part of the uncertainty quantification step. This helps to improve the accuracy of leak source predictions. The estimated leak locations are then displayed on a graphical user interface, providing a clear visual representation for operators.

[0011] Finally, the system quantifies the uncertainty associated with the estimated leak locations. Using probabilistic models like the Markov-Chain Monte Carlo (“MCMC”) method, the system calculates the uncertainty associated with the identified leak sources (in location and rate), and possibly to yield a certainty level for each. This uncertainty quantification is essential for providing operators with an understanding to the variation associated with the leak estimation procedure, and thus, to enable targeted maintenance actions. By integrating real-time sensor data, advanced algorithms, and uncertainty analysis, the system offers a robust and efficient solution for detecting multiple methane leaks in industrial settings.

[0012] Both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the examples, as claimed.BRIEF DESCRIPTION OF THE DRAWINGS

[0013] FIG. 1 is an illustration of an example method for detecting multiple methane leaks at a site.

[0014] FIG. 2 shows a series of example charts depicting the methane leak number identification procedure.

[0015] FIG. 3 is an illustration of a site map depicting marked solution leak locations.

[0016] FIG. 4 is an illustration of an example method for identifying source locations of multiple leaks at a site.

[0017] FIGS. 5A-5D are visual representations of the example method of FIG. 4.

[0018] FIG. 6 is an illustration of an example system diagram for detecting multiple methane leaks.DESCRIPTION OF THE EXAMPLES

[0019] Reference will now be made in detail to the present examples, including examples illustrated in the accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings to refer to the same or like parts.

[0020] The present invention provides a system and method for detecting and identifying multiple methane leaks at an industrial site, such as an oil or gas facility. The system uses a network of methane sensors strategically placed throughout the site to continuously monitor methane levels. The sensor data is analyzed by a central computing device, which applies advanced algorithms to distinguish between background methane fluctuations and actual leak events. The method generates spatial distributions of methane concentrations to identify potential leak regions and estimate their exact locations. Additionally, the system employs probabilistic models to quantify the uncertainty of the estimated leak sources, providing a certainty level for each identified leak. This comprehensive approach ensures informed detection and localization of multiple leaks given the sensor data collected, even in complex and overlapping scenarios.

[0021] FIG. 1 illustrates a method for detecting multiple methane leaks at a site, using information collected from fixed-point sensors and an onsite weather station. The stages described below can be executed by a computing device, which can be any processor-based device, such as a computer, tablet, cell phone, or server. At stage 102, the computing device can initialize key parameters of the procedure (including Ntop, Nlim, Rmax, Gmax, and Atols).These parameters define algorithmic properties, where Ntop is the number of subspaces initially selected for from the full set of specified subspaces, Nlim represents the maximum number of subspaces that can be considered, Rmax defines a relaxation factor for an active leak , Gmax specifies the maximum number of subspaces that can be selected in a group , and Atols is the acceptance threshold used to accept a potential leak.

[0022] Once these parameters are initialized, the computing device can establish two initial properties: fnull and nrecs at stage 104. Fnull indicates the upper error bound, and nrecs refers to the number of records extracted from the data collected by the sensors over the specified time period T. At stage 106, if the number of records, nrecs, is less than five (nrecs < 5), the data is deemed insufficient for further analysis at stage 108, and the method terminates at stage 110.

[0023] If, however, nrecs is greater than or equal to five (nrecs > 5), the computing device can apply a multi-leak detection algorithm beginning at stage 112, where the computing device defines the initial set of subspaces in consideration as NIND. NIND refers to the set of defined subspaces retained for evaluation. The goal is to reduce this set by carefully eliminating subspaces that are not good candidates, eventually resulting in the few that indicate potential leaks. Notably, NIND0 can be defined over the range from 2 to Ns, where Ns is the total number of potential subspaces, to perform a sensitivity evaluation of each subspace.

[0024] At stage 114, the computing device can evaluate the impact of each subspace in turn. This means that the inverse problem is solved to identify only one leak on the selected subspace. The value of the error measure used for inversion indicates the merit of a leak on the subspace. As a result, subspaces with low error estimates are more likely candidates because the mismatch between observed data from the sensors to the forward model predications is minimized, and those with high values are far less likely candidates for a leak. In an example,the computing device can solve the single-leak inversion method as an underlying solver(referred to hereinafter as the kernel method) by running cases with NSETj = [j] for j 6 [2 ns] to assess potential contribution value vj(X) of each space, as stated above. Subsequently, the computing device can select a predetermined portion of the subspaces (e.g., 2 / 3 given by Ntop) subject to the group selection limit (Gmax = 2). This effectively removes a portion of the leak candidates as undesirable or poor candidates. At stage 116, the computing device can select Ntop spaces, where Ntop refers to the most likely candidate spaces given the sensitivity results established stage 114.

