Combustion stability control method and system based on coal-fired boiler

Abstract

The invention relates to the technical field of coal-fired boiler control, in particular to a control method and system based on combustion stability of a coal-fired boiler. The method comprises the steps that boiler process and physical property parameters are synchronously collected through a multi-source sensing module, and multi-dimensional real-time data are obtained; comprehensively calculating a stability index through edge side processing so as to judge a combustion stability level; based on the grade and the physical property parameter, calling a combustion characteristic mapping relation to match the basic parameter, and fusing model predictive control and an intelligent optimization algorithm to generate a multi-dimensional cooperative control instruction; issuing the instruction to an execution mechanism for driving adjustment, and acquiring feedback data in real time and comparing the feedback data with a target value to generate a result; and on the basis of a comparison result, updating model core parameters through an online learning algorithm. Through data fusion, intelligent decision making and online self-adaption, accurate sensing and optimal control of the combustion state are achieved, and the stability, efficiency and environmental protection performance of boiler operation are effectively improved.

Description

Technical Field

[0001] This invention relates to the fields of industrial process control and energy technology, and in particular to a method and system for controlling the combustion stability of coal-fired boilers. Background Technology

[0002] As a crucial energy conversion device, the combustion stability of coal-fired boilers directly affects unit operating efficiency, pollutant emissions, and equipment safety. Traditional control methods rely primarily on feedback regulation using a few process parameters, which has significant limitations. Existing technologies suffer from the following shortcomings: First, control strategies fail to adequately consider the dynamic changes in fuel properties, leading to a significant deterioration in control effectiveness when coal quality fluctuates. Second, control models are mostly statically designed, unable to adaptively adjust based on equipment status and operating conditions. Third, the lack of an effective fault-tolerant control mechanism during the coordinated operation of multiple boilers means that a single boiler instrument failure can impact the stability of the entire system. Finally, traditional PID control exhibits hysteresis, making it difficult to handle the nonlinear and high-inertia characteristics of the combustion process.

[0003] Therefore, there is an urgent need in this field for a method and system for controlling the combustion stability of coal-fired boilers that can integrate multi-source information, have forward-looking prediction capabilities, and self-learning characteristics. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for controlling the combustion stability of coal-fired boilers, so as to solve the problems of control lag, rigid models and lack of fault tolerance mechanism in the prior art.

[0005] In some embodiments of this application, a method for controlling the combustion stability of a coal-fired boiler is provided, characterized by comprising:

[0006] Multi-dimensional real-time monitoring data is obtained through a multi-source sensing module, including process parameters and physical property parameters.

[0007] The process parameters and physical property parameters are subjected to edge processing to obtain a stability index characterizing the combustion state, and the combustion stability level is determined based on the stability index.

[0008] Based on the pre-constructed combustion characteristic mapping relationship, combustion stability level and the physical property parameters, basic control parameters are generated. The combustion state in the future period is predicted based on the model predictive control algorithm, and multi-dimensional collaborative control commands are generated in combination with the intelligent optimization algorithm.

[0009] The multi-dimensional collaborative control command is sent to each actuator to drive each actuator to perform adjustment actions. Combustion state data after the adjustment actions is collected in real time and compared with preset target data to generate comparison results.

[0010] Based on the comparison results and the online learning algorithm, the combustion characteristic mapping relationship and the core parameters of the model predictive control algorithm are updated.

[0011] In some embodiments of this application, when obtaining multi-dimensional real-time monitoring data through a multi-source sensing module, and the multi-dimensional real-time monitoring data includes process parameters and physical property parameters, the method includes:

[0012] A measurement group of multi-source sensing modules is built on each coal-fired boiler, in which two sets of acquisition devices with the same performance are configured for key process parameter acquisition points.

[0013] Based on the aforementioned measurement group, when the system contains multiple coal-fired boilers, the central control unit sends a unified synchronous acquisition command to all measurement groups to generate the original dataset.

[0014] In the original dataset, the coal-fired boiler system monitors the output of each acquisition device in real time. When the data of any unit continuously deviates from the median range of its hot standby unit and is determined to be invalid, the mean of its hot standby unit is used as the valid output to generate the dataset.

[0015] The dataset is compared with corresponding data from other measurement groups in the central processing unit to identify the consistency of key features, thereby generating cross-validation results.

[0016] Based on the cross-validation results, data from different measurement groups are fused and weighted to generate the multi-dimensional real-time monitoring data.

[0017] In some embodiments of this application, the step of performing edge-side processing on the process parameters and physical property parameters to obtain a stability index characterizing the combustion state, and determining the combustion stability level based on the stability index, includes:

[0018] The local stability index is calculated using edge computing based on real-time data from a single measurement group.

[0019] The key features of the real-time data of the single measurement group are compared in real time with the group feature interval formed by all other normally functioning measurement groups in the system, and a group consistency index is generated to characterize the degree of data deviation.

[0020] A stability index is generated by performing a weighted fusion calculation based on the local stability index and the group consistency index.

[0021] The stability level of the current combustion is determined based on the numerical range of the stability index.

