CN-121978272-A - Dynamic calibration method for oxygen content in boiler air-smoke system

CN121978272ACN 121978272 ACN121978272 ACN 121978272ACN-121978272-A

Abstract

The invention belongs to the technical field of oxygen detection and dynamic calibration, and particularly relates to a dynamic calibration method of oxygen in a boiler flue gas system. The method comprises the steps of multi-working condition data acquisition and standardization processing, failure feature extraction and feature library establishment, reference oxygen amount calculation model establishment, data driving model establishment, anomaly detection and calibration, parameter self-updating and the like, and a new calibration effect real-time evaluation link is added. The method is characterized in that a 'mechanism+data' dual-drive mode is adopted, a reference model is built on the basis of a Nernst equation of a zirconia sensor, a data-driven model is built by fusing an improved bidirectional LSTM network, and the accurate identification and dynamic calibration of abnormal signals are realized by combining a self-adaptive threshold algorithm and a multi-stage calibration strategy. The method is adaptive to complex smoke environment and variable load operation scene, remarkably improves oxygen measurement precision, ensures low-nitrogen combustion and environment-friendly compliance, reduces coal consumption and operation cost, and provides reliable data support for digital transformation of intelligent power plants.

Inventors

HE SHANHONG
ZHENG GUIBO
LV DESHENG
LOU JIE
TANG TIAN
Si Shuaiqi

Assignees

广东粤电靖海发电有限公司

Dates

Publication Date: 20260505
Application Date: 20260108

Claims (10)

