Gas turbine NOx emission prediction method and system based on ARX model

Abstract

The invention relates to the technical field of gas emission prediction, and discloses a gas turbine NOx emission prediction method and system based on an ARX model, and the method comprises the steps: constructing an ARX model with a residual error according to the historical data of a gas turbine combustion system; according to the input quantity and the output quantity under each working condition in the combustion system, identifying time delay and identification parameters in the discretized ARX model by adopting a parameter identification method; taking the historical NOx concentration under each working condition as a verification set, inputting the verification set into the ARX model to obtain a prediction data set of the NOx concentration, and judging the prediction precision of data in the prediction set; carrying out uncertainty analysis on the ARX model by adopting an optimization algorithm, and evaluating the reliability and stability of a prediction result; the working condition of the combustion system is identified by using a mode identification technology, and the ARX model is improved according to an identification result. The NOx prediction method provided by the invention has higher accuracy and strong robustness to external interference, and compared with hardware measurement, the development cost is lower.

Description

Technical Field

[0001] The present invention relates to the technical field of gas emission prediction, and in particular to a method and system for predicting NOx emissions of a gas turbine based on an ARX model. Background Art

[0002] NOx generated during the combustion process is one of the important sources of atmospheric pollutants, which seriously affects air quality and human health. In order to cope with the severe environmental pollution problem, countries are constantly tightening NOx emission limits. The NOx emission standards of gas turbines vary by region and model. Prior knowledge or a posteriori knowledge is used in the process of establishing a prediction model. Prior knowledge includes knowledge such as NOx generation mechanism, combustion mechanism, and fluid mechanics. A posteriori knowledge mainly uses data for system identification to construct unknown internal relationships of the system.

[0003] The linear auto regressive model with exogenous inputs (ARX) is a time series analysis method used to model, predict and control dynamic systems. It combines the autoregressive (AR) model and the model of extra inputs. The characteristic of this model is that it can capture important information about the system state by analyzing time series data, thereby expressing the operating rules of the system in a profound and concentrated manner. The core idea of ​​the ARX model is to describe the dynamic characteristics of the system by fitting autoregressive (AR) and moving average (MA) coefficients based on past input and output data. This model consists of two parts, autoregressive (AR) and moving average (MA), and is a commonly used linear time-invariant system model.

[0004] The ARX model has a wide range of applications, including but not limited to modeling, prediction, and control of dynamic systems. It not only considers the impact of historical data on the current state, but also incorporates external input factors, such as external meteorological parameters, to improve prediction accuracy. This method has been applied in many fields. For example, in the dynamic prediction of building cooling load, the ARX model can provide higher prediction accuracy than the traditional AR model by combining historical data and external meteorological parameters. Summary of the invention

[0006] In view of the above existing problems, the present invention is proposed.

[0007] Therefore, the present invention provides a gas turbine NOx emission prediction method based on the ARX model, which has higher accuracy and strong robustness to external interference and reduces development costs.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for predicting NOx emissions from a gas turbine based on an ARX model, comprising: modeling the NOx generation mechanism according to historical data of a gas turbine combustion system, constructing an ARX model with residuals, and performing time discretization processing on the constructed model; using a parameter identification method to identify the time delay and identification parameters in the discretized ARX model according to the input and output under each operating condition in the combustion system; using the historical NOx concentration under each operating condition as a validation set, inputting it into the ARX model to obtain a prediction data set of NOx concentration, and judging the prediction accuracy of the data in the prediction set; using an optimization algorithm to perform uncertainty analysis on the ARX model, and evaluating the reliability and stability of the prediction results; using pattern recognition technology to identify the operating conditions of the combustion system, and improving the ARX model according to the identification results.

[0009] As a preferred solution of the gas turbine NOx emission prediction method based on the ARX model described in the present invention, the historical data includes the operating status of the gas turbine under different working conditions, including temperature, pressure, flow rate, speed; the mixing ratio of fuel and air, combustion temperature, combustion chamber pressure; historical NOx emission concentration data; fuel type, composition, calorific value; environmental conditions and parameters in the gas turbine control system;

[0010] The data can be made consistent by removing duplicate records, correcting obvious errors, and processing outliers. At the same time, the data can be converted into a unified format and unit for comparison and analysis with data from different sources and types to make the data from different data sources consistent.