[0022] In some embodiments of this application, when generating basic control parameters based on pre-constructed combustion characteristic mapping relationships, combustion stability levels, and the physical property parameters, predicting the combustion state in future time periods based on model predictive control algorithms, and generating multi-dimensional collaborative control commands in conjunction with intelligent optimization algorithms, the process includes:

[0023] Based on the combustion stability level and the physical property parameters, a pre-built combustion characteristic mapping relationship is invoked to match the basic control parameters;

[0024] Based on the aforementioned basic control parameters, the combustion state in future time periods is predicted and the target is optimized by the model predictive control algorithm to generate a feedforward control sequence.

[0025] The feedforward control sequence and the stability index of real-time feedback are fused and calculated using an intelligent optimization algorithm to generate the multi-dimensional collaborative control command.

[0026] In some embodiments of this application, when the basic control parameters are matched by calling a pre-constructed combustion characteristic mapping relationship based on the combustion stability level and the physical property parameters, the following steps are included:

[0027] Construct a dynamic multi-dimensional similarity evaluation function to calculate the similarity distance between the current boiler and all other normally operating boilers in the system in the multi-dimensional operating space in real time;

[0028] Based on the similarity distance, the top K normal boilers that are most similar to the current boiler are selected to form a dynamic boiler group with similar operating conditions.

[0029] Based on the physical property parameters and corresponding control parameters provided by the boiler group under similar operating conditions, which have been cross-validated as reliable, alternative control parameters suitable for the current boiler are generated by mapping through an interpolation algorithm.

[0030] The alternative control parameters are input into a pre-built rapid combustion stability assessment model for calculation and verification;

[0031] If the predicted stability index of its output is higher than the preset safety threshold, then the alternative control parameter is adopted as the basic control parameter.

[0032] Otherwise, the weights of the similarity evaluation function are readjusted, and a new round of calculation is performed.

[0033] In some embodiments of this application, the step of predicting the combustion state in future time periods and performing target optimization based on the basic control parameters to generate a feedforward control sequence includes:

[0034] A feedforward control parameter window is constructed using the current basic control parameters as the initial sequence.

[0035] The feedforward control parameter window is input into the prediction model to obtain a prediction sequence of key combustion state parameters for the same future time period;

[0036] Construct a multi-objective optimization function, and optimize the sequence values ​​in the feedforward control parameter window under the preset conditions, with the multi-objective optimization function as the objective.

[0037] The first control parameter of the control parameter window obtained from each solution is used as the actual instruction to be executed, thereby generating the feedforward control sequence.

[0038] In some embodiments of this application, the step of fusing the feedforward control sequence and the stability index of real-time feedback through an intelligent optimization algorithm to generate the multi-dimensional collaborative control command includes:

[0039] Based on the aforementioned feedforward control sequence and the stability index of real-time feedback, a real-time dynamic optimization problem is constructed.

[0040] Solve the real-time dynamic optimization problem to generate a set of preliminary control commands;

[0041] The preliminary control commands are compared with the physical adjustment limits of each actuator to generate a feasibility verification result;

[0042] Based on the feasibility verification results, the preliminary control instructions are modified and quantified, and finally the multi-dimensional collaborative control instructions are generated and output to each actuator.

[0043] In some embodiments of this application, the step of issuing the multi-dimensional collaborative control command to each actuator, driving each actuator to perform adjustment actions, collecting combustion state data after the adjustment actions in real time, comparing it with preset target data, and generating comparison results includes:

[0044] The multi-dimensional collaborative control commands are synchronously sent to the burner regulating mechanism, the combustion stabilization auxiliary device and the air-coal regulating mechanism, driving each actuator to coordinate and execute the regulating actions, and generating a coordinated action execution status.

[0045] Based on the execution status of the coordinated action, process parameters reflecting changes in combustion status are collected in real time, and combined with multi-dimensional real-time monitoring data, a real-time status dataset is generated.

[0046] Based on the real-time status dataset, the real-time comprehensive stability index is calculated, and pollutant emission concentration data and combustion efficiency data are obtained to generate an evaluation parameter set that includes stability, emissions and efficiency.

[0047] The comprehensive stability index of the evaluation parameters is compared in real time with the preset stability index threshold, the pollutant emission concentration data is compared with the preset pollutant emission threshold, and the combustion efficiency data is compared with the preset combustion efficiency target value to generate comparison results for each dimension.

[0048] Based on the comparison results of each dimension, and combined with the status markers of the controlled follower nodes in the system, the active control nodes and the controlled follower nodes are distinguished and evaluated, generating a comprehensive comparison result that includes stability status, emission compliance status, and efficiency status.

[0049] In some embodiments of this application, when updating the combustion characteristic mapping relationship and the core parameters of the model predictive control algorithm based on the comparison results and the online learning algorithm, the following steps are included:

[0050] Based on the comprehensive comparison results, a multi-objective optimization function for parameter updating is constructed.

[0051] Using the aforementioned multi-objective optimization function, the parameter association weights in the combustion characteristic mapping relationship are dynamically adjusted through an online learning algorithm;

[0052] Meanwhile, based on the model prediction deviation reflected in the comprehensive comparison results, the parameters of the prediction model in the model prediction control algorithm are identified and corrected online.

[0053] When there are controlled follower nodes in the system, constraints are applied to the learning process of the controlled follower nodes based on their state flags to limit the impact of their abnormal data on the updating of core parameters.

