Boiler decoking method, device, equipment and medium

Abstract

The invention provides a boiler decoking method, a boiler decoking device, boiler decoking equipment and a medium, and the method comprises the following steps: collecting multi-mode coke residue distribution data in a boiler; carrying out fusion processing on the multi-mode coke residue distribution data to obtain coke residue distribution characteristic parameters in the boiler; based on the coke residue distribution characteristic parameters, determining a coke residue growth trend by using a coke residue growth prediction model; according to the growth trend of the coke slag, acoustic wave control parameters and steam control parameters are determined through a digital twinborn simulation technology; and controlling a sound wave generator and a steam pulse device to perform targeted decoking on the boiler based on the sound wave control parameters and the steam control parameters. The decoking efficiency and accuracy can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of polycrystalline silicon, in particular to a boiler decoking method, device, equipment and medium. BACKGROUND

[0002] During long-term operation, the heating surface of the boiler will gradually form a layer of coke due to the influence of high temperature and fuel characteristics. These cokes will significantly reduce the heat exchange efficiency of the boiler, increase the exhaust gas temperature, increase energy consumption, and even cause local overheating and structural damage in severe cases.

[0003] Traditional boiler decoking technology mainly includes mechanical cleaning, soot blowing by soot blower and water flushing methods. Conventional decoking adopts a periodic and comprehensive strategy, without considering the actual distribution of the cokes inside the boiler, which not only wastes energy but also makes it difficult to achieve targeted and precise decoking, resulting in incomplete cleaning in some areas and excessive treatment in other areas, affecting the overall operation efficiency and equipment service life of the boiler.

[0004] Therefore, the existing boiler decoking method has low accuracy and low decoking efficiency. SUMMARY

[0005] The technical problem to be solved by the present application is to solve the above-mentioned deficiencies of the prior art, and to provide a boiler decoking method, device, equipment and medium, which can improve the decoking efficiency and accuracy.

[0006] In a first aspect, the present application provides a boiler decoking method, the method comprising:

[0007] Collecting multi-modal coke distribution data inside the boiler; the multi-modal coke distribution data is the original state information of the coke directly obtained by the multi-source sensor;

[0008] Fusing the multi-modal coke distribution data to obtain coke distribution characteristic parameters inside the boiler, the coke distribution characteristic parameters being used to represent the physical state of the coke and its evolution law;

[0009] Based on the coke distribution characteristic parameters, a coke growth prediction model is used to determine the coke growth trend, the coke growth prediction model being used to predict the dynamic change trend of the coke;

[0010] According to the coke growth trend, a sound wave control parameter and a steam control parameter are determined by a digital twin simulation technology, the digital twin technology being a simulation technology that reflects the digital twin model of the boiler thermal-structure coupling relationship, simulates the physical process of the sound wave and the steam on the coke and optimizes the parameters;

[0011] control an acoustic wave generator and a steam pulse device based on the acoustic wave control parameter and the steam control parameter to perform targeted decoking on the boiler.

[0012] In a second aspect, the present application provides a boiler decoking device, which comprises:

[0013] a collection module configured to collect multi-modal soot distribution data inside a boiler; the multi-modal soot distribution data is original soot state information directly obtained by a multi-source sensor;

[0014] an acquisition module connected to the collection module and configured to perform fusion processing on the multi-modal soot distribution data to obtain soot distribution characteristic parameters inside the boiler; the soot distribution characteristic parameters are used to represent the physical state of soot and its evolution law;

[0015] a first determination module connected to the acquisition module and configured to determine a soot growth trend based on the soot distribution characteristic parameters by using a soot growth prediction model; the soot growth prediction model is used to predict the dynamic change trend of soot;

[0016] a second determination module connected to the first determination module and configured to determine acoustic wave control parameters and steam control parameters by using digital twin simulation technology according to the soot growth trend; the digital twin technology is a simulation technology for simulating the physical process of the action of acoustic waves and steam on soot and optimizing parameters by using a digital twin model reflecting the thermal-structure coupling relationship of a boiler;

[0017] a control module connected to the second determination module and configured to control an acoustic wave generator and a steam pulse device based on the acoustic wave control parameter and the steam control parameter to perform targeted decoking on the boiler.

