| CPC F23N 5/265 (2013.01) [F23G 7/08 (2013.01); F23L 7/005 (2013.01); G06V 20/52 (2022.01); G06V 20/60 (2022.01); F23N 2229/20 (2020.01); F23N 2237/22 (2020.01)] | 15 Claims |

|
1. A method of assessment of the quality of the burn of gases in a flare system and adjustment to water vapor flow rate in a continuous and constant way, the method comprising:
a. obtaining a flare flame image set comprising images;
b. using computer vision techniques, quantifying a flare flame height of each of the images;
c. using a deep learning model, classifying the images into four states: flame with excess vapor, optimized flame, flame with soot and images with insufficient information;
d. combining the quantification of step “b” with the results of step “c” to classify the image set, obtained in step “a”, in one of the four states of the flare flame;
e. using retraining of an image classification model;
f. adjusting the water vapor flow rate in a continuous, automated and integrated way to the water vapor injection control system in the flare system, by means of the information obtained in steps “a”, “b”, “c” and “e”;
g. employing the following components: flare, camera, image stream manager, edge computer, data historian, alert manager, information visualization panels, distributed digital control system (DDCS), and cloud storage and computing.
|