[0054] The combustion characteristic mapping relationship and core parameters of the model predictive control algorithm, which have been updated through online learning, will be applied to the next control cycle.

[0055] In some embodiments of this application, a coal-fired boiler combustion stability control system is applied to the aforementioned coal-fired boiler combustion stability control method, characterized in that it includes:

[0056] A multi-source sensing module is used to obtain multi-dimensional real-time monitoring data, which includes process parameters and physical property parameters.

[0057] The edge computing and state assessment module is used to perform edge-side processing on the process parameters and physical property parameters to obtain a stability index characterizing the combustion state, and to determine the combustion stability level based on the stability index.

[0058] The intelligent decision-making and instruction generation module is used to generate basic control parameters based on the pre-built combustion characteristic mapping relationship, combustion stability level and the physical property parameters, predict the combustion state in the future period based on the model predictive control algorithm, and generate multi-dimensional collaborative control instructions in combination with the intelligent optimization algorithm.

[0059] The control execution and feedback monitoring module is used to send the multi-dimensional collaborative control commands to each actuator, drive each actuator to perform adjustment actions, collect combustion state data after the adjustment actions in real time, compare it with preset target data, and generate comparison results.

[0060] The online learning and adaptive update module is used to update the combustion characteristic mapping relationship and the core parameters of the model predictive control algorithm based on the comparison results and the online learning algorithm.

[0061] Compared with the prior art, the beneficial effects of the combustion stability control method for coal-fired boilers proposed in this application are as follows:

[0062] By synchronously collecting process and physical property parameters and utilizing edge-cloud collaborative computing, the shortcomings of traditional control—such as single data dimension and lagging state perception—are overcome, enabling accurate and real-time assessment of combustion stability. The integration of model predictive control and intelligent optimization algorithms not only achieves forward-looking prediction and optimization of future combustion states but also enables rapid generation of backup control strategies through system-wide collaboration when an anomaly occurs in a single boiler, ensuring the stable operation of the entire boiler system.

[0063] A feedback-based online learning mechanism was introduced, enabling the control model to continuously self-correct and optimize based on actual operating results, effectively adapting to changes in fuel characteristics and equipment aging, and continuously improving the intelligence and robustness of the control system. Attached Figure Description

[0064] Figure 1 This is a flowchart illustrating a preferred embodiment of a coal-fired boiler combustion stability control method. Detailed Implementation

[0065] The specific embodiments of this application will be described in further detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate this application, but are not intended to limit the scope of this application.

[0066] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0067] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0068] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0069] In some embodiments of this application, a method for controlling the combustion stability of a coal-fired boiler is provided, characterized by comprising:

[0070] Multi-dimensional real-time monitoring data is obtained through a multi-source sensing module, including process parameters and physical property parameters.

[0071] The process parameters and physical property parameters are subjected to edge processing to obtain a stability index characterizing the combustion state, and the combustion stability level is determined based on the stability index.

[0072] Based on the pre-constructed combustion characteristic mapping relationship, combustion stability level and the physical property parameters, basic control parameters are generated. The combustion state in the future period is predicted based on the model predictive control algorithm, and multi-dimensional collaborative control commands are generated in combination with the intelligent optimization algorithm.

[0073] The multi-dimensional collaborative control command is sent to each actuator to drive each actuator to perform adjustment actions. Combustion state data after the adjustment actions is collected in real time and compared with preset target data to generate comparison results.

[0074] Based on the comparison results and the online learning algorithm, the combustion characteristic mapping relationship and the core parameters of the model predictive control algorithm are updated.

[0075] In some embodiments of this application, when obtaining multi-dimensional real-time monitoring data through a multi-source sensing module, and the multi-dimensional real-time monitoring data includes process parameters and physical property parameters, the method includes:

[0076] A measurement group of multi-source sensing modules is built on each coal-fired boiler, in which two sets of acquisition devices with the same performance are configured for key process parameter acquisition points.

[0077] Based on the aforementioned measurement group, when the system contains multiple coal-fired boilers, the central control unit sends a unified synchronous acquisition command to all measurement groups to generate the original dataset.

[0078] In the original dataset, the coal-fired boiler system monitors the output of each acquisition device in real time. When the data of any unit continuously deviates from the median range of its hot standby unit and is determined to be invalid, the mean of its hot standby unit is used as the valid output to generate the dataset.

[0079] The dataset is compared with corresponding data from other measurement groups in the central processing unit to identify the consistency of key features, thereby generating cross-validation results.

[0080] Based on the cross-validation results, data from different measurement groups are fused and weighted to generate the multi-dimensional real-time monitoring data.

[0081] In this embodiment, the measurement group is a data acquisition unit responsible for a specific boiler and integrating multiple sensing functions. The key process parameter acquisition points include a furnace temperature field monitoring device, a furnace pressure sensor, and a flue gas oxygen meter.

[0082] In this embodiment, dynamic hot standby refers to the simultaneous operation and real-time comparison of two sets of data acquisition devices. When the primary device experiences data anomalies, the system can switch to the backup device without delay, ensuring the continuity of the data stream and achieving high reliability.