[0018] In a third aspect, the present application provides an electronic device, which comprises a processor, a memory, and a program or instruction stored on the memory and executable on the processor; when the program or instruction is executed by the processor, the method for decoking a boiler according to the first aspect is implemented.

[0019] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program; when the computer program is executed by a processor, the method for decoking a boiler according to the first aspect is implemented.

[0020] The boiler decoking method, device, equipment and medium provided by the application first collect multi-modal coke distribution data in the boiler, improve the breadth and accuracy of coke state information collection, then fuse the multi-source heterogeneous multi-modal coke distribution data, extract characteristic parameters representing the real physical state of the coke in the boiler and the evolution law thereof, predict the growth trend of the coke by using a coke growth prediction model based on the characteristic parameters, then optimize the combination of sound wave and steam control parameters in a virtual environment by using digital twin technology according to the growth trend of the coke, and finally realize accurate targeted and collaborative control of the sound wave generator and the steam pulse device. The multi-modal data collection, data fusion, intelligent prediction and digital twin parameter optimization are combined to realize accurate targeted and collaborative control of the boiler decoking, and improve the decoking efficiency and accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 A flowchart of a boiler decoking method according to Embodiment 1 of the application;

[0022] Figure 2 A reference diagram of the overall algorithm framework of another boiler decoking method according to Embodiment 1 of the application;

[0023] Figure 3 A structural schematic diagram of a boiler decoking device according to Embodiment 2 of the application;

[0024] Figure 4 A structural schematic diagram of an equipment according to Embodiment 3 of the application. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the technical solutions of the application, the embodiments of the application will be further described in detail below with reference to the drawings.

[0026] It can be understood that the specific embodiments and drawings described herein are only used to explain the application, but not to limit the application.

[0027] It can be understood that the embodiments in the application and the features in the embodiments can be combined with each other without conflict.

[0028] It can be understood that, for the convenience of description, only parts related to the application are shown in the drawings of the application, and parts unrelated to the application are not shown in the drawings.

[0029] It can be understood that each unit and module involved in the embodiments of the application can correspond to only one entity structure, or can be composed of multiple entity structures, or multiple units and modules can be integrated into one entity structure.

[0030] It can be understood that the terms "first", "second" and the like in the embodiments of the present application are used to distinguish different objects or different treatments of the same object, rather than to describe a specific order of the objects.

[0031] It can be understood that the functions, steps marked in the flowcharts and block diagrams of the present application can occur in an order different from that marked in the drawings, without conflict.

[0032] It can be understood that in the flowcharts and block diagrams of the present application, the architecture, functions and operations of the possible implementation of the system, device, equipment and method according to the embodiments of the present application are shown. Each block in the flowchart or block diagram can represent a unit, module, program segment, code, which contains executable instructions for realizing the specified function. Moreover, each block or combination of blocks in the block diagram and flowchart can be realized by a hardware-based system for realizing the specified function, or by a combination of hardware and computer instructions.

[0033] It can be understood that the units and modules involved in the embodiments of the present application can be realized by software or hardware, for example, the units and modules can be located in a processor.

[0034] It should be noted that the scene diagrams described in the embodiments of the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. It can be known by those skilled in the art that with the evolution of network architecture and the appearance of new business scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0035] The current boiler decoking field faces multiple technical challenges. The existing monitoring means relies on manual inspection, which is difficult to capture the dynamic growth process of the coke in real time, resulting in the lack of preventive maintenance. The accumulation of coke reduces the thermal efficiency of the boiler by 5-8%, and even causes local overheating, pipe explosion and other safety accidents. Traditional mechanical decoking methods include running soot blowing, stopping for manual removal and chemical decoking.

[0036] The commonly used runtime soot blowing methods include steam soot blower and acoustic soot blower. The steam soot blower is suitable for a wide range of applications and can be arranged at various parts of the boiler. It can directly blow soot on the coking area of the heating surface of the furnace, horizontal flue and tail vertical shaft, and has good soot blowing effect. It is particularly effective for strong coking, low ash melting point and sticky ash. However, the blowing range is limited due to factors such as blowing pressure and nozzle design. The acoustic soot blower can reach positions that other soot blowers cannot reach, but its cleaning capacity is limited due to the limitation of acoustic energy, and it has poor effect on stubborn deposits and fouling surfaces. Manual cleaning has a high safety risk, may cause mechanical damage to the pipeline, and can cause the boiler maintenance period to be extended. Chemical decoking can be operated online, but it has a high corrosion risk, drug residues pollute the environment, and the effect on deep coking (more than 20 cm) is limited. Therefore, the development of innovative technologies that integrate intelligent monitoring, precise decoking and energy efficiency optimization has become a key to the transformation and upgrading of the industry.

