Boiler oxygen amount self-adaptive optimization method and device and storage medium

Abstract

The invention discloses a boiler oxygen amount self-adaptive optimization method and device and a storage medium. According to the technical scheme, the method comprises the steps that a boiler efficiency model containing undetermined parameters is established; actual operation data are obtained in real time in the boiler operation process, and filtering preprocessing is conducted on the data; updating undetermined parameters of the efficiency model in real time according to the processed data by applying online identification algorithms such as a recursive least square method, and generating a real-time correction model capable of reflecting the current characteristics of the boiler; and finally, on the basis of the real-time correction model, calculating and outputting an optimal oxygen amount set value by solving a multi-objective optimization function giving consideration to energy-saving benefits and equipment loss. Preferably, the method further includes utilizing a predictive model to perform prospective control. Compared with the prior art, the method has the advantages that the accuracy, the robustness and the comprehensive economic benefit of oxygen control are remarkably improved through online self-adaption and prospective optimization of the model.

Description

Technical Field

[0001] The present invention relates to the technical field of boiler control, and in particular to a method and device for adaptively optimizing boiler oxygen content, and a computer-readable storage medium. Background Art

[0002] In industrial production processes such as thermal power generation and centralized heating, boilers are core energy conversion equipment, and their combustion efficiency is directly related to a company's energy consumption and economic benefits. The oxygen content in the boiler combustion process is a critical control parameter. Too low an oxygen content leads to incomplete fuel combustion, significant energy loss, and the generation of harmful substances such as black smoke and carbon monoxide. Excessive oxygen content removes a significant amount of heat, lowering exhaust gas temperatures, similarly reducing thermal efficiency, and potentially producing excessive nitrogen oxides. Therefore, precisely controlling the boiler's oxygen content at the optimal value is crucial for achieving energy conservation and emission reduction.

[0003] Existing methods for optimizing boiler oxygen levels typically employ techniques based on expert rules and big data analysis. These methods divide complex boiler systems into several control segments, develop control rules based on the experience of industry experts, and analyze historical operating data, combined with a pre-defined boiler efficiency model, to calculate the recommended oxygen level for the current load.

[0004] However, such existing technical solutions have the following inherent defects: 1. The boiler efficiency model is static. The actual operating characteristics of the boiler will change slowly but continuously with the aging and wear of the equipment, dust or coking on the heated surface, and real-time fluctuations in the quality of the fuel coal. The static efficiency model cannot track these changes, resulting in the gradual inaccuracy of the model. The "optimal" oxygen content calculated by it will deviate from the actual optimal point, and the optimization effect will decrease over time. 2. Insufficient adaptability and rigid control strategy. The reliance on historical big data makes it difficult for the system to effectively respond to new operating conditions or large disturbances that have not appeared in the historical database, which manifests as a delayed response. At the same time, expert rules are based on the summary of limited experience and are inherently rigid. They cannot cover all complex combustion scenarios, which limits the further exploration of optimization potential.

[0005] Therefore, how to overcome the static and poor adaptability problems of the efficiency model in the existing technology and develop a boiler oxygen optimization method that can adapt in real time, be forward-looking and more robust is a technical problem that needs to be solved urgently in this field. Summary of the Invention

[0006] The purpose of the present invention is to provide a method, device and storage medium for adaptive optimization of boiler oxygen content, so as to solve the technical problem in the background art that the boiler efficiency model is static and cannot adapt to changes in operating conditions, resulting in poor optimization effect.

[0007] To achieve the above-mentioned objectives, the present invention provides the following technical solutions: A method for adaptive optimization of boiler oxygen content, comprising the following steps: establishing a boiler efficiency model including pending parameters, wherein the boiler efficiency model is used to characterize the relationship between boiler efficiency and boiler input variables; obtaining the actual operating data of the boiler in real time during boiler operation; updating the pending parameters online based on the actual operating data and the boiler efficiency model to generate a real-time corrected efficiency model; and calculating and outputting the optimal oxygen content setting value under the current operating conditions based on the real-time corrected efficiency model.

[0008] Furthermore, the step of online updating of the pending parameters is specifically as follows: calculating the predicted output of the boiler efficiency model based on the estimated value of the pending parameters at the previous moment and the actual operating data at the current moment; comparing the actual efficiency measurement value of the boiler with the predicted output to obtain a prediction error; applying the recursive least squares method to update the pending parameters based on the prediction error, wherein the recursive least squares method uses a forgetting factor to adjust the weight of new and old data in parameter estimation.