[0083] In this embodiment, the consistency comparison of key features is based on the fact that under similar operating conditions, the operating parameters of multiple normally functioning boilers will exhibit certain similarities. If the data features of a certain measurement group deviate significantly from the feature range formed by other normally functioning boilers, the system will suspect that there may be undiscovered hidden faults or common-cause interference in that data group, thereby generating a cross-validation result that indicates its credibility is questionable.

[0084] In this embodiment, the system fuses and weights data from all measurement groups based on the cross-validation results. Data with high cross-validation confidence is assigned a higher weight, while data with questionable confidence is assigned a lower weight.

[0085] In some embodiments of this application, the step of performing edge-side processing on the process parameters and physical property parameters to obtain a stability index characterizing the combustion state, and determining the combustion stability level based on the stability index, includes:

[0086] The local stability index is calculated using edge computing based on real-time data from a single measurement group.

[0087] The key features of the real-time data of the single measurement group are compared in real time with the group feature interval formed by all other normally functioning measurement groups in the system, and a group consistency index is generated to characterize the degree of data deviation.

[0088] A stability index is generated by performing a weighted fusion calculation based on the local stability index and the group consistency index.

[0089] The stability level of the current combustion is determined based on the numerical range of the stability index.

[0090] In this embodiment, if the population consistency index shows a significant deviation in data characteristics, but the local stability index indicates normality, an advanced diagnostic procedure is triggered. The advanced diagnostic procedure comprehensively evaluates the sensor health status and the physical authenticity of the combustion process, and corrects the local stability index.

[0091] In this embodiment, the edge computing function is implemented by deploying embedded processors with specific data processing capabilities within each measurement group or in the adjacent field control unit. This includes real-time filtering and noise reduction preprocessing of raw sensor signals such as temperature and pressure; running lightweight algorithms to extract the fractal dimension and brightness distribution variance from the flame image in real time; and calculating the preliminary stability criterion for the boiler corresponding to that measurement group, i.e., the local stability index, based on these extracted features.

[0092] In this embodiment, the group feature interval is the key feature data reported by all measurement groups marked as operating normally, which are continuously collected by the central processing unit, and a reasonable upper and lower limit boundary is generated in real time for each key feature.

[0093] In this embodiment, the stability index is specifically formulated as follows: Stability Index = w1 × Local Stability Index + w2 × Group Consistency Index. The weighting coefficients w1 and w2 are dynamically and adaptively adjusted. When the group consistency index of a measurement group is good, meaning its data highly matches the group's characteristic range, the system tends to trust its local calculation results and assigns a higher weight w1 to the local stability index. Conversely, when the group consistency index of the measurement group deteriorates, indicating that its data significantly deviates from the normal range of the group, the system reduces the confidence level of the local data, correspondingly decreasing the weight of w1, while increasing the weight of the group consistency index w2, so that the final stability index better reflects the relative anomaly of the boiler within the group.

[0094] In this embodiment, the classification of combustion stability levels includes: when the stability index is between 0.8 and 1.0, the system determines the combustion state to be stable. At this time, the control system mainly executes conventional economic optimization strategies to improve combustion efficiency and reduce pollutant emissions. When the index drops to the range of 0.6 to 0.8, it is determined to be slightly unstable. The control system then activates a preventative adjustment mode. The control algorithm strengthens the observation of combustion process disturbances and makes small-scale adjustments to the air-fuel ratio to curb the decline in state. When the index further drops to the range of 0.4 to 0.6, it is determined to be moderately unstable. The system will trigger an active combustion stabilization strategy, dynamically adjust the swirl intensity of the burner, and, if necessary, activate a combustion stabilization auxiliary device to forcibly restore combustion intensity. When the index is below 0.4, it is determined to be severely unstable, indicating a rapid deterioration in the combustion state and a risk of fire extinguishing. The system will execute the highest level of protective intervention strategy, forcing the boiler to switch to a preset safe combustion mode and simultaneously activating an audible and visual alarm to prompt operator intervention.

[0095] In some embodiments of this application, when generating basic control parameters based on pre-constructed combustion characteristic mapping relationships, combustion stability levels, and the physical property parameters, predicting the combustion state in future time periods based on model predictive control algorithms, and generating multi-dimensional collaborative control commands in conjunction with intelligent optimization algorithms, the process includes:

[0096] Based on the combustion stability level and the physical property parameters, a pre-built combustion characteristic mapping relationship is invoked to match the basic control parameters;

[0097] Based on the aforementioned basic control parameters, the combustion state in future time periods is predicted and the target is optimized by the model predictive control algorithm to generate a feedforward control sequence.

[0098] The feedforward control sequence and the stability index of real-time feedback are fused and calculated using an intelligent optimization algorithm to generate the multi-dimensional collaborative control command.

[0099] In this embodiment, the pre-constructed combustion characteristic mapping relationship includes: collecting a large amount of historical normal operation data, which includes physical property parameter data, corresponding process parameter data, and control parameter data that have been proven to maintain stable combustion at different combustion stability levels. Machine learning methods are used to train this data to construct a nonlinear mapping model. The system inputs the real-time obtained combustion stability level and physical property parameters into this model, and the model outputs a set of initial basic control parameters. This set of parameters includes a basic air-coal ratio value, a recommended primary air pressure setpoint, and a mode code indicating the air distribution mode of each burner layer.