[0011] As a preferred solution of the gas turbine NOx emission prediction method based on the ARX model described in the present invention, wherein: the construction of the ARX model with residuals includes extracting a formula describing the basic relationship between system variables according to the mechanism formula between the input variables and the output variables, the fuel amount m affects the combustion temperature, the influence of the combustion temperature on the NOx generation rate is in the form of exp(c / T), then the influence of the fuel amount on NOx through the temperature is exp(c1 / m+b), c1 and b are unknown parameters, called primitives;

[0012] Determine the candidate input variables of the model, which consist of each primitive, the coupling form between primitives, and the time delay; use the candidate variables and experimental or operational data to identify unknown parameters and determine the model structure:

[0013]

[0014] φ0(t)=1

[0015]

[0016] Where y(t) represents the predicted NOx concentration, φ i (t) represents the input variable, k represents the time delay, u j represents the control variable, ε(t) represents the deviation, θ i represents the model parameters;

[0017] The constructed model is subjected to time discretization processing, and the continuous-time model is converted into a discrete-time model. In the discrete-time model, a difference equation is used to replace a differential equation, and the integral is converted into a summation form. The discretized ARX model is expressed as:

[0018]

[0019] t z = zΔt

[0020] Among them, t z represents a discrete time point, Δt is a sampling interval, z represents an integer index, and the discretized basis function is expressed as:

[0021]

[0022] Among them, k j and l are u j and the time delay parameters of y.

[0023] As a preferred solution of the gas turbine NOx emission prediction method based on the ARX model described in the present invention, wherein: the time delay and identification parameters in the discretized ARX model are identified,

[0024]

[0025] Where n represents the number of control variables, m represents the number of output variables, and λ represents the attenuation parameter;

[0026] According to the structure of the ARX model, the ARX model is parameterized into an estimable form and the objective function is established:

[0027]

[0028] in, Represents the output predicted by the model, and N represents the number of data points; the finalized model is applied to the actual combustion system to predict the NOx generation concentration, and the input variables are adjusted as needed to achieve the desired NOx emission level.

[0029] As a preferred solution of the gas turbine NOx emission prediction method based on the ARX model described in the present invention, wherein: the prediction accuracy of the data in the prediction set is determined, including collecting historical NOx concentration data under each operating condition as a validation set, cleaning and preprocessing the data, using the historical data to train the ARX model, identifying model parameters, including time delay parameters and model coefficients; inputting the data in the validation set into the trained ARX model, and calculating the predicted value of NOx concentration:

[0030]

[0031] Among them, P(t z ) represents the discrete time point t z The predicted value of NOx concentration;

[0032] The data missing rate does not exceed 5%. Outliers need to be detected and processed by statistical methods. 80% of the historical data is used as the training set, and the remaining 20% ​​is used as the validation set. The root mean square error (RMSE) and the coefficient of determination (R) are used to determine the 2 Determine the accuracy of the forecast;

[0033] When RMSE is less than 10ppm and R 2 When the RMSE is between 10-20ppm and R 2 When RMSE is between 0.7 and 0.85, the prediction accuracy is “medium”; when RMSE is above 20 ppm or R 2 When it is lower than 0.7, the prediction accuracy is “low”;

[0034] When the prediction accuracy is "medium" or "low", the model's input variables and time delay parameters are reconsidered, and the discretized time delay and identification parameters are re-identified based on the input and output quantities under each operating condition in the combustion system. When the prediction accuracy is "high", the model is used for actual prediction, verification and update.

[0035] As a preferred solution of the gas turbine NOx emission prediction method based on the ARX model described in the present invention, the uncertainty analysis includes considering the cooling coefficient in the simulated annealing algorithm, the uncertainty parameters of the model, the temperature of the gas turbine, and the amount of fuel, and performing the uncertainty analysis of the ARX model through the simulated annealing algorithm in the optimization algorithm to evaluate the reliability and stability of the gas turbine NOx emission prediction.

[0036]

[0037] Where R represents the evaluation result, λ1 represents the cooling coefficient in the simulated annealing algorithm, δ represents the uncertainty parameter of the model, T represents the temperature of the gas turbine, m' represents the amount of fuel, and p represents the number of model parameters;

[0038] The iterative process of the simulated annealing algorithm is expressed as:

[0039]

[0040] in, represents the value of the i-th model parameter at the b-th iteration, It represents the update amount of the i-th model parameter at the b-th iteration, b represents the number of iterations, and the optimized model parameters are used to calculate the evaluation result R. If R is greater than 0.75 and less than 1 and infinitely tends to 1, it means that the prediction reliability is high; if R is less than or equal to 0.75, it means that the prediction uncertainty is high.

