Real-time flare optimization using an edge device
Automated systems and methods are provided for continuous monitoring of the flaring of waste gas at an industrial facility, which employ an RGB camera operably coupled to a gateway device by a data communication interface. The RGB camera is configured to capture time-series color image frames of a flare and communicate the time-series color image frames to the gateway device. The gateway device includes an image processing module and a flare optimization module executing on the gateway device. The image processing module is configured to process the time-series color image frames to determine at least one flare parameter that provides a qualitative measurement of the combustion efficiency of the flare over time. The flare optimization module is configured to adjust relative amount of waste gas to at least one assist gas for the flare based on the at least one flare parameter to continuously optimize the combustion efficiency of the flare.
1 . An automated system for continuous optimization of a flaring of waste gas at an industrial facility, the automated system comprising:
an RGB camera operably coupled to a gateway device by a data communication interface; and
a pressure sensor configured to measure a waste gas flowline pressure of the waste gas;
wherein:
the RGB camera is configured to capture time-series color image frames of a flare produced at the industrial facility and communicate the time-series color image frames to the gateway device;
the gateway device includes an image processing module and a flare optimization module executing on the gateway device;
the image processing module is configured to process the time-series color image frames to determine at least one flare parameter that provides a qualitative measurement of combustion efficiency of the flare over time; and
the flare optimization module is configured to;
evaluate the waste gas flowline pressure to determine if the waste gas is flowing to a flare tip of the flare, the evaluation including determining if the waste gas flowline pressure is greater than a predetermined criterion which indicates the waste gas is flowing to the flare tip; and
adjust a relative amount of the waste gas to at least one assist gas for the flare based on the at least one flare parameter to continuously optimize the combustion efficiency of the flare.
2 . The automated system according to claim 1 , wherein:
the flare optimization module is configured to control the relative amount of the waste gas to the at least one assist gas that produces the flare to optimize the at least one flare parameter.
3 . The automated system according to claim 2 , further comprising:
a flare stack with the flare tip;
a first flow control device operably coupled to the gateway device by the data communication interface, the first flow control device fluidly coupled to a supply of the waste gas; and
a second flow control device operably coupled to the gateway device by the data communication interface, the second flow control device fluidly coupled to a supply of one or more assist gases;
wherein:
the first flow control device comprises a first electric valve and the second flow control device comprises a second electric valve; and
the gateway device is configured to communicate signals or commands to at least one of the first electric valve or the second electric valve based on an execution of the flare optimization module in order to adjust the relative amount of the waste gas to the one or more assist gases supplied to the flare tip for mixing and combustion that produces the flare to continuously optimize the combustion efficiency of the flare.
4 . The automated system according to claim 3 , further comprising:
the pressure sensor operably coupled to the gateway device by the data communication interface, the pressure sensor configured to measure a flow line pressure of the waste gas supplied to the flare tip and communicate data representing the flow line pressure to the gateway device; and
an ignitor operably coupled to the gateway device by the data communication interface, the ignitor configured to supply an ignition flame to the flare tip when activated;
wherein the gateway device is further configured to process the at least one flare parameter determined by the image processing module and the data representing the flow line pressure communicated from the pressure sensor, and selectively activate the ignitor based on such processing.
5 . The automated system according to claim 1 , wherein:
the at least one assist gas comprises at least one of air, steam, or other assist gas.
6 . The automated system according to claim 1 , wherein:
the at least one flare parameter further represents an amount of efficient combustion of the flare and an amount of inefficient combustion of the flare; and
the flare optimization module optimizes the at least one flare parameter by yielding a more efficient combustion and a less efficient combustion of the flare.
7 . The automated system according to claim 1 , wherein:
the at least one flare parameter represents at least one of a ratio of a smoke to a fire of the flare or a ratio of the fire to the smoke of the flare; and
the flare optimization module optimizes the at least one flare parameter by minimizing the ratio of the smoke to the fire of the flare or by maximizing the ratio of the fire to the smoke of the flare.
8 . The automated system according to claim 7 , wherein:
the time-series color image frames of the flare comprise an RGB image of the flare, the RGB image including pixels;
the ratio of the smoke to the fire of the flare is represented by a first ratio of a first pixel count of the pixels representing the smoke to a second pixel count of the pixels representing the fire; and
the ratio of the fire to the smoke of the flare is represented by a second ratio of the second pixel count to the first pixel count.
9 . The automated system according to claim 1 , wherein:
the at least one flare parameter further represents a color temperature of a combustion of the flare; and
the flare optimization module further optimizes the at least one flare parameter by adjusting the color temperature of the combustion of the flare.
10 . The automated system according to claim 1 , wherein:
the image processing module comprises at least one machine learning model that determines the at least one flare parameter based on a first input including an RGB image of the flare.
11 . The automated system according to claim 10 , wherein:
the at least one machine learning model includes a first machine learning model that is trained to generate a pixel-wise label mask for an arbitrary RGB image of the flare supplied as a second input to the first machine learning model;
the pixel-wise label mask classifies pixels of the arbitrary RGB image of the flare as corresponding to a set of predefined labels; and
the set of predefined labels include a first label representing an efficient combustion of the flare and a second label representing an inefficient combustion of the flare.
12 . The automated system according to claim 11 , wherein:
the image processing module is further configured to process the pixel-wise label mask to determine a first pixel count and a second pixel count;
the first pixel count represents an amount of the efficient combustion of the flare; and
the second pixel count represents an amount of the inefficient combustion of the flare.
13 . The automated system according to claim 11 , wherein:
the first machine learning model is trained with training data that includes at least one RGB image of the flare and a corresponding pixel-wise label mask for the at least one RGB image of the flare; and
the pixel-wise label mask is generated by first image processing operations that segment pixels of the RGB image of the flare that correspond to the efficient combustion of the flare as well as second image processing operations that segment pixels of the RGB image of the flare that correspond to the inefficient combustion of the flare.