[0100] In this embodiment, the cross-validation involves the central processing unit comparing the dataset from a certain measurement group with the corresponding data from other normal measurement groups in the system, focusing on the consistency of key features. When the value of a physical property parameter provided by a certain measurement group consistently and significantly deviates from the dynamic reasonable range constituted by similar parameters from most other normal units, the system will assign a logical label of doubtful reliability to that parameter. When the combustion characteristic mapping relationship requires this doubtful parameter, the system will activate fault-tolerant logic. The system will search for other boiler measurement groups with similar loads, fuel characteristics, and reliable data in the real-time operating boiler group to obtain their reliable physical property parameters. Based on the parameters of these reliable neighboring boilers, the system estimates the physical property parameter values ​​that the current boiler should have under normal conditions using interpolation calculation methods, and inputs this estimated value into the combustion characteristic mapping relationship. The control parameters output by the mapping relationship based on this estimated value serve as alternative control parameters, thereby effectively avoiding control errors that may be caused by a single data source failure.

[0101] In this embodiment, the model predictive control algorithm is a model-based multivariate predictive controller. This model can predict the boiler's combustion state over a future period based on the current control operation.

[0102] In this embodiment, in terms of system architecture, each normally operating coal-fired boiler is regarded as an active control node, and its control commands are mainly generated independently from its own real-time data and model prediction results. When the system determines through continuous cross-validation that the measurement group of a certain boiler is in a continuous abnormal state, the system will switch the control authority of that boiler from an active control node to a controlled follower node.

[0103] In this embodiment, the final generated multi-dimensional collaborative control commands include: inputting the feedforward control sequence output by the model predictive control algorithm, together with the stability index fed back in real time, into an optimization algorithm for fusion calculation. The final output command set precisely specifies the opening value of the secondary air damper of each burner layer, the opening value of the burnout air damper, the speed set value of the coal feeder, and the start / stop status and power level of the combustion stabilization auxiliary device.

[0104] In some embodiments of this application, when the basic control parameters are matched by calling a pre-constructed combustion characteristic mapping relationship based on the combustion stability level and the physical property parameters, the following steps are included:

[0105] Construct a dynamic multi-dimensional similarity evaluation function to calculate the similarity distance between the current boiler and all other normally operating boilers in the system in the multi-dimensional operating space in real time;

[0106] Based on the similarity distance, the top K normal boilers that are most similar to the current boiler are selected to form a dynamic boiler group with similar operating conditions.

[0107] Based on the physical property parameters and corresponding control parameters provided by the boiler group under similar operating conditions, which have been cross-validated as reliable, alternative control parameters suitable for the current boiler are generated by mapping through an interpolation algorithm.

[0108] The alternative control parameters are input into a pre-built rapid combustion stability assessment model for calculation and verification;

[0109] If the predicted stability index of its output is higher than the preset safety threshold, then the alternative control parameter is adopted as the basic control parameter.

[0110] Otherwise, the weights of the similarity evaluation function are readjusted, and a new round of calculation is performed.

[0111] In this embodiment, the key dimensions of the multi-dimensional operating space include: real-time load rate, volatile matter content of the received coal, and the ratio of total air volume to coal feed.

[0112] In some embodiments of this application, the step of predicting the combustion state in future time periods and performing target optimization based on the basic control parameters to generate a feedforward control sequence includes:

[0113] A feedforward control parameter window is constructed using the current basic control parameters as the initial sequence.

[0114] The feedforward control parameter window is input into the prediction model to obtain a prediction sequence of key combustion state parameters for the same future time period;

[0115] Construct a multi-objective optimization function, and optimize the sequence values ​​in the feedforward control parameter window under the preset conditions, with the multi-objective optimization function as the objective.

[0116] The first control parameter of the control parameter window obtained from each solution is used as the actual instruction to be executed, thereby generating the feedforward control sequence.

[0117] In this embodiment, the preset conditions include: the stability index is not lower than a preset safety threshold, and the constraint of physical adjustment limits for each actuator.

[0118] In this embodiment, the key combustion state parameters include: stability index, NOx emission concentration, and boiler thermal efficiency.

[0119] In this embodiment, a multi-objective optimization function is constructed, which aims to simultaneously minimize the fluctuation of the stability index, the cumulative value of NOx emission concentration, and maximize the cumulative value of boiler thermal efficiency.

[0120] In some embodiments of this application, the step of fusing the feedforward control sequence and the stability index of real-time feedback through an intelligent optimization algorithm to generate the multi-dimensional collaborative control command includes:

[0121] Based on the aforementioned feedforward control sequence and the stability index of real-time feedback, a real-time dynamic optimization problem is constructed.

[0122] Solve the real-time dynamic optimization problem to generate a set of preliminary control commands;

[0123] The preliminary control commands are compared with the physical adjustment limits of each actuator to generate a feasibility verification result;

[0124] Based on the feasibility verification results, the preliminary control instructions are modified and quantified, and finally the multi-dimensional collaborative control instructions are generated and output to each actuator.

[0125] In some embodiments of this application, the step of issuing the multi-dimensional collaborative control command to each actuator, driving each actuator to perform adjustment actions, collecting combustion state data after the adjustment actions in real time, comparing it with preset target data, and generating comparison results includes:

[0126] The multi-dimensional collaborative control commands are synchronously sent to the burner regulating mechanism, the combustion stabilization auxiliary device and the air-coal regulating mechanism, driving each actuator to coordinate and execute the regulating actions, and generating a coordinated action execution status.