[0041] As a preferred solution of the gas turbine NOx emission prediction method based on the ARX model described in the present invention, the improvement of the ARX model includes collecting the current operating data of the combustion system, monitoring and recording the data in real time, inputting the real-time data under the current working condition into the constructed ARX model, comparing the real-time data with the historical data, identifying the working condition mode of the current combustion system,

[0042]

[0043] Among them, F represents the recognition result of the combustion system operating mode, X t Indicates real-time data, X h Represents historical data, T d Indicates time delay;

[0044] According to the prediction accuracy of the model, analyze the performance of the model under the current working conditions, and improve the ARX model based on the uncertainty analysis results and prediction accuracy evaluation;

[0045] By comparing the predicted NOx emissions with the actual measured values, the model improvement process is initiated when the prediction error exceeds 5%;

[0046] Determine the prediction stability by evaluating the change in prediction error of the model in different time periods. When the prediction stability is lower than 0.95, start the model improvement process;

[0047] The frequency of model adjustment is determined by recording the number of times the model is adjusted. When the model needs to be adjusted within 3 consecutive days, the optimization algorithm is started;

[0048] The recognition accuracy of the combustion system operating condition is determined by comparing the matching degree between the recognized operating condition mode and the actual operating condition. When the recognition accuracy is lower than 90%, the optimization algorithm is started.

[0049] Another object of the present invention is to provide a gas turbine NOx emission prediction system based on the ARX model, which can better adapt to different operating conditions and improve the operating efficiency and environmental protection performance of the gas turbine.

[0050] As a preferred solution of the gas turbine NOx emission prediction system based on the ARX model described in the present invention, it includes: a data acquisition and processing module, an operating condition identification module, an ARX model construction module, and a prediction and control module;

[0051] The data acquisition and processing module collects key parameter data of the gas turbine during operation, cleans the collected data, removes noise and outliers, normalizes the data, and eliminates the influence of different dimensions on the model;

[0052] The operating condition identification module uses pattern recognition technology to classify data, identify different operating modes of the combustion system, determine the characteristic parameters under each operating mode, and provide a basis for the selection and parameter adjustment of the ARX model;

[0053] The ARX model building module selects a suitable ARX model for prediction, determines the model's input variables, time delay parameters and model parameters, builds the ARX model, and performs training and verification;

[0054] The prediction and control module inputs real-time data into the optimized ARX model, performs NOx emission prediction, outputs prediction results, and controls and adjusts the combustion system according to the prediction results.

[0055] A computer device includes a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, it implements the steps of a gas turbine NOx emission prediction method based on an ARX model.

[0056] A computer-readable storage medium stores a computer program thereon, wherein when the computer program is executed by a processor, the computer program implements the steps of a gas turbine NOx emission prediction method based on an ARX model.

[0057] The beneficial effects of the present invention are as follows: the present method can derive model parameters through known input and output sequences, thereby reducing computational costs; the model can be used for real-time monitoring of pollutants, replacing hardware measurement with soft measurement methods, thereby greatly reducing the economic cost of NOx monitoring; when the noise is relatively small, the ARX model has the advantages of robustness and simple algorithm; when the noise is relatively large, the order of the model can be appropriately increased to compensate for the influence of noise on identification accuracy; compared with the prior art, the NOx prediction method proposed in the present invention has higher accuracy and strong robustness to external interference, and compared with hardware measurement, the development cost is also lower. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0059] Figure 1 A schematic flow chart of a method for predicting NOx emissions from a gas turbine based on an ARX model is provided as an embodiment of the present invention.

[0060] Figure 2 A schematic diagram of working modules of a gas turbine NOx emission prediction system based on an ARX model is provided in accordance with an embodiment of the present invention. DETAILED DESCRIPTION

[0061] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0062] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0063] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0064] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0065] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0066] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0067] Example 1

[0068] Reference Figure 1 , which is the first embodiment of the present invention, provides a gas turbine NOx emission prediction method based on an ARX model, comprising:

[0069] S1: Based on the historical data of the gas turbine combustion system, the NOx generation mechanism is modeled, an ARX model with residuals is constructed, and the constructed model is time discretized.