[0025] At stage 118, the computing device can filter Ntop subspaces according to the list of all subspaces, the associated parent groups (GroupSpec) and the limit on the number of subspaces permitted per group (Gmax) to give the set of retained subspaces as NIND1. Next, the kernel solver can be applied to the set NIND1, yielding the lower and upper bounds on the error measure (as fLB and fUB respectively) along with the resulting group membership list as GLIST in stage 120. Here, fLB and fUB refer to the lower and upper bounds of the error measure, respectively, and GLIST is the group membership of each retained subspace in solution set NIND2.

[0026] Subsequently, at stage 122, the computing device can evaluate the error measure for every possible combination of subspaces in the retained set NIND2. This results in the Pareto front of error measure versus potential leak count. Note the single subspaces results (are retrieved from the results of sensitivity study (step 114). The Pareto front serves to indicate the trade-off between the expected number of leaks and the resulting error measure. A larger leak number will reduce the error measure, but the at the cost of more complexity. The leak count identification procedure is intended to provision the expected number of leaks such that the gainin each additional leak (starting from none) is manifestly worthwhile. That is the gain exceeds the specified acceptance tolerance given by Atols. Based on the Pareto front and the acceptance conditions imposed the computing device can establish the most likely number of leaks (Lbest) from the results stored in RESTAB - a table showing error measure with increasing leak count. The solution set of candidates (NOPT) are returned at stage 124. At stage 126, the computing device completes the final evaluation, which outputs the identified number of leaks (nopt), the identified subspaces (NOPT) and other pertinent properties as: Lbest, NOPT, nopt, fopt, and SOPT.

[0027] Once these values are determined, the computing device has effectively identified the potential number of methane leaks and their respective subspace locations given the sensor data made available over time period T. The method concludes at stage 110 for the given period, and the process can resume over the next period after a suitable wait time. The detection of methane leaks is based on a systematic evaluation of sensor data, meteorological conditions, and the refined selection of candidate subspaces through the use of the single-leak kernel solver, multi-leak evaluation, and Pareto Front analysis, to effectively identify the number and location of leaks on the site. The procedure makes a hard problem of identifying multiple concurrent leaks tractable.

[0028] FIG. 2 illustrates the leak count identification process and analysis of methane leaks at a site using various methods of error analysis and gains evaluation. The charts in FIG. 2 depict the methane leak detection process at several stages of method 100, allowing for a comprehensive understanding of how leaks are identified and quantified.

[0029] Chart 210 includes a methane leak count determined using the Pareto front derived from an error function analysis. The Pareto front represents a set of optimal solutions,balancing the error measure with methane leak count estimation . The chart 210 highlights the number of leaks identified through the optimization of error functions (solving the inverse problem for the given data), ensuring that the solutions with the lowest error values are prioritized. The Pareto Front shows the reduction in the error measure with leak number count that is limited to the maximum number in consideration (Nlim as stated above). The example plot shows invalid values (as a default upper bound) for leaks greater than three.

[0030] Chart 220 depicts the Pareto Front with error measure versus leak count along with the lower and upper error bound estimates. Here, the upper bound indicates the worst case scenario (if no leak is identified) and the lower bounds is based on the prospect of Nlim identified leaks. The bounded error approach establishes limits for acceptable error margins, within which the system can effectively estimate the number of detected leaks. As illustrated in FIG. 2, the identified number of leaks is marked by a circle 228, based on the iterative evaluation process noted in plot 230.

[0031] Chart 230 depicts the values of iterative and normalized gains with leak count number. These metrics are used to select leaks one-by-one starting from no leaks. As the labels indicate, a normalized gain measures the resulting error value with respect to the noted lower and upper bounds. The iterative gain, on the other hand, measures the gain with respect to each incremental step. In any case, the procedure continues to accept one new leak (on the x-axis) if the gain is above the acceptance threshold and stops when it is below. This helps to identify the number of leaks. As the minimum solution for a given number of leaks on the Pareto Front is associated with a given set of subspaces, the location of the estimated number of leaks is also ascertained. The line 232 in the chart represents the normalized gain value and the line 224 indicates the iterative gain values.