[0127] Based on the execution status of the coordinated action, process parameters reflecting changes in combustion status are collected in real time, and combined with multi-dimensional real-time monitoring data, a real-time status dataset is generated.

[0128] Based on the real-time status dataset, the real-time comprehensive stability index is calculated, and pollutant emission concentration data and combustion efficiency data are obtained to generate an evaluation parameter set that includes stability, emissions and efficiency.

[0129] The comprehensive stability index of the evaluation parameters is compared in real time with the preset stability index threshold, the pollutant emission concentration data is compared with the preset pollutant emission threshold, and the combustion efficiency data is compared with the preset combustion efficiency target value to generate comparison results for each dimension.

[0130] Based on the comparison results of each dimension, and combined with the status markers of the controlled follower nodes in the system, the active control nodes and the controlled follower nodes are distinguished and evaluated, generating a comprehensive comparison result that includes stability status, emission compliance status, and efficiency status.

[0131] In this embodiment, the coordinated action execution status includes: after the central control unit issues multi-dimensional coordinated control commands, the system receives feedback signals from each actuator in real time. These actuators include the servo motor position feedback of the burner regulating mechanism, the operating status signal of the combustion stabilization auxiliary device, and the frequency feedback of the inverter in the air-fuel regulating mechanism. The system uses a status monitoring module to analyze and fuse these feedback signals to determine whether each mechanism has accurately executed the regulating action according to the command requirements, and whether the timing coordination between the actions meets the preset coordination requirements. Finally, the system generates a structured status report, which includes the individual action completion status of each actuator, as well as a comprehensive status index characterizing the degree of coordination of all mechanism actions. This structured report is the coordinated action execution status.

[0132] In this embodiment, the system calculates a real-time stability index based on the real-time status dataset using an embedded evaluation algorithm. Simultaneously, the system acquires real-time emission concentration data for pollutants such as nitrogen oxides and sulfur dioxide through connected online pollutant monitoring instruments, and calculates the current combustion efficiency data in real-time using a thermal efficiency model based on heat balance calculations. These three types of data—the real-time stability index, pollutant emission concentration data, and combustion efficiency data—are encapsulated to form an evaluation parameter set for comprehensively assessing the control effect.

[0133] In this embodiment, the generation of the comparison results for each dimension is achieved by comparing each parameter in the evaluation parameter set with its corresponding preset target value one by one. Specifically, the real-time stability index is compared with a preset stability index threshold to generate a comparison result for the stability dimension; the pollutant emission concentration data is compared with a preset pollutant emission threshold to generate a comparison result for the environmental emission dimension; and the combustion efficiency data is compared with a preset combustion efficiency target value to generate a comparison result for the economic dimension. Then, the system enters the advanced evaluation stage to generate a comprehensive comparison result. During this process, if the system identifies the existence of controlled follower nodes, it will attach special status markers to the above-mentioned comparison results for each dimension from such nodes. In the final comprehensive comparison result generated by the system, it will clearly distinguish and record which results come from active control nodes and which come from controlled follower nodes, thereby realizing differentiated performance evaluation and management of boilers in different states within the system.

[0134] In some embodiments of this application, when updating the combustion characteristic mapping relationship and the core parameters of the model predictive control algorithm based on the comparison results and the online learning algorithm, the following steps are included:

[0135] Based on the comprehensive comparison results, a multi-objective optimization function for parameter updating is constructed.

[0136] Using the aforementioned multi-objective optimization function, the parameter association weights in the combustion characteristic mapping relationship are dynamically adjusted through an online learning algorithm;

[0137] Meanwhile, based on the model prediction deviation reflected in the comprehensive comparison results, the parameters of the prediction model in the model prediction control algorithm are identified and corrected online.

[0138] When there are controlled follower nodes in the system, constraints are applied to the learning process of the controlled follower nodes based on their state flags to limit the impact of their abnormal data on the updating of core parameters.

[0139] The combustion characteristic mapping relationship and core parameters of the model predictive control algorithm, which have been updated through online learning, will be applied to the next control cycle.

[0140] In this embodiment, constructing a multi-objective optimization function for parameter updates includes assigning corresponding weights to each state based on the importance of the stability state, emission compliance state, and efficiency state in the comprehensive comparison results.

[0141] In this embodiment, when a controlled follower node exists in the system, the data generated by the controlled follower node is analyzed to determine whether it is abnormal data. For data determined to be abnormal, its influence in the parameter update process is reduced.

[0142] In this embodiment, before applying the online-learned and updated combustion characteristic mapping relationship and the core parameters of the model predictive control algorithm to the next control cycle, an offline simulation test is performed. By inputting historical operating data, the combustion process of the next control cycle is simulated to verify whether the updated parameters can effectively improve combustion stability, reduce pollutant emissions, and increase combustion efficiency. If the simulation test results meet the expected requirements, they are applied to actual control; if not, the parameters or constraints of the online learning algorithm are readjusted, and a new round of parameter updates is performed until a satisfactory parameter configuration is obtained.