[0037] Based on the above technical problems, the present application provides a boiler decoking method to improve the efficiency and accuracy of boiler decoking.

[0038] Embodiment 1:

[0039] The present embodiment provides a boiler decoking method, which can be used for intelligent monitoring and collaborative removal of high-temperature and high-pressure boiler coking in polysilicon production enterprises, thermal power plants, chemical plants, metallurgical enterprises and large industrial production facilities. Specifically, it can be applied to large coal-fired boilers, circulating fluidized bed boilers, waste incineration boilers, and process steam boilers, tail gas treatment boilers and heat recovery boilers in the polysilicon production process, and other equipment prone to serious coking.

[0040] As shown in Figure 1 The method comprises the following steps S101-S105.

[0041] S101, collect multi-modal coking distribution data inside the boiler.

[0042] The multi-modal coking distribution data is the original state information of the coking obtained directly by the multi-source sensor. "Multi-modal" means that the data sources are diverse, including various signal data, and can also involve image or parameter data.

[0043] Specifically, the original state information of the coking inside the boiler (such as sound waves, infrared, steam parameters, etc.) is directly collected by deploying multiple types of sensors to form a comprehensive perception of the coking state of the boiler.

[0044] S102, fuse the multi-modal coking distribution data to obtain coking distribution characteristic parameters inside the boiler.

[0045] The coking distribution characteristic parameters are used to represent the physical state of the coking and its evolution law.

[0046] Specifically, different types of coke residue data are fused, that is, multi-source heterogeneous data is fused, and characteristic parameters that can truly reflect the coking state are extracted. The data utilization rate and analysis accuracy are improved.

[0047] S103, based on the coke residue distribution characteristic parameters, using a coke residue growth prediction model to determine the growth trend of the coke residue.

[0048] The coke residue growth prediction model is used to predict the dynamic change trend of the coke residue.

[0049] Specifically, the growth trend of the coke residue is inferred using the prediction model, which can accurately determine the growth trend of the coke residue in the current working condition of the boiler.

[0050] S104, according to the coke residue growth trend, determining the sound wave control parameter and the steam control parameter through digital twin simulation technology.

[0051] The digital twin technology is a simulation technology that simulates the physical process of the sound wave and the steam acting on the coke residue through a digital twin model reflecting the thermal-structure coupling relationship of the boiler and optimizes the parameters.

[0052] Specifically, by applying digital twin simulation technology, different combinations of control parameters are evaluated and optimized in a virtual environment, without the need for actual trial and error, greatly shortening the test cycle and improving the accuracy of parameter matching. The sound wave and steam control parameters determined by the digital twin simulation technology not only achieve accurate control and prediction of the descaling effect, but also form a linkage and synergistic mechanism of "sound wave loosening-steam impact" through the phase coordination control of sound wave-steam, improve the descaling efficiency, and reduce the energy consumption.

[0053] S105, based on the sound wave control parameter and the steam control parameter, controlling the sound wave generator and the steam pulse device to target the descaling of the boiler.

[0054] Specifically, the sound wave generator and the steam pulse device are controlled to accurately match the output parameters of different parts and different degrees of coke residue conditions, avoiding energy waste and ineffective cleaning in the traditional descaling process, while effectively improving the thoroughness and timeliness of the cleaning.

[0055] In the embodiments of the present application, the multi-modal coke distribution data inside the boiler is first collected directly through multi-source sensors, which can comprehensively and accurately obtain the original state information of the coke, thereby overcoming the shortcomings of limited information collection and limited monitoring range of traditional single sensing means. Through fusion processing of the multi-modal coke distribution data, characteristic parameters representing the real physical state of the coke and its evolution law are extracted, making the subsequent analysis more scientific and accurate. Further, based on these characteristic parameters, the future dynamic development trend of the coke is predicted in advance by means of a coke growth prediction model, avoiding the disadvantages of experience-based and blind operation. Subsequently, the digital twin model is used to simulate and optimize the sound wave and steam decoking parameters in the virtual environment, reducing the actual trial and error risk and realizing the scientific optimization of the control parameters. Finally, through the cooperative control of the sound wave generator and the steam pulse device, targeted decoking operation can be implemented, improving the accuracy and efficiency of the decoking operation.