1. The dynamic calibration method for the oxygen content in the boiler air smoke system is characterized by comprising the following steps of: collecting zirconium battery output signals, smoke parameters and boiler operation data under different loads, coal types and operation environments, preprocessing the collected data, and obtaining a standardized multi-working-condition data set; the failure feature extraction and feature library establishment, namely, based on the standardized multi-task data set, extracting static statistical features and dynamic failure features, and constructing an oxygen sensor failure feature library of a related working condition type-failure feature-drift mode; Constructing a reference oxygen amount calculation model, namely substituting a Nernst equation of a zirconia sensor into the flue gas temperature, the pressure and the working temperature parameters of the zirconium battery in the standardized multi-working-condition data set, fusing a temperature correction coefficient and a pressure correction coefficient to construct the reference oxygen amount calculation model, and outputting a reference oxygen amount under a static working condition; constructing a data driving model by using a time sequence learning algorithm, taking the standardized multi-working-condition data set, a failure characteristic library and a reference oxygen value as model input, training the model by using time sequence characteristics of load fluctuation and smoke temperature change to capture nonlinear influence, and outputting a real-time corrected oxygen value; The abnormal detection and calibration are carried out by taking the real-time corrected oxygen value as a reference, adopting a self-adaptive threshold algorithm to identify abnormal signals of the oxygen sensor, combining the failure feature library to match a drift mode, triggering a corresponding type of calibration signal according to the operation working condition of the boiler, and implementing automatic calibration; and the parameter self-updating is to automatically adjust the calibration period and the correction coefficient according to the real-time working condition data.
2. The method for dynamically calibrating oxygen in a boiler flue gas system according to claim 1, further comprising a step of evaluating the effect of calibration in real time by calculating a relative deviation ratio of oxygen values before and after calibration: wherein In order to calibrate the value of the pre-oxygen, For the oxygen value after calibration And when the calibration is abnormal, judging the calibration and triggering a secondary calibration flow.
3. The method for dynamically calibrating oxygen in a boiler flue gas system according to claim 1, wherein the static statistical features comprise a mean μ, a variance σ2, a skewness S and a kurtosis K, and the calculation formula is: Wherein, the Is the first The filtered zirconium battery output signal values at various time points, Calculating a window length for the static feature; the dynamic failure feature includes drift rate Frequency of fluctuation Autocorrelation coefficient The calculation formula is as follows: Wherein, the In order to drift the length of the statistical time window, Representing the fourier transform of the signal, Is a static mean value of the values, Calculating window length for autocorrelation, and feature extraction using adaptive sliding window control algorithm Dynamically adjust to , For the initial window length to be the same, In order for the attenuation coefficient to be a factor, In order to achieve a rate of change of the load of the boiler, , For the load factor of the boiler, Is a time interval.
4. The method for dynamically calibrating oxygen in a boiler flue gas system according to claim 1, wherein the reference oxygen amount calculation model is constructed by fusing Nernst equation with real-time working condition parameters, and the core formula is as follows: Wherein, the As a result of the ideal gas constant, Is a function of the faraday constant, For the temperature correction coefficient, the formula is calculated: is a coefficient of temperature linearity which is a function of the temperature, Is a temperature secondary coefficient, and the temperature secondary coefficient is the temperature secondary coefficient, For the zirconium battery operating temperature (K), Is the reference temperature; for the pressure correction factor, the formula is calculated: As a result of the pressure correction factor, Measuring pressure for sensor, inputting parameters for model Is collected in real time by a thermocouple array, As measured by the pressure sensor(s), And Dynamically correcting by an adaptive PID controller, and outputting by the controller: As an error signal, the signal is a signal, 、、 As a function of the PID parameters, , , 。
5. The method for dynamically calibrating oxygen in a boiler flue gas system according to claim 1, wherein the data driving model adopts an improved bidirectional LSTM network structure and comprises two layers of LSTM units, an input layer comprises a standardized multi-working-condition data set, a failure feature library and a reference oxygen value, an output layer generates a real-time corrected oxygen value, and model training adopts an Adam optimization algorithm and a learning rate The loss function is the weighted mean square error: Wherein, the The true oxygen label value for the ith training sample, As the weight coefficient of the sample, As a coefficient of complexity of the operating mode, , 、 As the characteristic weight of the object to be processed, For the load factor of the boiler, In order for the maximum load factor to be the highest, Is the temperature of the flue gas and is equal to the temperature of the flue gas, For the maximum temperature of the flue gas, Is the attenuation coefficient.
6. The method for dynamically calibrating oxygen in a boiler flue gas system according to claim 1, wherein the adaptive threshold algorithm adopts a dynamic adjustment mechanism, and a threshold calculation formula is: Wherein, the To correct an exponentially weighted moving average of oxygen values in real time, , In order to smooth the coefficient of the coefficient, Is the current reference oxygen value; Correcting the standard deviation of the oxygen value in real time; in order for the coefficients to be adaptive, , As the reference coefficient of the reference value, In order to adjust the coefficient of the coefficient, Is the minimum load factor; in order for the drift rate to be high, The algorithm realizes abnormality detection through an adaptive threshold controller, and the controller outputs , , To adapt to the coefficients, the threshold dynamically expands upon abrupt load changes.
7. The method for dynamically calibrating oxygen in a boiler flue gas system according to claim 1, wherein in said step of self-updating parameters, the calibration period is defined The dynamic adjustment formula of (2) is: Wherein, the For the current calibration period, For the initial calibration period to be a period of time, In order for the drift rate to be high, For the current boiler load rate, In order for the maximum load factor to be the highest, The value range is 0.05-0.2 for attenuation coefficient.
8. The method for dynamically calibrating oxygen in a boiler flue gas system according to claim 1, wherein the calibration signal triggering conditions are: Wherein, the For the real-time measurement of the sensor, As the reference oxygen value, the oxygen value, In order to adapt the threshold value to be used, In order for the drift rate to be high, As a threshold value for the drift rate, For the current boiler load rate, In order for the load to be at a minimum, In order for the maximum load factor to be the highest, The value range of the load factor triggering coefficient is 0.7-0.9.
9. The method for dynamically calibrating the oxygen content in the boiler flue gas system according to claim 1, wherein after the calibration signal is triggered, a multistage calibration strategy is executed, wherein the multistage calibration strategy comprises two modes of quick response calibration and high-precision calibration, the quick response calibration is used for a frequent load fluctuation scene, instantaneous correction is achieved based on sliding window historical data interpolation, the high-precision calibration is used for a steady-state operation scene, and the sensor output is optimized by adopting an iterative least square method.
10. The method for dynamically calibrating oxygen in a boiler flue gas system according to claim 1, wherein the failure feature library comprises an online updating mechanism, wherein the feature weight is updated by using an incremental learning algorithm based on new acquired data, and the updating formula is as follows: Wherein, the As the weight of the original characteristic is given, In order to update the post-weight of the weight, The feature variation calculated for the new data, Is a forgetting factor.