14 . The automated system according to claim 11 , wherein:
the first machine learning model comprises a convolution encoder-decoder machine learning model.
15 . The automated system according to claim 1 , wherein:
the image processing module is further configured to process an arbitrary RGB image of the flare to generate output data that represents a color temperature of combustion of the flare.
16 . The automated system according to claim 1 , wherein:
the industrial facility comprises an oil production site, a refinery, or a chemical processing plant.
17 . A method for continuous optimization of a flaring of waste gas at an industrial facility, the method comprising:
providing an RGB camera operably coupled to a gateway device by a data communication interface, wherein the RGB camera is configured to capture time-series color image frames of a flare produced at the industrial facility and communicate the time-series color image frames to the gateway device;
providing a pressure sensor configured to measure a waste gas flowline pressure of the waste gas; and
configuring the gateway device to execute an image processing module and a flare optimization module, wherein:
the image processing module is configured to process the time-series color image frames to determine at least one flare parameter that provides a qualitative measurement of combustion efficiency of the flare over time; and
the flare optimization module is configured to;
evaluate the waste gas flowline pressure to determine if the waste gas is flowing to a flare tip of the flare, the evaluation including determining if the waste gas flowline pressure is greater than a predetermined criterion which indicates the waste gas is flowing to the flare tip; and
adjust a relative amount of the waste gas to at least one assist gas for the flare based on the at least one flare parameter to continuously optimize the combustion efficiency of the flare.
18 . The method according to claim 17 , wherein:
the flare optimization module is configured to control the relative amount of the waste gas to the at least one assist gas that produces the flare to optimize the at least one flare parameter.
19 . The method according to claim 17 , wherein:
the at least one assist gas comprises at least one of air, steam, or other assist gas.
20 . The method according to claim 17 , further comprising:
providing a flare stack with the flare tip at the industrial facility, a first flow control device operably coupled to the gateway device by the data communication interface, the first flow control device fluidly coupled to a supply of the waste gas, and second flow control device operably coupled to the gateway device by the data communication interface, the second flow control device fluidly coupled to a supply of one or more assist gases;
configuring the gateway device to communicate signals or commands to at least one of a first electric valve of the first flow control device or a second electric valve of the second flow control device based on the execution of the flare optimization module in order to adjust the relative amount of the waste gas to the one or more assist gases supplied to the flare tip for a mixing and a combustion that produces the flare to continuously optimize the combustion efficiency of the flare.
21 . The method according to claim 17 , wherein:
the at least one flare parameter further represents an amount of an efficient combustion of the flare and an amount of an inefficient combustion of the flare; and
the flare optimization module optimizes the at least one flare parameter by yielding a more efficient combustion of the flare and a less efficient combustion of the flare.
22 . The method according to claim 17 , wherein:
the at least one flare parameter represents a ratio of a smoke to a fire of the flare or a ratio of the fire to the smoke of the flare; and
the flare optimization module optimizes the at least one flare parameter by minimizing the ratio of the smoke to the fire of the flare or by maximizing the ratio of the fire to the smoke of the flare.
23 . The method according to claim 17 , wherein:
the at least one flare parameter further represents a color temperature of a combustion of the flare; and
the flare optimization module further optimizes the at least one flare parameter by adjusting the color temperature of the combustion of the flare.
24 . The method according to claim 17 , wherein:
the image processing module comprises at least one machine learning model that determines the at least one flare parameter based on a first input including an RGB image of the flare.
25 . The method according to claim 24 , further comprising:
training a first machine learning model of the at least one machine learning model to generate a second input including a pixel-wise label mask for an arbitrary RGB image of the flare, wherein the second input is supplied to the first machine learning model, wherein the pixel-wise label mask classifies pixels of the arbitrary RGB image of the flare as corresponding to a set of predefined labels, wherein the set of predefined labels include a first label representing an efficient combustion of the flare and a second label representing an inefficient combustion of the flare; and
deploying the first machine learning model as part of the image processing module that executes on the gateway device.
26 . The method according to claim 25 , wherein:
the image processing module is further configured to process the pixel-wise label mask to determine a first pixel count and a second pixel count;
the first pixel count represents an amount of the efficient combustion of the flare; and
the second pixel count represents an amount of the inefficient combustion of the flare.
27 . The method according to claim 25 , wherein:
the training of the first machine learning model employs training data that includes at least one RGB image of the flare and a corresponding pixel-wise label mask for the RGB image of the flare; and
the pixel-wise label mask is generated by first image processing operations that segment pixels of the RGB image of the flare that correspond to the efficient combustion of the flare as well as second image processing operations that segment pixels of the RGB image of the flare that correspond to the inefficient combustion of the flare.
28 . The method according to claim 25 , wherein:
the first machine learning model comprises a convolution encoder-decoder machine learning model.
29 . The method according to claim 17 , further comprising:
configuring the image processing module to process an arbitrary RGB image of the flare to generate data that represents a color temperature of a combustion of the flare.
30 . The method according to claim 17 , further comprising:
providing the pressure sensor operably coupled to the gateway device by the data communication interface, the pressure sensor configured to measure a flow line pressure of the waste gas supplied to the flare tip and communicate data representing the flow line pressure to the gateway device, and an ignitor operably coupled to the gateway device by the data communication interface, the ignitor configured to supply an ignition flame to the flare tip when activated; and
configuring the gateway device to process the at least one flare parameter determined by the image processing module and the data representing the flow line pressure communicated from the pressure sensor, and selectively activate the ignitor based on the processing.
31 . The method according to claim 17 , wherein:
the industrial facility comprises an oil production site, a refinery, or a chemical processing plant.