[0070] Furthermore, the historical data includes the operating status of the gas turbine under different operating conditions, including temperature, pressure, flow rate, speed; the fuel-air mixture ratio, combustion temperature, combustion chamber pressure; historical NOx emission concentration data; fuel type, composition, calorific value; environmental conditions and parameters in the gas turbine control system;

[0071] The data can be made consistent by removing duplicate records, correcting obvious errors, and processing outliers. At the same time, the data can be converted into a unified format and unit for comparison and analysis with data from different sources and types to make the data from different data sources consistent.

[0072] It should be noted that the gas turbine is a complex system with extremely strong nonlinear characteristics. The NOx emission characteristics are affected by many factors such as fuel type, onboard load, burner type and operating mode, involving turbulence, combustion, heat transfer and mass transfer, and it is difficult to reflect it with a simple process. Generally speaking, there are three main mechanisms for the generation of NOx:

[0073] (1) Thermal NOx: When the temperature is higher than 1500℃, the nitrogen oxides generated by the reaction of nitrogen and oxygen in the air are called thermal NOx. This type of nitrogen oxide is only formed at high temperatures, so it is usually called thermal NOx. The formation mechanism of thermal NOx is the result of a series of unbranched chain reactions caused by oxygen atoms colliding with nitrogen molecules at high temperatures. This process is closely related to temperature, oxygen concentration and residence time in the high temperature zone;

[0074] (2) Rapid NOx: The formation mechanism of rapid NOx is different from that of thermal NOx. It mainly occurs when hydrocarbons are burned and the fuel is rich, that is, when there is more CH in hydrocarbons and the oxygen concentration is relatively low. It is generated quickly and is formed on the flame surface. The formation of rapid NOx has little to do with temperature, but is closely related to the decomposition of nitrogen compounds in the fuel and the combustion conditions;

[0075] (3) Fuel-type NOx: Fuel-type NOx is nitrogen oxides formed by thermal decomposition and oxidation of nitrogen-containing compounds in the fuel during combustion. The generation process of this type of NOx is very complex, involving multiple chemical reactions and chemical kinetic parameters. Nitrogen compounds in the fuel decompose at high temperatures to generate intermediate products such as HCN and NH2, and gradually generate NO. This reaction process can be carried out at temperatures below 1600°C. The nitrogen content and form of the fuel have a great influence on the generation of fuel-type NOx.

[0076] Furthermore, the construction of the ARX model with residuals includes extracting a formula describing the basic relationship between system variables according to the mechanism formula between the input variables and the output variables, the fuel amount m affects the combustion temperature, the combustion temperature affects the NOx generation rate in the form of exp(c / T), then the fuel amount affects NOx through temperature in the form of exp(c1 / m+b), c1 and b are unknown parameters, called primitives;

[0077] Determine the candidate input variables of the model, which consist of each primitive, the coupling form between primitives, and the time delay; use the candidate variables and experimental or operational data to identify unknown parameters and determine the model structure:

[0078]

[0079] φ0(t)=1

[0080]

[0081] Where y(t) represents the predicted NOx concentration, φ i (t) represents the input variable, k represents the time delay, u j represents the control variable, ε(t) represents the deviation, θ i represents the model parameters;

[0082] The constructed model is subjected to time discretization processing, and the continuous-time model is converted into a discrete-time model. In the discrete-time model, a difference equation is used to replace a differential equation, and the integral is converted into a summation form. The discretized ARX model is expressed as:

[0083]

[0084] t z = zΔt

[0085] Among them, t z represents a discrete time point, Δt is a sampling interval, z represents an integer index, and the discretized basis function is expressed as:

[0086]

[0087] Among them, k j and l are u j and the time delay parameters of y.

[0088] S2: According to the input and output under each operating condition in the combustion system, the parameter identification method is used to identify the time delay and identification parameters in the discretized ARX model.

[0089] Furthermore, the identification of the time delay and identification parameters in the discretized ARX model includes:

[0090]

[0091] Where n represents the number of control variables, m represents the number of output variables, and λ represents the attenuation parameter;

[0092] According to the structure of the ARX model, the ARX model is parameterized into an estimable form and the objective function is established:

[0093]

[0094] in, Represents the output predicted by the model, and N represents the number of data points; the finalized model is applied to the actual combustion system to predict the NOx generation concentration, and the input variables are adjusted as needed to achieve the desired NOx emission level.