[0032] The iterative selection process ensures that the system improves its leak count estimation given the trade-off between the resulting error measure with the number of leaks. This method is particularly effective for identifying multiple concurrent methane leaks in complex environments, and efficient by way of the specified subspace scheme and the filter process that selectively eliminates less desirable subspaces.

[0033] Collectively, the charts in FIG. 2 demonstrate the use of error function analysis, bounded error measurements, and iterative gain evaluation to provide a comprehensive solution for identifying multiple methane leaks on a site. These methods allow for the system to improve leak estimation, while maintaining error bounds on the inverse solution given the sensor data available over the time period T.

[0034] FIG. 3 illustrates a visual representation of the methane leak detection process over a defined site layout, illustrating both the identified leaks and the actual leaks within the predefined subspaces (for a known controlled test). The method 100, described in FIG. 1, can be applied to the site in FIG. 3, where fixed-point sensors and environmental data are used to detect methane leaks with high accuracy.

[0035] The Site Layout 300 shows a comprehensive overview of the area being monitored for methane leaks, segmented into predefined subspaces 306. Each subspace 306 represents a portion of the overall site, facilitating detailed, localized leak detection. The site is monitored using fixed-point sensors, which are marked as triangles 308 in FIG. 3. These sensors 308 continuously collect methane concentration data, which is then processed using the methodology of method 100 to determine the potential locations of methane leaks.

[0036] In this embodiment, the system identifies multiple methane leaks, as depicted by the identified leaks 302, which are marked at specific locations within the subspaces. Theseidentified leaks 302 represent the system’s best estimation, based on the analysis performed in earlier steps of method 100, including application of the kernel solver and Pareto analysis.

[0037] For comparison, FIG. 3 also shows the actual leaks 304, which represent the ground truth or confirmed methane leak locations within the site from a controlled test. The actual leaks provide a reference point for evaluating the accuracy of the system’s leak detection capabilities. By comparing the identified leaks 302 to the actual leaks 304, the system’s precision in detecting methane leaks can be assessed.

[0038] The layout and analysis shown in FIG. 3 highlight the system’s ability to accurately pinpoint methane leaks in complex environments. The division of the site into predefined subspaces indicative of locations with potential leak sources allows for localized sensitivity and error analysis, while the use of fixed-point sensors ensures continuous monitoring and real-time data collection. This approach ensures that the system can efficiently identify leaks with minimal false positives or undetected leaks.

[0039] The integration of the identified leaks 302 and the actual leaks 304 into the predefined subspaces 306 provides a clear visual representation of how the method 100 operates in practice, facilitating a thorough understanding of the system’s efficacy in methane leak detection across a large site.

[0040] FIG. 4 illustrates a method for identifying source locations of multiple leaks at a site. FIGS. 5A-5D provide a visual representation of an example application of the method depicted in FIG. 4 and are referenced throughout the explanation of FIG. 4. At stage 410, a computing device can identify multiple leaks at a site that includes a variety of equipment components that may act as potential sources of methane emissions. A network of permanently installed fixed point methane sensors can be used to continuously monitor the site. Thesesensors are strategically placed to provide comprehensive coverage of the site, ensuring that potential methane leaks from any component can be detected in real-time. The sensors can be any kind of sensors that can be used for detecting methane, such as IR gas sensors, electrochemical sensors, catalytic bead sensors, laser-based sensors, acoustic sensors, optical gas imaging (“OGI”) cameras, or light detection and ranging (“LiDAR”) sensors.

[0041] The identification of multiple leaks can be facilitated through an advanced algorithm that analyzes the data collected from the sensors. The algorithm aggregates and processes the raw sensor data to distinguish between different emission sources. Using statistical and probabilistic models, the system identifies regions where elevated methane levels are detected and distinguishes between background methane fluctuations and actual leak events.