[0056] Optionally, the multi-modal coke distribution data includes: sound wave reflection signal, infrared thermal imaging signal and steam parameter signal.

[0057] Among them, the sound wave reflection signal is collected by arranging a high-temperature piezoelectric ceramic sound wave sensor array in the horizontal flue and vertical flue area; the infrared thermal imaging signal is collected by arranging an endoscopic high-temperature infrared thermal imager in the coke deposition prone position; and the steam parameter signal is collected by arranging pressure, flow and temperature sensors on the steam pipeline.

[0058] Here, the sound wave reflection signal provides coke thickness and hardness information. The sound wave reflection signal can include: acoustic parameters such as reflection wave arrival time (which can calculate thickness), amplitude, main frequency, reflection energy distribution, etc. It also includes signal spatial distribution, which can be used to locate the position and local thickness of the coke.

[0059] The infrared thermal imaging signal provides temperature distribution and heat flow information. The infrared thermal imaging signal can include: temperature distribution image, temperature gradient (for density, thermal resistance analysis), time series temperature data. The regional heat flux density, temperature difference, heat distribution uniformity, etc. can be further calculated by using these data.

[0060] The steam parameter signal can include: steam pressure, temperature, flow and other related auxiliary data (such as transient change, reflecting the state of soot blowing operation). The steam parameter signal provides indirect heat exchange efficiency information.

[0061] In the embodiment, by constructing a comprehensive perception network of sound, light, heat, pressure and other multi-physical quantities, the technical problem that the traditional single parameter monitoring is difficult to accurately capture the complex coke distribution of the boiler is solved. Among them, the high-temperature piezoelectric ceramic acoustic sensor array can obtain the flue area coke thickness information in harsh environment; the endoscopic high-temperature infrared thermal imager directly visualizes the temperature anomaly of the key parts such as the water cooling wall; the multi-parameter monitoring on the steam pipeline provides direct evidence of the influence of coke on heat exchange efficiency.

[0062] Through the multi-dimensional and multi-regional collaborative perception strategy, not only the comprehensiveness, accuracy and timeliness of the coke distribution data are greatly improved, but also the precise characterization of the coke characteristics of different parts is realized through differentiated monitoring methods, and the accuracy of subsequent data fusion processing and coke growth trend prediction is improved.

[0063] Optionally, the multi-modal coke distribution data can further include: the boiler operating condition data can include: load, combustion parameter, air supply amount, temperature, etc. The boiler operating condition signal provides the overall environment and boundary conditions. The boiler operating condition data can be collected through a distributed control system.

[0064] Optionally, the S102 specifically includes:

[0065] The multi-modal coke distribution data is subjected to time-space alignment by using a Kalman filter algorithm to obtain first distribution data.

[0066] The first distribution data is subjected to filtering and denoising processing to obtain second distribution data.

[0067] The second distribution data is subjected to principal component analysis dimension reduction processing to obtain third distribution data.

[0068] The third distribution data is subjected to normalization processing by using a deviation standardization method to obtain target distribution data.

[0069] Coke distribution characteristic parameters are extracted from the target distribution data.

[0070] Specifically, the above steps include the following contents:

[0071] Time-space alignment processing: the multi-modal coke distribution data from different sensors, different time points and different spatial positions is subjected to time-space coordinate system unification and alignment by using a Kalman filter algorithm to generate first distribution data.

[0072] Signal denoising processing: the first distribution data is subjected to denoising processing. For example, a Savitzky-Golay filter can be used for denoising. The filter is based on the least square polynomial fitting principle and can reduce noise while retaining the shape characteristics of the original signal to generate second distribution data.

[0073] Dimension reduction processing: applying principal component analysis (PCA) technique to the second distribution data for dimension reduction, extracting the main variation components in the data, eliminating redundant information, and obtaining third distribution data with lower dimension but higher information content.