[0009] Preferably, before the online updating of the pending parameters, a step of filtering and preprocessing the actual operating data is also included; the filtering and preprocessing step is specifically: applying a Kalman filter to process the sensor measurement values ​​containing noise to obtain the posterior state estimation values ​​of the boiler key state variables, and using the posterior state estimation values ​​for the online updating of the pending parameters.

[0010] More specifically, the step of applying the Kalman filter is an adaptive robust Kalman filtering method, including: calculating the new information sequence of the Kalman filter and its actual covariance within a preset time window; comparing the actual covariance with the theoretical covariance of the new information sequence; when the actual covariance does not match the theoretical covariance, online adjusting the measurement noise covariance matrix of the Kalman filter to suppress the influence of abnormal measurement data on the filtering results.

[0011] In particular, the method further includes a step of dynamically selecting a filtering strategy based on intelligent identification of operating conditions, the step including: using a support vector machine classifier to perform real-time identification of the current boiler operating condition type; according to the identified operating condition type, selecting and loading the corresponding process noise covariance matrix and measurement noise covariance matrix for the adaptive robust Kalman filter from a preset parameter library.

[0012] Furthermore, the step of calculating the optimal oxygen setting value based on the real-time corrected efficiency model is achieved by solving a multi-objective optimization function, which includes: a fuel cost function related to the coal consumption rate; and an actuator loss cost function related to the rate of change of the oxygen setting value.

[0013] Specifically, in the multi-objective optimization function, the weight coefficient between the fuel cost function and the actuator loss cost function is dynamically adjusted; the dynamic adjustment method of the weight coefficient is: using a fuzzy logic controller, based on at least one of the boiler's load change rate, electricity price and accumulated equipment operating time, real-time inference and output of the weight coefficient value of the actuator loss cost function.

[0014] Furthermore, the method is also a forward-looking optimization method. Before calculating the optimal oxygen setting value, it also includes: using a prediction model based on a long short-term memory network to predict the future load demand of the boiler based on historical operating data; and using the predicted future load demand as input for calculating the optimal oxygen setting value.

[0015] Preferably, the prediction model is a multivariate input model, whose input, in addition to historical load data, also includes at least one variable that can indirectly reflect changes in coal quality; the method also includes online soft measurement estimation of the calorific value of coal in the short term in the future based on the energy conservation equation, and uses the calorific value estimate together with the predicted future load demand to calculate the optimal oxygen setting value.

[0016] As an optional implementation, the step of calculating the optimal oxygen content setting value based on the real-time corrected efficiency model is specifically as follows: in the offline stage, based on a high-fidelity boiler model, optimization calculations are performed for multiple assumed operating conditions, and the calculation results are trained into an agent model; in the online stage, real-time operating condition data is input into the agent model, and the agent model quickly outputs the optimal oxygen content setting value.

[0017] Based on the above optional implementation, the method also includes a step of automatically updating the proxy model, which includes: online monitoring of a performance degradation index between the output of the proxy model and a true optimal solution calculated based on the real-time corrected efficiency model; when the performance degradation index exceeds a preset threshold, automatically triggering the retraining and deployment of the proxy model using the latest real-time corrected efficiency model.

[0018] To achieve the above-mentioned objectives, the present invention also provides a boiler oxygen content adaptive optimization device, comprising: a model establishment module for establishing a boiler efficiency model including undetermined parameters; a data acquisition module for acquiring the actual operating data of the boiler in real time during the operation of the boiler; a parameter updating module for updating the undetermined parameters online based on the actual operating data and the boiler efficiency model to generate a real-time corrected efficiency model; and an optimization calculation module for calculating and outputting the optimal oxygen content set value under the current operating conditions based on the real-time corrected efficiency model.

[0019] Furthermore, the device further comprises: a feedforward prediction module, configured to predict the future load demand of the boiler before the optimization calculation module calculates the optimal oxygen content setting value, and provide the predicted future load demand to the optimization calculation module.

[0020] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements any of the aforementioned methods when executed by a processor.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] 1. By updating the pending parameters of the efficiency model online, the present invention enables the optimization model to track changes in boiler operating characteristics in real time, overcoming the fundamental flaws of static models in the prior art and ensuring that the optimal oxygen setpoint calculated at any time is closer to the actual optimal point, thereby significantly improving energy savings.

[0023] 2. By introducing adaptive robust filtering and intelligent identification of working conditions, the present invention can effectively filter out sensor noise, suppress abnormal data impact, and adopt optimal filtering strategies for different working conditions, greatly enhancing the stability and reliability of the system under complex, changeable, and non-ideal working conditions.