Description

Dynamic calibration method for oxygen content in boiler air-smoke system Technical Field The invention belongs to the technical field of oxygen detection and dynamic calibration, and particularly relates to a dynamic calibration method of oxygen in a boiler flue gas system. Background The oxygen measurement precision in the boiler air-smoke system is a core link for guaranteeing combustion efficiency and controlling pollutant emission, and is directly related to the economical efficiency, environmental protection and safety of boiler operation. However, the existing boiler oxygen calibration technology has the remarkable limitations that the traditional offline calibration and fixed period calibration modes lack of real-time performance, cannot adapt to complex working conditions such as variable load operation of a boiler, coal switching, flue gas parameter fluctuation and the like, so that a sensor is easy to drift and misalign, measurement deviation is difficult to dynamically correct, meanwhile, a single mechanism model is limited by working condition adaptability, and the problem of insufficient interpretability exists in a pure data driving model, so that accurate compensation of dynamic deviation cannot be realized. Under the background of the intensive promotion of the 'double carbon' strategy and the increasingly strict environmental protection regulations, the 'emission standard of atmospheric pollutants of boilers' clearly requires the ultralow emission modified coal-fired boilerThe emission is less than or equal to 50mg/m 3, and the accurate oxygen amount control becomes a key premise for realizing low-nitrogen combustion and meeting the emission limit value. If the oxygen amount is misaligned, the efficiency of the denitration system is reduced, ammonia escape exceeds standard, enterprises face severe environmental protection penalties, insufficient combustion is caused, and energy consumption and pollutant emission are increased. In addition, according to measurement and calculation, the oxygen measurement accuracy is improved by 1%, the boiler efficiency can be improved by 0.3% -0.5%, for a 600MW coal-fired unit, the annual coal saving amount can reach thousands of tons, the fuel cost is reduced by millions of yuan, and the energy saving and consumption reduction requirements are urgent. With the continuous promotion of construction of intelligent power plants and industrial Internet, the boiler system needs to be converted into digital and intelligent, and reliable bottom data is the basis for effective operation of a combustion optimization algorithm. The hysteresis and model limitation of the existing calibration technology cannot meet the multiple requirements of variable working condition operation, environment protection up to standard, energy saving, consumption reduction and intelligent upgrading of the boiler, and the research and development of a dynamic calibration method adapting to complex working conditions has extremely high reality urgency. Disclosure of Invention The method aims to solve the problems of poor instantaneity, insufficient working condition adaptability, model limitation and the like of the traditional boiler oxygen calibration technology, realizes dynamic and accurate oxygen calibration through a mechanism and data dual-drive mode, adapts to complex operation scenes, ensures environment protection compliance and energy conservation and consumption reduction, and provides reliable data support for digital transformation of intelligent power plants. In order to achieve the above object, the present invention adopts the following technical scheme: the dynamic calibration method of the oxygen content in the boiler air smoke system comprises the following steps: collecting zirconium battery output signals, smoke parameters and boiler operation data under different loads, coal types and operation environments, preprocessing the collected data, and obtaining a standardized multi-working-condition data set; the failure feature extraction and feature library establishment, namely, based on the standardized multi-task data set, extracting static statistical features and dynamic failure features, and constructing an oxygen sensor failure feature library of a related working condition type-failure feature-drift mode; Constructing a reference oxygen amount calculation model, namely substituting a Nernst equation of a zirconia sensor into the flue gas temperature, the pressure and the working temperature parameters of the zirconium battery in the standardized multi-working-condition data set, fusing a temperature correction coefficient and a pressure correction coefficient to construct the reference oxygen amount calculation model, and outputting a reference oxygen amount under a static working condition; constructing a data driving model by using a time sequence learning algorithm, taking the standardized multi-working-condition data set, a failure characteristic library and