[0074] Normalization processing: using the dispersion standardization method to normalize the different dimension parameters in the third distribution data, converting them into dimensionless comparison benchmarks, and forming target distribution data with uniform scale.

[0075] Feature extraction: extracting feature parameters that can accurately represent the internal slag distribution state of the boiler from the target distribution data after the above processing.

[0076] In this embodiment, a systematic data processing link of "alignment-noise reduction-dimension reduction-normalization-extraction" is constructed, and the introduction of Kalman filter algorithm effectively handles the inconsistency problems of different sensor data in time and space; the filter processing can remove random noise interference while retaining the real characteristics of slag distribution to the greatest extent; the principal component analysis dimension reduction processing not only reduces data redundancy, but also improves calculation efficiency, highlighting the key information of slag distribution; the dispersion standardization solves the problem that different physical quantities such as sound waves, infrared, and pressure cannot be directly compared. The original disordered multi-source data are converted into a high-quality, low-dimensional, standardized feature parameter set, significantly enhancing the extraction accuracy and representativeness of the slag distribution characteristics.

[0077] Optionally, the above extracting slag distribution feature parameters from the target distribution data specifically includes:

[0078] The slag thickness is determined by the time-frequency of the sound wave reflection signal in the target distribution data; the temperature gradient data of the slag are determined by the temperature gradient of the infrared thermal imaging signal in the target distribution data; the slag distribution density is determined according to the temperature gradient data; the slag growth rate is determined by the change data of the slag thickness in a continuous time period; and the slag thermal resistance value is determined by the steam parameter signal and the temperature gradient data in the target distribution data.

[0079] The slag distribution feature parameters include the slag thickness, the slag distribution density, the slag growth rate, and the slag thermal resistance value.

[0080] Specifically, the physical thickness of the slag can be quantified by using the time delay and frequency change characteristics of the sound wave when it reflects at different medium interfaces. Optionally, the time-frequency characteristics such as the arrival time and main frequency component of the echo can be obtained by acquiring the sound wave reflection signal and performing time-frequency analysis (such as short-time Fourier transform, etc.); based on the time delay, energy attenuation, and spectrum change of the echo, combined with the propagation path information of the sound wave in the medium, the echo difference of the detected surface relative to the reference interface is determined, so as to calculate the normal thickness of the slag layer.

[0081] By analyzing the infrared thermal imaging signal, the temperature gradient change of different regions is determined, and the temperature gradient data of the cinder is obtained. Alternatively, the temperature field can be extracted from the infrared thermal image frame / matrix, and the gradient of the temperature in the spatial dimension (e.g. the gradient amplitude along the surface coordinate direction) can be calculated; the temperature gradient calculation result is taken as the "temperature gradient data", and is associated with the corresponding time stamp.

[0082] The temperature gradient reflects the non-uniformity of heat conduction, thereby indirectly representing the density of the cinder distribution. Alternatively, the target region can be judged by threshold or segmentation using the temperature gradient data (for example, the region is divided according to the gradient amplitude), and the regions with significant and non-significant temperature gradients are identified; the distribution degree of the cinder in space is quantified in the form of the area ratio of the identified corresponding region or the frequency per unit area, that is, the cinder distribution density is determined.

[0083] By the change data of the cinder thickness in the continuous time period, the growth amount of the cinder thickness per unit time is calculated, thereby determining the growth rate of the cinder and establishing the time dimension data of the cinder formation. Alternatively, the time series comparison of the cinder thickness at the same spatial position in the continuous time period can be performed; the growth rate is calculated according to the change amount of the thickness with time and the time interval.

[0084] Combined with the parameter data such as steam temperature and pressure in the steam parameter signal and the temperature gradient data, the degree of hindering heat transfer by the cinder, i.e. the thermal resistance value, is calculated according to the heat transfer principle of thermodynamics. Alternatively, the steam parameter signal associated with the target region and the temperature gradient data corresponding to the time can be taken; the equivalent thermal resistance value of the cinder layer is calculated by using the working condition related information provided by the steam parameter signal and combining the temperature field change corresponding to the temperature gradient data.