[0024] 3. By introducing the coordinated prediction of load and coal quality based on the long-short-term memory network, the present invention upgrades the control mode from "reactive" to "forward-looking", which can foresee future operating condition changes and perform "pre-control" in advance, effectively reducing control lag, and achieving better energy-saving and environmental protection effects, especially under variable load conditions.

[0025] 4. By constructing a multi-objective optimization function that includes equipment loss costs and using fuzzy logic to dynamically adjust the strategy, the present invention avoids excessive loss of actuators while pursuing energy conservation, extends the life of the equipment, and achieves the comprehensive optimization of operating economic benefits and long-term equipment health. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solution of the present invention, the accompanying drawings are briefly described below:

[0027] Figure 1 Schematic diagram of the overall process of a method for adaptively optimizing boiler oxygen content according to the present invention;

[0028] Figure 2 Schematic diagram of the data filtering and model updating sub-process in the method of the present invention;

[0029] Figure 3 Schematic diagram of the process of forward-looking optimization control in the method of the present invention;

[0030] Figure 4 This is a functional module diagram of an embodiment of a boiler oxygen adaptive optimization device of the present invention. DETAILED DESCRIPTION

[0031] To make the purpose, technical solutions and advantages of the present invention more clear, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the scope of protection of the present invention.

[0032] Example 1

[0033] This embodiment provides a method for adaptively optimizing the oxygen content of a boiler. Figure 1 , this method is an overall process of the present invention, and can specifically include the following steps:

[0034] Step S101: Establish an initial boiler efficiency model. First, establish a mathematical model that describes the relationship between boiler combustion efficiency and other key variables. This model contains a set of undetermined parameters that will be updated online in subsequent steps. A simplified linear regression model can be expressed as:

[0035] in, is a discrete time point; For The system output at that moment, specifically, the actual boiler efficiency can be calculated by measuring parameters such as steam flow, feed water temperature, and fuel consumption; It is by An observation data vector consisting of multiple boiler input variables (such as coal feed rate, primary air volume, secondary air volume, oxygen set value, etc.) measured at all times; is the unknown parameter vector that characterizes the boiler characteristics.

[0036] Step S102: Acquire and pre-process real-time operation data. In each control cycle of boiler operation, various sensor data of the boiler are acquired through the data acquisition system to form an observation data vector and actual efficiency measurements Preferably, in order to improve the stability and accuracy of subsequent model updates, the raw data are filtered and preprocessed before being used for calculation. Figure 2 , this pre-processing step can use Kalman filter. First, establish the state space model of key state variables (such as exhaust temperature, steam pressure, etc.): ,

[0037] in, is the real value vector of the system state; is the sensor measurement value vector; is the control input vector; , , They are the state transfer matrix, control input matrix and observation matrix respectively; and They are process noise and measurement noise, which are usually assumed to be Gaussian white noise, and their covariances are and The standard prediction and update steps of the Kalman filter can be used to obtain the noisy measurements. A smoother state estimate that is closer to the true value is obtained .

[0038] Furthermore, to cope with sudden disturbances of non-Gaussian nature (e.g., large pieces of wet coal entering the furnace causing a sudden change in measurement data), an adaptive robust Kalman filter can be used. This method monitors the covariance of the filter's innovation sequence (i.e., the difference between the measured value and the predicted value) in real time and compares it with the theoretical covariance. If there is a significant deviation between the two, indicating that the current noise characteristics are abnormal, the system will automatically increase the measurement noise covariance matrix. The value of , thereby reducing the weight of the current abnormal measurement data in the state update, achieving the purpose of robust filtering. In particular, a working condition intelligent identification module based on support vector machine (SVM) can also be introduced. This module analyzes the pattern of a set of sensor data and classifies the current working condition into "normal", "coking", "wet coal", etc., and then selects a set of optimal covariance matrices ( and ), to achieve precise and customized filtering of specific disturbances.

[0039] Step S103: Update the efficiency model parameters online. Use the high-quality data pre-processed in step S102 to update the efficiency model parameters vector. Perform online update. Specifically, the recursive least squares method (RLS) with forgetting factor can be used. First, calculate the prediction error :

[0040] in, is the actual efficiency after filtering, is the estimated value of the parameter vector at the previous moment. Then, the parameters are updated according to the following recursive formula: , ,

[0041] in, is the gain matrix; is the parameter covariance matrix; is the forgetting factor ( ), which is used to adjust the influence of new and old data. Through this step, the efficiency model (consisting of parameter vector definition) can be corrected in real time to accurately reflect the current characteristics of the boiler.