[0095] S3: The historical NOx concentration under each operating condition is used as a validation set and input into the ARX model to obtain a prediction data set of NOx concentration, and the prediction accuracy of the data in the prediction set is determined.

[0096] Furthermore, the prediction accuracy of the data in the prediction set is determined by collecting historical NOx concentration data under each operating condition as a validation set, cleaning and preprocessing the data, training the ARX model using the historical data, and identifying model parameters, including time delay parameters and model coefficients; inputting the data in the validation set into the trained ARX model, and calculating the predicted value of NOx concentration:

[0097]

[0098] Among them, P(t z ) represents the discrete time point t z The predicted value of NOx concentration;

[0099] It should be noted that the data missing rate does not exceed 5%, and outliers need to be detected and processed by statistical methods. 80% of the historical data is used as the training set, and the remaining 20% ​​is used as the validation set. The root mean square error RMSE and the determination coefficient R 2 Determine the accuracy of the forecast;

[0100] When RMSE is less than 10ppm and R 2 When the RMSE is between 10-20ppm and R 2 When RMSE is between 0.7 and 0.85, the prediction accuracy is “medium”; when RMSE is above 20 ppm or R 2 When it is lower than 0.7, the prediction accuracy is “low”;

[0101] When the prediction accuracy is "medium" or "low", the model's input variables and time delay parameters are reconsidered, and the discretized time delay and identification parameters are re-identified based on the input and output quantities under each operating condition in the combustion system. When the prediction accuracy is "high", the model is used for actual prediction, verification and update.

[0102] S4: Use optimization algorithms to perform uncertainty analysis on the ARX model and evaluate the reliability and stability of the prediction results.

[0103] Furthermore, the uncertainty analysis includes considering the cooling coefficient in the simulated annealing algorithm, the uncertainty parameters of the model, the temperature of the gas turbine, and the amount of fuel, and performing uncertainty analysis of the ARX model through the simulated annealing algorithm in the optimization algorithm to evaluate the reliability and stability of the gas turbine NOx emission prediction.

[0104]

[0105] Where R represents the evaluation result, λ1 represents the cooling coefficient in the simulated annealing algorithm, δ represents the uncertainty parameter of the model, T represents the temperature of the gas turbine, m' represents the amount of fuel, and p represents the number of model parameters;

[0106] The iterative process of the simulated annealing algorithm is expressed as:

[0107]

[0108] in, represents the value of the i-th model parameter at the b-th iteration, It represents the update amount of the i-th model parameter at the b-th iteration, b represents the number of iterations, and the optimized model parameters are used to calculate the evaluation result R. If R is greater than 0.75 and less than 1 and infinitely tends to 1, it means that the prediction reliability is high; if R is less than or equal to 0.75, it means that the prediction uncertainty is high.

[0109] S5: Use pattern recognition technology to identify the operating conditions of the combustion system and improve the ARX model based on the identification results.

[0110] Furthermore, the improvement of the ARX model includes collecting the current operating data of the combustion system, monitoring and recording the data in real time, inputting the real-time data under the current working condition into the constructed ARX model, comparing the real-time data with the historical data, identifying the working mode of the current combustion system,

[0111]

[0112] Among them, F represents the recognition result of the combustion system operating mode, X t Indicates real-time data, X h Represents historical data, T d Indicates time delay;

[0113] According to the prediction accuracy of the model, analyze the performance of the model under the current working conditions, and improve the ARX model based on the uncertainty analysis results and prediction accuracy evaluation;

[0114] By comparing the predicted NOx emissions with the actual measured values, the model improvement process is initiated when the prediction error exceeds 5%;

[0115] Determine the prediction stability by evaluating the change in prediction error of the model in different time periods. When the prediction stability is lower than 0.95, start the model improvement process;

[0116] The frequency of model adjustment is determined by recording the number of times the model is adjusted. When the model needs to be adjusted within 3 consecutive days, the optimization algorithm is started;

[0117] The recognition accuracy of the combustion system operating condition is determined by comparing the matching degree between the recognized operating condition mode and the actual operating condition. When the recognition accuracy is lower than 90%, the optimization algorithm is started.

[0118] Example 2

[0119] The second embodiment of the present invention is different from the first two embodiments in that:

[0120] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0121] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0122] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0123] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0124] Example 3

[0125] Reference Figure 2 , which is an embodiment of the present invention, provides a gas turbine NOx emission prediction system based on the ARX model, characterized by: including a data acquisition and processing module, an operating condition identification module, an ARX model construction module, and a prediction and control module.