[0085] Finally, the cinder distribution characteristic parameter set including the cinder thickness, the cinder distribution density, the cinder growth rate and the cinder thermal resistance value is formed.

[0086] In this embodiment, the cinder thickness is obtained by time-frequency analysis of the acoustic wave reflection signal, realizing quantitative evaluation of the physical existence of the cinder; the distribution density is determined by the temperature gradient and the steam parameter of the infrared thermal imaging signal, revealing the spatial non-uniformity of the cinder; the growth rate is calculated based on the historical data, introducing the time dimension of the dynamic change of the cinder; and the influence degree of the cinder on the heat exchange efficiency of the boiler is directly quantified by the temperature gradient and the thermal resistance value. Through the multi-dimensional and complementary characteristic parameter set, not only the physical characteristics, spatial distribution, time evolution and thermal influence of the cinder are comprehensively described, but also more detailed and multi-angle data support is provided for subsequent cinder growth trend prediction and cinder removal parameter optimization, improving the scientificity and accuracy of the boiler cinder removal.

[0087] Alternatively, the above method further comprises:

[0088] obtain a boiler operation history data, the history data including a history collected acoustic wave reflection signal, an infrared thermal imaging signal and a steam parameter;

[0089] extract a history cinder distribution characteristic parameter based on the boiler operation history data, the history cinder distribution characteristic parameter including a cinder thickness, a cinder distribution density, a cinder growth rate and a cinder thermal resistance value;

[0090] construct a neural network model according to the history cinder distribution characteristic parameter;

[0091] adjust a hyperparameter of the neural network until a prediction error satisfies a preset condition, and obtain a cinder growth prediction model.

[0092] Here, the neural network model is a deep learning model for establishing a mapping relationship from input features (such as history characteristic parameters and optional working condition quantities) to output targets (such as cinder growth related prediction quantities). It can be a Long Short-Term Memory (LSTM), a Gated Recurrent Unit (GRU), a Convolutional Neural Network (CNN), etc.

[0093] The hyperparameter refers to a parameter that needs to be preset before the model training starts, which determines the structural characteristics and training process of the model. The hyperparameter can include: network structure related hyperparameters, such as the number of layers, the number of neurons, the hidden layer dimension, the activation function type, etc.; training process related hyperparameters, such as the learning rate, the batch size, the number of training rounds (epochs), the optimizer type, etc.; regularization related hyperparameters, such as the dropout rate: the proportion of randomly inactivated nodes to prevent overfitting, the L1 / L2 regularization coefficient, etc. How to determine and adjust the hyperparameter can be selected according to the actual situation, which is not limited in the present application.

[0094] Specifically, first, the history collected acoustic wave reflection signal, the infrared thermal imaging signal and the steam parameter in the boiler operation process are obtained, and the cinder growth trend corresponding to these data (which can be obtained by actual detection or simulation) is also obtained. Then, based on the obtained boiler operation history data, the cinder distribution characteristic parameter is extracted, including the cinder thickness, the cinder distribution density, the cinder growth rate and the cinder thermal resistance value. The original data is converted into key parameters that can directly represent the characteristics of the cinder; then, the extracted cinder distribution characteristic parameter is used as input to construct a neural network model. By adjusting the hyperparameters of the neural network, such as the learning rate, the number of hidden layers, the number of neurons, etc., repeated training and verification are performed until the prediction error of the model satisfies the preset accuracy condition, and finally a model that can accurately predict the cinder growth trend is obtained.

[0095] In the embodiment, by obtaining and utilizing historical data of boiler operation, the historical coking residue distribution characteristic parameters directly related to the coking state are extracted first, and then a neural network model is constructed and optimized according to the super parameters, so that the coking residue growth prediction model can be obtained under the condition of meeting the preset prediction error. The prediction method based on deep learning improves the coking residue monitoring from traditional static observation and experience judgment to dynamic prediction driven by data, and improves the accuracy of the coking control parameters.

[0096] Optionally, the S104 specifically comprises:

[0097] According to the coking residue growth trend, different combinations of the sound wave control parameters and the steam control parameters are set by using the digital twin model of the boiler to perform multi-round simulation, and the coking efficiency and system energy consumption data under each combination are obtained.

[0098] The digital twin model is a digital twin model reflecting the thermal-structure coupling relationship of the boiler, which is constructed based on the structural data and operating condition parameters of the boiler.