[0143] In some embodiments of this application, a combustion stability control system for a coal-fired boiler is applied to a combustion stability control method for a coal-fired boiler, including:

[0144] A multi-source sensing module is used to obtain multi-dimensional real-time monitoring data, which includes process parameters and physical property parameters.

[0145] The edge computing and state assessment module is used to perform edge-side processing on the process parameters and physical property parameters to obtain a stability index characterizing the combustion state, and to determine the combustion stability level based on the stability index.

[0146] The intelligent decision-making and instruction generation module is used to generate basic control parameters based on the pre-built combustion characteristic mapping relationship, combustion stability level and the physical property parameters, predict the combustion state in the future period based on the model predictive control algorithm, and generate multi-dimensional collaborative control instructions in combination with the intelligent optimization algorithm.

[0147] The control execution and feedback monitoring module is used to send the multi-dimensional collaborative control commands to each actuator, drive each actuator to perform adjustment actions, collect combustion state data after the adjustment actions in real time, compare it with preset target data, and generate comparison results.

[0148] The online learning and adaptive update module is used to update the combustion characteristic mapping relationship and the core parameters of the model predictive control algorithm based on the comparison results and the online learning algorithm.

[0149] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of this application, and these improvements and substitutions should also be considered within the scope of protection of this application.

Claims

1. A method for controlling the combustion stability of a coal-fired boiler, characterized in that, include: Multi-dimensional real-time monitoring data is obtained through a multi-source sensing module, including process parameters and physical property parameters. The process parameters and physical property parameters are subjected to edge processing to obtain a stability index characterizing the combustion state, and the combustion stability level is determined based on the stability index. Based on the pre-constructed combustion characteristic mapping relationship, combustion stability level and the physical property parameters, basic control parameters are generated. The combustion state in the future period is predicted based on the model predictive control algorithm, and multi-dimensional collaborative control commands are generated in combination with the intelligent optimization algorithm. The multi-dimensional collaborative control command is sent to each actuator to drive each actuator to perform adjustment actions. Combustion state data after the adjustment actions is collected in real time and compared with preset target data to generate comparison results. Based on the comparison results and the online learning algorithm, the combustion characteristic mapping relationship and the core parameters of the model predictive control algorithm are updated.

2. The method for controlling combustion stability of a coal-fired boiler according to claim 1, characterized in that, When multi-dimensional real-time monitoring data is obtained through a multi-source sensing module, and the multi-dimensional real-time monitoring data includes process parameters and physical property parameters, it includes: A measurement group of multi-source sensing modules is built on each coal-fired boiler, in which two sets of acquisition devices with the same performance are configured for key process parameter acquisition points. Based on the aforementioned measurement group, when the system contains multiple coal-fired boilers, the central control unit sends a unified synchronous acquisition command to all measurement groups to generate the original dataset. In the original dataset, the coal-fired boiler system monitors the output of each acquisition device in real time. When the data of any unit continuously deviates from the median range of its hot standby unit and is determined to be invalid, the mean of its hot standby unit is used as the valid output to generate the dataset. The dataset is compared with corresponding data from other measurement groups in the central processing unit to identify the consistency of key features, thereby generating cross-validation results. Based on the cross-validation results, data from different measurement groups are fused and weighted to generate the multi-dimensional real-time monitoring data.

3. The method for controlling combustion stability of a coal-fired boiler according to claim 1, characterized in that, The process parameters and physical property parameters are subjected to edge processing to obtain a stability index characterizing the combustion state. When determining the combustion stability level based on the stability index, the following steps are included: The local stability index is calculated using edge computing based on real-time data from a single measurement group. The key features of the real-time data of the single measurement group are compared in real time with the group feature interval formed by all other normally functioning measurement groups in the system, and a group consistency index is generated to characterize the degree of data deviation. A stability index is generated by performing a weighted fusion calculation based on the local stability index and the group consistency index. The stability level of the current combustion is determined based on the numerical range of the stability index.

4. The method for controlling combustion stability of a coal-fired boiler according to claim 2, characterized in that, When generating basic control parameters based on pre-constructed combustion characteristic mapping relationships, combustion stability levels, and physical property parameters, predicting the combustion state in future time periods based on model predictive control algorithms, and generating multi-dimensional collaborative control commands in conjunction with intelligent optimization algorithms, the process includes: Based on the combustion stability level and the physical property parameters, a pre-built combustion characteristic mapping relationship is invoked to match the basic control parameters; Based on the aforementioned basic control parameters, the combustion state in future time periods is predicted and the target is optimized by the model predictive control algorithm to generate a feedforward control sequence. The feedforward control sequence and the stability index of real-time feedback are fused and calculated using an intelligent optimization algorithm to generate the multi-dimensional collaborative control command.

5. The method for controlling combustion stability of a coal-fired boiler according to claim 4, characterized in that, When the basic control parameters are matched by calling a pre-built combustion characteristic mapping relationship based on the combustion stability level and the physical property parameters, the following steps are included: Construct a dynamic multi-dimensional similarity evaluation function to calculate the similarity distance between the current boiler and all other normally operating boilers in the system in the multi-dimensional operating space in real time; Based on the similarity distance, the top K normal boilers that are most similar to the current boiler are selected to form a dynamic boiler group with similar operating conditions. Based on the physical property parameters and corresponding control parameters provided by the boiler group under similar operating conditions, which have been cross-validated as reliable, alternative control parameters suitable for the current boiler are generated by mapping through an interpolation algorithm. The alternative control parameters are input into a pre-built rapid combustion stability assessment model for calculation and verification; If the predicted stability index of its output is higher than the preset safety threshold, then the alternative control parameter is adopted as the basic control parameter. Otherwise, the weights of the similarity evaluation function are readjusted, and a new round of calculation is performed.