[0126] The data acquisition and processing module collects key parameter data during gas turbine operation, cleans the collected data, removes noise and outliers, normalizes the data, and eliminates the impact of different dimensions on the model.

[0127] The operating condition identification module uses pattern recognition technology to classify data, identify different operating conditions of the combustion system, and determine the characteristic parameters of each operating condition, providing a basis for the selection and parameter adjustment of the ARX model.

[0128] The ARX model building module selects the appropriate ARX model for prediction, determines the model's input variables, time delay parameters and model parameters, builds the ARX model, and performs training and verification.

[0129] The prediction and control module inputs real-time data into the optimized ARX model to predict NOx emissions, outputs the prediction results, and controls and adjusts the combustion system based on the prediction results.

[0130] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for predicting NOx emissions from a gas turbine based on an ARX model, characterized in that: include, Based on the historical data of the gas turbine combustion system, the NOx generation mechanism is modeled, an ARX model with residuals is constructed, and the constructed model is time discretized; According to the input and output under each working condition in the combustion system, the parameter identification method is used to identify the time delay and identification parameters in the discretized ARX model; The historical NOx concentration under each operating condition is used as a validation set and input into the ARX model to obtain a prediction data set of NOx concentration, and the prediction accuracy of the data in the prediction set is determined; Use optimization algorithms to perform uncertainty analysis on the ARX model and evaluate the reliability and stability of the prediction results; The combustion system operating conditions are identified using pattern recognition technology, and the ARX model is improved based on the identification results.

2. A method for predicting NOx emissions from a gas turbine based on an ARX model as claimed in claim 1, characterized in that: The historical data includes the operating status of the gas turbine under different working conditions, including temperature, pressure, flow rate, speed; fuel-air mixing ratio, combustion temperature, combustion chamber pressure; historical NOx emission concentration data; fuel type, composition, calorific value; environmental conditions and parameters in the gas turbine control system; The data can be made consistent by removing duplicate records, correcting obvious errors, and processing outliers. At the same time, the data can be converted into a unified format and unit for comparison and analysis with data from different sources and types to make the data from different data sources consistent.

3. A method for predicting NOx emissions from a gas turbine based on an ARX model as claimed in claim 2, characterized in that: The construction of the ARX model with residuals includes extracting a formula describing the basic relationship between system variables according to the mechanism formula between the input variables and the output variables. The fuel amount m affects the combustion temperature. The influence of the combustion temperature on the NOx generation rate is in the form of exp(c / T). Then, the influence of the fuel amount on NOx through the temperature is exp(c1 / m+b). c1 and b are unknown parameters, which are called primitives. Determine candidate input variables for the model, consisting of each primitive, the coupling form between primitives, and time delays; use candidate variables and experimental or operational data to identify unknown parameters and determine the model structure: φ0(t)=1 Where y(t) represents the predicted NOx concentration, φ i (t) represents the input variable, k represents the time delay, u j represents the control variable, ε(t) represents the deviation, θ i represents the model parameters; The constructed model is subjected to time discretization processing, and the continuous-time model is converted into a discrete-time model. In the discrete-time model, a difference equation is used to replace a differential equation, and the integral is converted into a summation form. The discretized ARX model is expressed as: t z =zΔt Among them, t z represents a discrete time point, Δt is a sampling interval, z represents an integer index, and the discretized basis function is expressed as: Among them, k j and l are u j and the time delay parameters of y.

4. A method for predicting NOx emissions from a gas turbine based on an ARX model as claimed in claim 3, characterized in that: The identification of the time delay and identification parameters in the discretized ARX model includes, Where n represents the number of control variables, m represents the number of output variables, and λ represents the attenuation parameter; According to the structure of the ARX model, the ARX model is parameterized into an estimable form and the objective function is established: in, Represents the output predicted by the model, and N represents the number of data points; the finalized model is applied to the actual combustion system to predict the NOx generation concentration, and the input variables are adjusted as needed to achieve the desired NOx emission level.