[0099] Based on the data obtained by the multi-round simulation, a multi-objective optimization function is created with the maximum coking efficiency and the minimum energy consumption as the target.

[0100] The non-dominated sorting genetic algorithm is used to process the multi-objective optimization function to generate a Pareto optimal solution set of the sound wave control parameters and the steam control parameters.

[0101] The model-free reinforcement learning algorithm based on the action value function is used to determine the initial sound wave and steam control parameter combination matched with the current operating condition from the Pareto optimal solution set.

[0102] The Bayesian optimization algorithm is applied to adjust the initial sound wave and steam control parameter combination to obtain the sound wave control parameters and the steam control parameters.

[0103] Here, the digital twin model is constructed based on the boiler structure data and operating condition parameters, and is a calculation model for representing the thermal-structure coupling relationship of the boiler. It can reproduce the coking behavior and energy consumption response of the entity boiler under the action of different sound waves and steam parameters in a virtual environment, and is used for multi-round simulation and performance evaluation. The boiler structure data can be obtained through design and operation archives, such as heating surface geometry, heat transfer structure, pressure part arrangement, and ash removal device position, which are used to construct the geometry and boundary conditions of the digital twin model. The operating condition parameters can be obtained through historical / real-time DCS / SCADA systems, such as boiler load, fuel properties, flue gas temperature and flow rate, steam parameters (pressure, temperature, flow rate), air volume ratio, etc., which are used to define the operating condition input and boundary conditions of the digital twin model.

[0104] The multi-objective optimization function is a mathematical model that considers multiple interdependent objectives (in this case, descaling efficiency and energy consumption) to find the best balance point between multiple objectives. For example, in this embodiment, objective 1 of the multi-objective optimization function can be to maximize descaling efficiency, and objective 2 can be to minimize energy consumption, with the Pareto frontier being used to evaluate conflicts.

[0105] The Non-dominated Sorting Genetic Algorithm II (NSGA-II) is an efficient evolutionary algorithm for handling multi-objective optimization problems by maintaining population diversity through non-dominated sorting and crowding distance calculation.

[0106] The model-free reinforcement learning algorithm based on action value function refers to a reinforcement learning method (such as Q-Learning) that does not rely on an environmental model and directly learns the state-action value function Q(s, a), which is used to select an initial parameter combination that better matches the current operating conditions from candidate solutions.

[0107] Bayesian optimization is a global optimization method for black-box functions with high computational cost, which balances exploration of unknown parameter space and utilization of known optimal points by constructing a probabilistic model using a Gaussian process.

[0108] Here, the acoustic wave control parameters refer to a set of adjustable control quantities used to drive the acoustic wave cleaning / decoking unit, such as parameterized descriptions of frequency, amplitude (sound pressure level), action duration, duty cycle, injection / radiation timing, phase relationship, etc. The steam control parameters refer to a set of adjustable control quantities used to drive the steam cleaning / decoking unit, such as parameterized descriptions of steam pressure, flow rate, temperature, injection duration, injection interval, valve opening, injection sequence, etc.

[0109] Specifically, first, based on the digital twin model of the boiler thermal-structure coupling, multiple simulations are performed for different combinations of acoustic wave and steam parameters to obtain descaling efficiency and energy consumption data. Then, a multi-objective optimization function is constructed with the objectives of maximizing descaling efficiency and minimizing energy consumption. Next, the non-dominated sorting genetic algorithm is used to process the optimization function to generate a set of Pareto optimal solutions representing different efficiency-energy consumption balances. Subsequently, the model-free reinforcement learning algorithm based on action value function is used to select the most suitable initial parameter combination from the optimal solution set according to the current boiler operating conditions. Finally, the Bayesian optimization algorithm is applied to fine-tune the initial solution to obtain the final acoustic wave and steam control parameters

[0110] In this embodiment, the basic data is provided through digital twin simulation; the multi-objective function construction converts discrete data into a continuous model; the NSGA-II algorithm balances multiple objectives to generate a series of optional solutions; the model-free reinforcement learning algorithm based on action value function selects the appropriate initial solution according to the actual working conditions; and the Bayesian optimization performs the final fine tuning to obtain the optimal control parameter combination. Through the multi-level and multi-algorithm collaborative parameter optimization technology architecture, the efficiency and accuracy of control parameter determination are improved.