[0042] When multiple leak sources are present, the algorithm can apply the proposed multi-leak identification method to determine the effective number of leaks and consequently, applying the uncertainty quantification method to establish the variability of the solution. This is done by generating distributions of leak location and rate over all leaks and comparing them across the site. The distributions derived from the inverse solutions can also be displayed showing methane leak locations estimated from data collected from multiple sensors placed around a site. The raw methane concentrations are only known at the sensor location points. These can be processed and visualized as a time series to show how methane levels change over time. The method involves marking "incumbent" solution spaces — regions where the likelihood of a leak is highest — based on the data available using the scheme outline herein. The process enables the system to account for complex scenarios where multiple leaks may occur simultaneously, ensuring a robust identification process even in dense regions or with overlapping leaks.

[0043] This initial identification step is crucial for the subsequent steps of regional grouping and leak source determination. By effectively identifying the number and location of the leaks, the system lays the foundation for further uncertainty analysis, narrowing down the search to related equipment components within the parent group of the selected subspaces. This process is illustrated in Fig. 5.

[0044] Using FIG. 5A as an example, site 502 represents a site where one or more methane leaks have been identified using the multi-leak method described previously herein. Leak regions 504 and 506 represent areas (the noted solution subspaces) where the computing device considers a possible methane leak from stage 410.

[0045] At stage 420, the computing device can re-activate the other subspaces belonging to the same parent group as each of the identified solution subspaces (see FIG. 5B). This predefined grouping is based on the physical layout and proximity of the equipment components, which may include pipelines, valves, compressors, storage tanks, and other infrastructure commonly found in oil and gas facilities.

[0046] The regional or spatial grouping serves to localize the analysis and narrow down potential leak sources within each specific area. The facility can be divided into zones or regions that encompass clusters of equipment components that are geographically or functionally related. These regions can be defined either manually, by an operator familiar with the site layout, or automatically by the system, based on predefined criteria such as the presence of oilfield assets, distance between components, the type of equipment, and historical leak patterns, or by some uniform partitioning scheme (e.g., a grid).

[0047] The purpose of segmenting the site into regional groups is to enhance the effectiveness of the leak identification process. By breaking down the site into smaller,manageable areas, the system can focus on target regions and the subspaces enclosed therein, to reduce the complexity and improve efficiency of the overall search space. The use of subspaces and parent groups to identify locales with potential leaks makes the problem tractable. Without this subspace and parent group definition the process to filtering candidate locations is not amenable. Thus, this localized analysis improves leak detection, as it allows the system to differentiate between overlapping leak signals given the wind conditions.

[0048] Each marked regional group represents a potential source area for one of the identified leaks. For uncertainty quantification, the defined groups are used to bound the search when Markov-Chain Monte Carlo (“MCMC”) method is applied, which illustrated shown in FIG. 5C. The MCMC method undertakes a number of random walks to establish the merit of the error measure with perturbation of the variable space. That is, it accounts for the change in the error measure as solution moves away from the noted optimal configuration. Here, the control set concerns the location (x,y,z) and rate (r) of each leak. Thus, for ns identified leaks, the MCMC search is performed over 2ns dimensions. For the example in FIG. 5C, with ns=2, the search is over 8 dimensions. The uncertainty quantification process will therefore be performed within the bounds of the defined regions, but notably over 2ns dimensions. Several MCMC chains are evaluated and the collected samples remaining after truncation of the burn-in period are used to establish the distributions indicating parameter variability. Importantly, the sample statistics can be gathered by group or by subspace. The mean and minimum values of the distribution of the error in each subspace are then used for selection purposes. That is, a dual objective measure that accounts for the lowest error value and the lowest mean value in a subspace is used to select the most likely leak candidate for reporting purposes. This is illustrated in FIG. 5D, where the highlighted subspaces are those indicated to the client for ease of understanding. All thestatistical data gathered can be reported for uncertainty estimation, or to arrange an alternative selection process as needed.

[0049] By structuring the equipment components into regional groups, the system improves its ability to identify the number of leaks in the first instance, and then to subsequently, better isolate and attribute the leaks to specific subspaces based on the statistical properties gathered.

[0050] An example of the uncertainty quantification process for multiple leaks is illustrated in FIGS. 5B, 5C and 5D as discussed above. FIG. 5B identifies the groups 508 and 510 related to the identified subspaces 504 and 506. The blocks in each group 508 and 510 represent the related subspaces by spatial design. FIG. 5C shows the expanded search bounds 512 and 514 over each group. An expanded search bound is a region of the site 502 that includes all the equipment of a particular group. For example, the expanded search bound 512 includes all the equipment of equipment group 508, and the expanded search bound 514 includes all the equipment of equipment group 510.

