Various embodiments include a combustion apparatus comprising: a facility for open- and/or closed-loop control of the apparatus; a combustion chamber; an actuator adjusting an air supply; and a combustion sensor in a region of a flame of the chamber. The controller stores a list of support points. A first air supply value is assigned to each support point. A drift test value and an index for ascertainment of a test result are assigned to each support point. The controller: generates a specified air supply; selects a support point as a function of the air supply; and decides on a test result using the index for the support point. To ascertain a test result: receives a signal from the combustion sensor; determines a new test result; ascertains a changed drift test value for the selected support point; and stores the changed drift test value as the drift test value.

A burner system is disclosed. In one example, the burner system includes an artificial intelligence configured to be executed on a processing element. The burner control system may define a control envelope and include a burner, an oxidizer subsystem, and a fuel subsystem. The oxidizer subsystem and the fuel subsystem may include one or more control devices operative to supply an oxidizer and a fuel to the burner to support a combustion process within the burner. The artificial intelligence may be operative to control the burner control system on a trim control curve within the control envelope.

This disclosure relates to method and system for training of deep learning based time-series models based on self-supervised learning. The problem of missing data is taken care of by introducing missing-ness masks. The deep learning model for univariate and multivariate time series data is trained with the distorted input data using the self-supervised learning to reconstruct the masked input data. Herein, the one or more distortion techniques include quantization, insertion, deletion, and combination of the one or more such distortion techniques with random subsequence shuffling. Different distortion techniques in the form of reconstruction of masked input data are provided to solve. The deep learning model performs these different distortion techniques, which force the deep learning model to learn better features. It is to be noted that the system uses a lot of unlabeled data available cheaply as compared to the label or annotated data which is very hard to get.

In an embodiment, a method of controlling flaring of a combustion gas including a flare gas, a supplemental fuel gas, and an assist gas is provided. Models estimating, based on flow rates and in-situ speed of sound measurements in the gases, net heating value of the combustion gas within a flare combustion zone, combustion efficiency of the combustion gas, and smoke yield of the combustion gas are maintained. The method also includes receiving measurements of the gas flow rates and determining set points for flow rates of the fuel gas and/or the assist gas based upon the models that achieve a target combustion efficiency. When a difference between a determined set point and its corresponding flow rate for the fuel gas and/or the assist gas is greater than a corresponding predetermined tolerance amount, that flow rate can be adjusted to reduce the determined difference below the predetermined tolerance amount.

A flame detection device that uses a breakthrough voltage across a pair of electrodes located in a flame zone to detect the presence of a flame. The flame detection device may be used with a burner that is part of a furnace in a central heating system for a home or building. Unlike conventional flame detection devices that measure ionization current in a flame, the flame detection device detects a flame by determining the voltage required for a spark event across a spark gap located in a flame zone (also referred to as the breakthrough voltage), and evaluating the breakthrough voltage and/or its various characteristics to detect the presence or absence of a flame. According to one example, the flame detection device includes a power supply, an ignition unit, output wires, insulators, and electrodes.

This disclosure relates to systems and methods of predicting physical parameters for a combustion fuel system. In one embodiment of the disclosure, a method of predicting physical parameters of a combustion fuel system includes causing water injection in at least one combustor. The water injection is associated with at least one time and performed during gaseous fuel operations or after liquid fuel operations. The method includes measuring exhaust spread data associated with the water injection and allows correlating the exhaust spread data to at least one physical parameter associated with a nozzle or a valve of the fuel system. The method further includes storing the exhaust spread data, the at least one physical parameter, and the at least one time to a database. The method further provides stored historical data from the database to an analytical model. The analytical model is operable to predict, based at least partially on the stored historical data, at least one future physical parameter associated with a future time.

A combustion control system that collects and accumulates, at predetermined intervals, combustion controlling information including at least a signal indicating an operating state of an ignition device, a signal indicating the supplying state of fuel to a pilot burner, a signal indicating the supplying state of fuel to a main burner, a signal indicating the supplying state of air to the main burner, a flame detection signal indicating the strengths of the flames of the burners (the pilot burner and the main burner), and information indicating presence or absence of the flames determined based on the value of the flame detection signal, and displays the combustion controlling information as trend data of combustion control in a graph in such a form that the relationship between time zones of the combustion sequences from the start of a combustion device to normal combustion and the trend data is understandable.

Systems and methods for monitoring emissions of a combusted gas are provided. The method includes determining a first net heating value of a flare gas. The method also includes determining a second net heating value of a combustion gas including the flare gas. The second net heating value can be determined based upon the first net heating value and a volumetric flow rate of the flare gas. Based upon the value of the second net heating value, an empirical model or a non-parametric machine learning model can be selected. A combustion efficiency of the combustion gas can be determined using the selected model, the second net heating value, and selected ones of the process conditions and the environmental conditions. Total emissions of the combustion mixture can be further determined from the combustion efficiency and a volumetric flow rate of the combustion gas.

A control device is a control device for a gas turbine including a plurality of combustors and is configured to select combustors to ignite in accordance with a target load on the basis of a predictor which defines a relationship between a load and the number and arrangement of combustors to ignite and a combustion temperature.

An apparatus for combustion optimization is provided. The apparatus for combustion optimization includes a management layer configured to collect currently measured real-time data for boiler combustion, and to determine whether to perform combustion optimization and whether to tune a combustion model and a combustion controller by analyzing the collected real-time data, a data layer configured to derive learning data necessary for designing the combustion model and the combustion controller from the real-time data and previously measured past data for the boiler combustion, a model layer configured to generate the combustion model and the combustion controller through the learning data, and an optimal layer configured to calculate a target value for the combustion optimization by using the combustion model and the combustion controller, and to output a control signal according to the calculated target value.