[0042] Step S104: Calculate and output the optimal oxygen setting value. Based on the real-time corrected efficiency model obtained in step S103, calculate the optimal oxygen setting value under the current working conditions. Preferably, this step adopts a forward-looking control strategy and considers multi-objective optimization. Figure 3 First, a prediction model based on the long short-term memory network (LSTM) is used to predict the load demand in a short time window in the future based on historical load, exhaust temperature and other data. and coal quality change trends (such as equivalent calorific value Then, these predicted values ​​are used as input to solve a multi-objective optimization problem. The objective function of this problem is It can be defined as:

[0043] Among them, the first term represents the fuel cost, is the efficiency function of real-time correction, and its negative value indicates maximum efficiency; the second term is the penalty term, which represents the loss cost of actuators such as dampers, and is used to suppress the oxygen set value The weight and It is used to balance energy saving benefits and equipment health. More preferably, the weight It can be dynamically adjusted by a fuzzy logic controller. The controller takes the current load fluctuation severity, market electricity price, etc. as input and adjusts the The value of is used to achieve a more intelligent decision balance. The numerical optimization algorithm (such as gradient descent method) is used to solve the problem of smallest , which is the optimal oxygen setting value and output it to the boiler's execution control system.

[0044] As an optional implementation method, in order to solve the problem of time-consuming online calculation of complex models, step S104 can adopt an offline-online hybrid optimization strategy. Offline, high-performance computers are used to optimize tens of thousands of hypothetical working conditions based on high-fidelity models, and the results ("working condition-optimal oxygen content" data pairs) are trained into a proxy model with extremely fast calculation speed (such as a deep neural network). When running online, the system only needs to input the predicted working conditions into the proxy model to instantly obtain an approximate optimal solution. To ensure that the proxy model is not "outdated", the system is also equipped with an automatic update mechanism. When it is detected that the output of the proxy model deviates too much from the calculation result of the latest high-fidelity model, the background process will be automatically triggered to retrain and deploy the proxy model with the latest high-fidelity model.

[0045] Example 2

[0046] This embodiment provides a boiler oxygen quantity adaptive optimization device, the structure of which is as follows: Figure 4 As shown. The device can be a computing server or dedicated controller integrated into a boiler distributed control system (DCS). The device includes: a model building module 401, which is used to establish a boiler efficiency model including undetermined parameters during system initialization, as described in formula ([eq:model]) in Example 1.

[0047] The data acquisition module 402 is used to acquire various sensor data of the boiler in real time during the operation of the boiler by communicating with the field bus or DCS system, such as coal feed rate, air volume, steam parameters, exhaust gas temperature, etc., to form actual operation data.

[0048] Parameter update module 403 is connected to model building module 401 and data acquisition module 402. This module is used to execute the filtering and parameter update logic described in Example 1. Specifically, it may include a filtering submodule (preferably an adaptive robust Kalman filter) and a parameter identification submodule (preferably an RLS algorithm). Based on the actual operating data obtained, it updates the undetermined parameters of the efficiency model online, thereby generating a real-time corrected efficiency model.

[0049] The optimization calculation module 404 is connected to the parameter updating module 403. This module is used to calculate and output the optimal oxygen content setting value under the current working conditions based on the real-time corrected efficiency model output by the parameter updating module 403.

[0050] Preferably, the device also includes a feedforward prediction module 405, connected to the data acquisition module 402 and the optimization calculation module 404. This module deploys a prediction model, such as an LSTM, to predict the boiler's future load demand and coal quality trends based on historical operating data. The prediction results are sent to the optimization calculation module 404 for proactive control. The optimization calculation module 404 internally executes the multi-objective optimization algorithm described in Example 1, or deploys a proxy model for rapid decision-making.

[0051] In actual operation, the data acquisition module 402 continuously collects data, the feedforward prediction module 405 predicts future operating conditions, the parameter update module 403 continuously adjusts the efficiency model, and the optimization calculation module 404 calculates the optimal oxygen setpoint based on the latest model and prediction results, and sends it to the boiler's underlying control loop for execution. These modules work together to form a closed-loop adaptive optimization system.