[0051] At stage 430, the computing device can designate a plurality of these groups as potential sources of the identified leaks. This designation is based on the data collected by the methane sensors, which provide real-time information on methane concentrations across the various regions of the site. The system can analyze the sensor data within each regional group to determine which groups exhibit elevated methane levels or other indicators of a potential leak. The goal is to identify the regions most likely to contain the sources of the leaks.

[0052] The process of designating groups as leak sources can include statistical analysis and comparison of methane concentration patterns, spatial distribution, and historical data for each group. For example, the computing device can analyze real-time methane sensor data usingtime-series analysis and statistical thresholds to detect anomalies in concentration levels. Spatial distribution techniques, such as geospatial interpolation and gradient analysis, can help visualize and pinpoint areas of high methane intensity within each group. By comparing current data with historical baselines and past leak events, the computing device can refine its leak predictions and identify recurring problem areas.

[0053] Additionally, data fusion and cross-correlation methods integrate sensor readings across multiple regions to assess whether leaks affect adjacent areas or are confined to a specific group. Multivariate analysis and uncertainty quantification, including MCMC methods, can further enhance the accuracy of the designation process by accounting for sensor noise and variability. This comprehensive approach ensures precise identification of regional groups with the highest likelihood of hosting methane leaks.

[0054] The computing device can assign a likelihood score to each regional group based on the strength of the sensor signals, the consistency of elevated methane readings, and the proximity of the readings to critical equipment components within the group. Groups that show a clear pattern of elevated methane levels are designated as probable leak sources, while those with lower or more sporadic readings may be deprioritized or excluded from further analysis.

[0055] This designation process can handle situations where multiple leaks are present by allowing the system to identify several groups as leak candidates simultaneously. Each group is included collectively in the MCMC step to establish parametric variability. The system ensures that the overall number of leaks identified in earlier steps is consistent with the designated groups, i.e., the number of leaks is identified in the first step and the uncertainty quantified in the second step. If the system has detected two distinct leaks, it will designate at least two regional groups as probable sources, but providing the uncertainty distributions at both the group andsubspace level The statistical data can be used to infer the most likely leak location to report as needed.

[0056] By designating a subset of regional groups as potential leak sources, the computing device ensures that subsequent efforts to pinpoint the exact source within each group are targeted and efficient. This step also prepares the computing device for the final step of determining which specific equipment component, within each designated group, is responsible for the leak.

[0057] In summary, the system designates a plurality of regional groups as potential leak sources by analyzing methane concentration data and other relevant parameters. This helps prioritize areas for further investigation and ensures that the subsequent leak identification steps are conducted with higher precision.

[0058] Using FIG. 5B as an example, other equipment groups not shown can be present at the site 502. The computing device can designate equipment groups 508 and 510 as potential leak sources because they are proximally located to the identified leak subspaces 504 and 506.

[0059] At stage 440, the computing device can determine, within each designated group, a specific equipment component most likely to be the source of a respective leak. The computing device can analyze the methane concentration data in conjunction with the spatial distribution of equipment components within each designated group. The goal is to identify patterns in the sensor readings that correlate most strongly with the proximity and configuration of specific equipment. Components such as valves, compressors, pipelines, and storage tanks, which are common in oil and gas facilities, can be prioritized based on their history of leak risks and their distance from the highest methane concentration points.

[0060] The computing device can employ advanced statistical models, such as proximitybased analysis and sensor gradient detection, to further isolate the equipment component that is most likely to be the source of the leak. By assessing how methane concentrations taper off around different components and how the sensor readings change over time, the system can identify equipment that lies closest to the center of the leak signal. This analysis can also take into account the structural and operational characteristics of the equipment, such as age, maintenance history, and operational stress factors, to weigh the likelihood of each component being the culprit.

[0061] In addition, the computing device can integrate historical data for specific equipment components, comparing current readings with past incidents of leaks or equipment failures. This provides further refinement, allowing the system to identify recurring problem areas or components that have exhibited similar behavior in previous leak events. Once the most likely source is determined, the computing device can mark the corresponding equipment component, which can then be targeted for inspection, repair, or replacement in the final stages of leak management.

[0062] Using FIG. 5D as an example, the computing device determines that equipment components 516 (differs from 504) and 518 (same as 506) are the most likely sources of the leaks from their respective groups given the statistical data gathered and the selection measure employed.