5. A method for predicting NOx emissions from a gas turbine based on an ARX model as claimed in claim 4, characterized in that: The method for determining the prediction accuracy of the data in the prediction set includes collecting historical NOx concentration data under each operating condition as a validation set, cleaning and preprocessing the data, training the ARX model using the historical data, and identifying model parameters, including time delay parameters and model coefficients; inputting the data in the validation set into the trained ARX model, and calculating the predicted value of NOx concentration: Among them, P(t z ) represents the discrete time point t z The predicted value of NOx concentration; The data missing rate does not exceed 5%. Outliers need to be detected and processed by statistical methods. 80% of the historical data is used as the training set, and the remaining 20% ​​is used as the validation set. The root mean square error (RMSE) and the coefficient of determination (R) are used to determine the 2 Determine the accuracy of the forecast; When RMSE is less than 10ppm and R 2 When the RMSE is between 10-20ppm and R 2 When RMSE is between 0.7 and 0.85, the prediction accuracy is "medium". When RMSE is above 20 ppm or R 2 When it is lower than 0.7, the prediction accuracy is "low"; When the prediction accuracy is "medium" or "low", the model's input variables and time delay parameters are reconsidered, and the discretized time delay and identification parameters are re-identified based on the input and output quantities under each operating condition in the combustion system. When the prediction accuracy is "high", the model is used for actual prediction, verification and update.

6. A method for predicting NOx emissions from a gas turbine based on an ARX model as claimed in claim 5, characterized in that: The uncertainty analysis includes considering the cooling coefficient in the simulated annealing algorithm, the uncertainty parameters of the model, the temperature of the gas turbine, and the amount of fuel, performing uncertainty analysis of the ARX model through the simulated annealing algorithm in the optimization algorithm, and evaluating the reliability and stability of the gas turbine NOx emission prediction. Where R represents the evaluation result, λ1 represents the cooling coefficient in the simulated annealing algorithm, δ represents the uncertainty parameter of the model, T represents the temperature of the gas turbine, m' represents the amount of fuel, and p represents the number of model parameters; The iterative process of the simulated annealing algorithm is expressed as: in, represents the value of the i-th model parameter at the b-th iteration, It represents the update amount of the i-th model parameter at the b-th iteration, b represents the number of iterations, and the optimized model parameters are used to calculate the evaluation result R. If R is greater than 0.75 and less than 1 and infinitely tends to 1, it means that the prediction reliability is high; if R is less than or equal to 0.75, it means that the prediction uncertainty is high.

7. A method for predicting NOx emissions from a gas turbine based on an ARX model as claimed in claim 6, characterized in that: The improvement of the ARX model includes collecting the current operating data of the combustion system, monitoring and recording the data in real time, inputting the real-time data under the current working condition into the constructed ARX model, comparing the real-time data with the historical data, identifying the working mode of the current combustion system, Among them, F represents the recognition result of the combustion system operating mode, X t Indicates real-time data, X h Represents historical data, T d Indicates time delay; According to the prediction accuracy of the model, analyze the performance of the model under the current working conditions, and improve the ARX model based on the uncertainty analysis results and prediction accuracy evaluation; By comparing the predicted NOx emissions with the actual measured values, the model improvement process is initiated when the prediction error exceeds 5%; Determine the prediction stability by evaluating the change in prediction error of the model in different time periods. When the prediction stability is lower than 0.95, start the model improvement process; The frequency of model adjustment is determined by recording the number of times the model is adjusted. When the model needs to be adjusted within 3 consecutive days, the optimization algorithm is started; The recognition accuracy of the combustion system operating condition is determined by comparing the matching degree between the recognized operating condition mode and the actual operating condition. When the recognition accuracy is lower than 90%, the optimization algorithm is started.

8. A system using a gas turbine NOx emission prediction method based on an ARX model as claimed in any one of claims 1 to 7, characterized in that: It includes data acquisition and processing module, working condition identification module, ARX model building module, prediction and control module; The data acquisition and processing module collects key parameter data of the gas turbine during operation, cleans the collected data, removes noise and outliers, normalizes the data, and eliminates the influence of different dimensions on the model; The operating condition identification module uses pattern recognition technology to classify data, identify different operating modes of the combustion system, determine the characteristic parameters under each operating mode, and provide a basis for the selection and parameter adjustment of the ARX model; The ARX model building module selects a suitable ARX model for prediction, determines the model's input variables, time delay parameters and model parameters, builds the ARX model, and performs training and verification; The prediction and control module inputs real-time data into the optimized ARX model, performs NOx emission prediction, outputs prediction results, and controls and adjusts the combustion system according to the prediction results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images