[0052] Example 3

[0053] This embodiment provides a computer-readable storage medium. This computer-readable storage medium can be non-volatile, such as a hard disk, solid-state drive (SSD), flash memory, or optical disk (CD, DVD, etc.), or volatile, such as random access memory (RAM). This computer-readable storage medium stores executable computer program instructions. When these instructions are loaded and executed by one or more processors (e.g., a central processing unit (CPU), a digital signal processor (DSP), or an application-specific integrated circuit (ASIC), the processors can implement all or part of the steps of the boiler oxygen content adaptive optimization method described in Example 1. For example, the program instructions are organized into code segments corresponding to the various functional modules in Example 2 (model building module 401, data acquisition module 402, parameter updating module 403, optimization calculation module 404, and feedforward prediction module 405). Executing these code segments implements the corresponding functions, thereby completing the entire optimization process.

[0054] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should be considered to fall within the scope of protection of the present invention.

Claims

1. A method for adaptively optimizing boiler oxygen content, characterized in that: The method comprises the following steps: establishing a boiler efficiency model including undetermined parameters, wherein the boiler efficiency model is used to characterize the relationship between boiler efficiency and boiler input variables; During the operation of the boiler, actual operation data of the boiler is obtained in real time; updating the undetermined parameters online according to the actual operation data and the boiler efficiency model to generate a real-time corrected efficiency model; Based on the real-time corrected efficiency model, the optimal oxygen content setting value under the current working conditions is calculated and output.

2. The method according to claim 1, characterized in that The step of online updating of the pending parameters specifically comprises: calculating the predicted output of the boiler efficiency model according to the estimated value of the pending parameters at the previous moment and the actual operation data at the current moment; The actual efficiency measurement value of the boiler is compared with the predicted output to obtain a prediction error; and a recursive least squares method is applied to update the undetermined parameter according to the prediction error, wherein the recursive least squares method uses a forgetting factor to adjust the weight of new and old data in parameter estimation.

3. The method according to claim 1 or 2, characterized in that Before the online updating of the pending parameters, the step of filtering and preprocessing the actual operating data is also included; the filtering and preprocessing step is specifically: applying a Kalman filter to process the sensor measurement values ​​containing noise to obtain the posterior state estimation values ​​of the boiler key state variables, and using the posterior state estimation values ​​for the online updating of the pending parameters.

4. The method according to claim 3, characterized in that The step of applying the Kalman filter is an adaptive robust Kalman filtering method, including: calculating the new information sequence of the Kalman filter and its actual covariance within a preset time window; comparing the actual covariance with the theoretical covariance of the new information sequence; when the actual covariance does not match the theoretical covariance, online adjusting the measurement noise covariance matrix of the Kalman filter to suppress the influence of abnormal measurement data on the filtering result.

5. The method according to claim 4, characterized in that It further includes the step of dynamically selecting a filtering strategy based on intelligent identification of operating conditions, the step comprising: using a support vector machine classifier to perform real-time identification of the current boiler operating condition type; according to the identified operating condition type, selecting and loading the corresponding process noise covariance matrix and measurement noise covariance matrix for the adaptive robust Kalman filter from a preset parameter library.

6. The method according to claim 1, characterized in that The step of calculating the optimal oxygen setting value based on the real-time corrected efficiency model is achieved by solving a multi-objective optimization function, which includes: a fuel cost function related to the coal consumption rate; and an actuator loss cost function related to the rate of change of the oxygen setting value.

7. The method according to claim 6, characterized in that In the multi-objective optimization function, the weight coefficient between the fuel cost function and the actuator loss cost function is dynamically adjusted; the dynamic adjustment method of the weight coefficient is: using a fuzzy logic controller, based on at least one of the boiler's load change rate, electricity price and accumulated equipment operating time, real-time inference and output of the weight coefficient value of the actuator loss cost function.

8. A boiler oxygen content adaptive optimization device, characterized in that: include: a model building module for building a boiler efficiency model including undetermined parameters, wherein the boiler efficiency model is used to characterize the relationship between boiler efficiency and boiler input variables; A data acquisition module, used to acquire the actual operation data of the boiler in real time during the operation of the boiler; a parameter updating module, configured to update the undetermined parameters online based on the actual operating data and the boiler efficiency model to generate a real-time corrected efficiency model; An optimization calculation module is used to calculate and output the optimal oxygen content setting value under the current working conditions based on the real-time corrected efficiency model.

9. The device according to claim 8, characterized in that Also includes: A feedforward prediction module is used to predict the future load demand of the boiler before the optimization calculation module calculates the optimal oxygen setting value, and provide the predicted future load demand to the optimization calculation module.

10. A computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the method according to any one of claims 1 to 7 when executed by a processor.

Images