[0063] This step ensures that the leak identification process moves from a group level back to the underlying subspace for diagnosis, ultimately facilitating faster and more efficient remediation efforts. Clearly, the reporting can include both the subspace and the group for each leak to aid remediation efforts. That is, the inspection team can target the group first, and thenconfirm the leak on the indicated subspace or one near it should the data have resulted in a different indication.

[0064] FIG. 6 is an illustration of an example system diagram for detecting multiple methane leaks. The system can include a site area 602 that can represent any industrial site where methane emissions are monitored, such as an oil and gas facility, a drilling site, or a production plant. Within the site area 602, there are multiple equipment groups 604, each containing various equipment components that are potential sources of methane leaks. The equipment groups 604 can include compressors, pipelines, valves, storage tanks, or any other infrastructure that may be involved in the production, transport, or storage of methane.

[0065] Distributed across the site area 602, methane sensors 606 are strategically placed to capture real-time methane concentration readings. These sensors can include IR gas sensors, electrochemical sensors, catalytic bead sensors, laser-based sensors, OGI cameras, or any combination of such technologies capable of detecting methane emissions. The methane sensors 606 continuously monitor the methane levels in their respective regions and communicate the gathered data to a centralized computing device 610.

[0066] The methane sensors 606 are configured to transmit methane concentration data via wireless or wired communication channels to the computing device 610. Communication protocols such as WI-FI, BLUETOOTH, ZIGBEE, or other suitable industrial communication standards can be utilized to ensure reliable data transmission. The computing device 610, which can be a server, desktop computer, or another processing unit, receives and processes the methane readings from the sensors.

[0067] The computing device 610 is equipped with an application 612 designed to detect methane leaks by analyzing the sensor data. The application 612 employs advanced algorithmsand probabilistic models to assess the likelihood of methane leaks at different points within the site area 602. By analyzing the spatial distribution of methane concentrations, the application 612 can attribute a certainty level to each detected leak source. This certainty level indicates the system’s confidence in the identification of a specific leak and helps prioritize areas for further inspection or remediation. The application 612 can also utilize machine learning techniques or statistical methods like MCMC to enhance the accuracy of leak detection and uncertainty quantification.

[0068] Additionally, the computing device 610 is equipped with a display 614, which provides visual representations of the detection process and results. The display 614 can show the locations of the identified methane leaks within the site area 602, the corresponding certainty levels, and the relevant equipment groups 604. This allows operators or site personnel to monitor the leak detection process in real-time and make informed decisions about repair and maintenance activities. The display 614 can include graphical user interface (GUI) elements, charts, and spatial maps that assist in visualizing the methane readings and leak sources, ensuring that operators can quickly identify and respond to leak events.

[0069] Through this system, the combination of methane sensors 606, computing device 610, and display 614 enables continuous, automated monitoring of methane leaks at the site, providing an efficient and accurate method for identifying and managing methane emissions.

[0070] Other examples of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the examples disclosed herein. Though some of the described methods have been presented as a series of steps, it should be appreciated that one or more steps can occur simultaneously, in an overlapping fashion, or in a different order. The order of steps presented are only illustrative of the possibilities and those steps can beexecuted or performed in any suitable fashion. Moreover, the various features of the examples described here are not mutually exclusive. Rather any feature of any example described here can be incorporated into any other suitable example. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

Claims

WHAT IS CLAIMED IS:

1. A method for detecting the presence of multiple methane leaks at a site, comprising: receiving methane concentration data from a plurality of methane sensors positioned at different locations within the site; analyzing the received methane concentration data to identify elevated methane levels at one or more regions within the site; applying a multi-leak detection algorithm to the identified elevated methane levels to detect the presence of multiple methane leaks, wherein the algorithm processes the methane concentration data from multiple sensors to distinguish between distinct leak sources; generating spatial distributions of the methane concentrations across the site based on the detected methane leaks; designating a plurality of potential leak regions within the site based on the spatial distributions, wherein each designated region is associated with at least one detected methane leak; and displaying, on a graphical user interface (“GUI”), the plurality of potential leak regions.

2. The method of claim 1, further comprising estimating the locations of the multiple methane leaks by determining the regions of highest methane concentration within the spatial distributions.