6. The method for controlling combustion stability of a coal-fired boiler according to claim 4, characterized in that, When generating a feedforward control sequence based on the aforementioned basic control parameters, by predicting the combustion state in future time periods using a model predictive control algorithm and performing target optimization, the process includes: A feedforward control parameter window is constructed using the current basic control parameters as the initial sequence. The feedforward control parameter window is input into the prediction model to obtain a prediction sequence of key combustion state parameters for the same future time period; Construct a multi-objective optimization function, and optimize the sequence values ​​in the feedforward control parameter window under the preset conditions, with the multi-objective optimization function as the objective. The first control parameter of the control parameter window obtained from each solution is used as the actual instruction to be executed, thereby generating the feedforward control sequence.

7. The method for controlling combustion stability of a coal-fired boiler according to claim 4, characterized in that, The step of fusing the feedforward control sequence and the stability index of real-time feedback through an intelligent optimization algorithm to generate the multi-dimensional collaborative control command includes: Based on the aforementioned feedforward control sequence and the stability index of real-time feedback, a real-time dynamic optimization problem is constructed. Solve the real-time dynamic optimization problem to generate a set of preliminary control commands; The preliminary control commands are compared with the physical adjustment limits of each actuator to generate a feasibility verification result; Based on the feasibility verification results, the preliminary control instructions are modified and quantified, and finally the multi-dimensional collaborative control instructions are generated and output to each actuator.

8. The method for controlling combustion stability of a coal-fired boiler according to claim 1, characterized in that, The step of issuing the multi-dimensional collaborative control command to each actuator, driving each actuator to perform adjustment actions, and collecting combustion state data after the adjustment actions in real time, comparing it with preset target data, and generating comparison results includes: The multi-dimensional collaborative control commands are synchronously sent to the burner adjustment mechanism, the combustion stabilization auxiliary device and the air-coal adjustment mechanism, driving each actuator to coordinate and execute adjustment actions, and generating a coordinated action execution status. Based on the execution status of the coordinated action, process parameters reflecting changes in combustion status are collected in real time, and combined with multi-dimensional real-time monitoring data, a real-time status dataset is generated. Based on the real-time status dataset, the real-time comprehensive stability index is calculated, and pollutant emission concentration data and combustion efficiency data are obtained to generate an evaluation parameter set that includes stability, emissions and efficiency. The comprehensive stability index of the evaluation parameters is compared in real time with the preset stability index threshold, the pollutant emission concentration data is compared with the preset pollutant emission threshold, and the combustion efficiency data is compared with the preset combustion efficiency target value to generate comparison results for each dimension. Based on the comparison results of each dimension, and combined with the status markers of the controlled follower nodes in the system, the active control node and the controlled follower node are distinguished and evaluated, generating comparison results that include stability status, emission compliance status, and efficiency status.

9. The method for controlling combustion stability of a coal-fired boiler according to claim 1, characterized in that, When updating the combustion characteristic mapping relationship and the core parameters of the model predictive control algorithm based on the comparison results and the online learning algorithm, the following is included: Based on the comparison results, a multi-objective optimization function for parameter updating is constructed. Using the aforementioned multi-objective optimization function, the parameter association weights in the combustion characteristic mapping relationship are dynamically adjusted through an online learning algorithm; Meanwhile, based on the model prediction deviation reflected in the comprehensive comparison results, the parameters of the prediction model in the model prediction control algorithm are identified and corrected online. When there are controlled follower nodes in the system, constraints are applied to the learning process of the controlled follower nodes based on their state flags to limit the impact of their abnormal data on the updating of core parameters. The combustion characteristic mapping relationship and core parameters of the model predictive control algorithm, which have been updated through online learning, will be applied to the next control cycle.

10. A combustion stability control system for a coal-fired boiler, applied to the combustion stability control method for a coal-fired boiler as described in any one of claims 1-9, characterized in that, include: A multi-source sensing module is used to obtain multi-dimensional real-time monitoring data, which includes process parameters and physical property parameters. The edge computing and state assessment module is used to perform edge-side processing on the process parameters and physical property parameters to obtain a stability index characterizing the combustion state, and to determine the combustion stability level based on the stability index. The intelligent decision-making and instruction generation module is used to generate basic control parameters based on the pre-built combustion characteristic mapping relationship, combustion stability level and the physical property parameters, predict the combustion state in the future period based on the model predictive control algorithm, and generate multi-dimensional collaborative control instructions in combination with the intelligent optimization algorithm. The control execution and feedback monitoring module is used to send the multi-dimensional collaborative control commands to each actuator, drive each actuator to perform adjustment actions, collect combustion state data after the adjustment actions in real time, compare it with preset target data, and generate comparison results. The online learning and adaptive update module is used to update the combustion characteristic mapping relationship and the core parameters of the model predictive control algorithm based on the comparison results and the online learning algorithm.

Images