3. The method of claim 2, wherein the estimated locations are refined by considering the proximity of the methane sensors to critical equipment components within the site and prioritizing regions near equipment components as probable leak sources.

4. The method of claim 2, wherein the spatial distributions are generated using geospatial interpolation techniques to provide a visual representation of methane concentration gradients across the site.

5. The method of claim 1, further comprising quantifying the likelihood that the estimated locations correspond to actual methane leak sources by applying a statistical model to the methane concentration data.

6. The method of claim 5, wherein the likelihood of each estimated leak location being an actual leak source is expressed as a certainty level, based on the consistency of elevated methane levels over time and the spatial proximity of the leaks to the methane sensors.

7. The method of claim 5, further comprising displaying, on the GUI, the corresponding uncertainty or likelihood values associated with each detected leak region.

8. A non-transitory, computer-readable medium containing instructions that, when executed by a hardware-based processor, causes the processor to perform stages for detecting the presence of multiple methane leaks at a site, comprising: receiving methane concentration data from a plurality of methane sensors positioned at different locations within the site; analyzing the received methane concentration data to identify elevated methane levels at one or more regions within the site;applying a multi-leak detection algorithm to the identified elevated methane levels to detect the presence of multiple methane leaks, wherein the algorithm processes the methane concentration data from multiple sensors to distinguish between distinct leak sources; generating spatial distributions of the methane concentrations across the site based on the detected methane leaks; designating a plurality of potential leak regions within the site based on the spatial distributions, wherein each designated region is associated with at least one detected methane leak; and displaying, on a graphical user interface (“GUI”), the plurality of potential leak regions.

9. The non-transitory, computer-readable medium of claim 8, the stages further comprising estimating the locations of the multiple methane leaks by determining the regions of highest methane concentration within the spatial distributions.

10. The non-transitory, computer-readable medium of claim 9, wherein the estimated locations are refined by considering the proximity of the methane sensors to critical equipment components within the site and prioritizing regions near equipment components as probable leak sources.

11. The non-transitory, computer-readable medium of claim 9, wherein the spatial distributions are generated using geospatial interpolation techniques to provide a visual representation of methane concentration gradients across the site.

12. The non-transitory, computer-readable medium of claim 8, further comprising quantifying the likelihood that the estimated locations correspond to actual methane leak sources by applying a statistical model to the methane concentration data.

13. The non-transitory, computer-readable medium of claim 12, wherein the likelihood of each estimated leak location being an actual leak source is expressed as a certainty level, based on the consistency of elevated methane levels over time and the spatial proximity of the leaks to the methane sensors.

14. The non-transitory, computer-readable medium of claim 12, the stages further comprising displaying, on the GUI, the corresponding uncertainty or likelihood values associated with each detected leak region.

15. A system for detecting the presence of multiple methane leaks at a site, comprising: a memory storage including a non-transitory, computer-readable medium comprising instructions; and at least one hardware-based processor that executes the instructions to carry out stages comprising: receiving methane concentration data from a plurality of methane sensors positioned at different locations within the site; analyzing the received methane concentration data to identify elevated methane levels at one or more regions within the site; applying a multi-leak detection algorithm to the identified elevated methane levels to detect the presence of multiple methane leaks, wherein the algorithmprocesses the methane concentration data from multiple sensors to distinguish between distinct leak sources; generating spatial distributions of the methane concentrations across the site based on the detected methane leaks; designating a plurality of potential leak regions within the site based on the spatial distributions, wherein each designated region is associated with at least one detected methane leak; and displaying, on a graphical user interface (“GUI”), the plurality of potential leak regions.

16. The system of claim 15, the stages further comprising estimating the locations of the multiple methane leaks by determining the regions of highest methane concentration within the spatial distributions.

17. The system of claim 16, wherein the estimated locations are refined by considering the proximity of the methane sensors to critical equipment components within the site and prioritizing regions near equipment components as probable leak sources.

18. The system of claim 16, wherein the spatial distributions are generated using geospatial interpolation techniques to provide a visual representation of methane concentration gradients across the site.

19. The system of claim 15, further comprising quantifying the likelihood that the estimated locations correspond to actual methane leak sources by applying a statistical model to the methane concentration data.

20. The system of claim 15, wherein the likelihood of each estimated leak location being an actual leak source is expressed as a certainty level, based on the consistency of elevated methane levels over time and the spatial proximity of the leaks to the methane sensors.

Images