[0111] In one example, through the parameter optimization driven by digital twin, the synergistic effect of acoustic wave and steam phase difference control precision less than 5° can be achieved, ensuring that the output parameter deviation from the measured value is not more than 2%, and the defocusing efficiency is stably maintained at a high level of more than 80%. The Bayesian optimization based on the synergy of Gaussian process proxy model and expected improvement function, combined with 10-core CPU parallel computing, can complete the complex parameter optimization task within 2 hours, greatly improving the computing efficiency. At the same time, the PID parameter self-tuning of the particle swarm algorithm realizes rapid convergence, enabling the system to quickly respond to changes in working conditions and avoiding the lag problem caused by long optimization time in traditional methods. Further, it can also intelligently match the optimal parameter combination according to different boiler working conditions: automatically reduce the acoustic wave power by 50% and switch to energy-saving mode at low load; increase the frequency to enhance the defocusing effect at high load; and trigger the enhanced coupling mode of acoustic wave and steam when detecting abnormal increase in the thickness of the slag. This differentiated control strategy enables the defocusing system to maintain optimal performance under different working conditions.

[0112] Optionally, the above-mentioned acoustic wave control parameters include the acoustic wave frequency, acoustic wave intensity and acoustic wave coverage angle of the acoustic wave generator, and the steam control parameters include the steam pressure, pulse frequency and jet angle of the steam pulse device.

[0113] Among them, the determination of acoustic wave control parameters: set the acoustic wave frequency parameter generated by the acoustic wave generator to determine the vibration characteristics of the acoustic wave; set the acoustic wave energy generated by the acoustic wave generator to control the force of the acoustic wave; set the acoustic wave radiation direction and spatial coverage range of the acoustic wave generator to ensure that the acoustic wave can accurately act on the target area.

[0114] The determination of steam control parameters: set the steam pressure of the steam pulse device to control the impact energy of the steam; set the time interval and frequency of the steam pulse to control the rhythm of the steam impact; set the jet direction and range of the steam pulse device to ensure that the steam can accurately act on the target area.

[0115] In this embodiment, by clearly defining the specific dimensions of the sound wave and steam control parameters, a "sound-steam synergy, time-space precision" multi-dimensional parameter control system is constructed, solving the technical defects of traditional deslagging methods such as single parameter and extensive control. Through the fine control method of multiple parameters and multiple dimensions, the deslagging system can realize the precise removal of different parts and different degrees of coke, ensuring the deslagging effect, avoiding energy waste and equipment damage, and improving the efficiency, precision and safety of the boiler deslagging.

[0116] Optionally, the S105 specifically includes: based on the sound wave control parameters and the steam control parameters, synchronizing the working cycle of the sound wave generator and the steam pulse device through the PID controller, and controlling the phase difference between the sound wave and the steam pulse to meet the preset range.

[0117] Specifically, the sound wave control parameters (sound wave frequency, sound wave intensity and sound wave coverage angle) and the steam control parameters (steam pressure, pulse frequency and jet angle) determined by the digital twin simulation and multi-layer optimization method are input into the control system; a proportional-integral-derivative (PID) controller is configured, the control target, control parameter and feedback mechanism are set, and a closed-loop control system of the sound wave generator and the steam pulse device is established; the running time sequence of the sound wave generator and the steam pulse device is adjusted through the PID controller to ensure that the working cycles of the two devices remain synchronized, realizing the time coordination of sound wave action and steam impact; the phase difference between the sound wave and the steam pulse is accurately adjusted to ensure that the phase difference is maintained within the preset range, optimizing the time sequence matching relationship of the sound wave and the steam.

[0118] In this embodiment, the proportional, integral and derivative control actions of the PID controller enable the system to quickly respond to parameter changes, eliminate steady-state errors and suppress overshoot, ensuring that the sound wave and steam devices operate accurately according to the optimized parameters; working cycle synchronization ensures that the two deslagging mechanisms coordinate in the time dimension, avoiding functional conflicts; phase difference control realizes the optimal time sequence matching of sound wave and steam action, forming a synergistic effect of "sound wave loosening-steam impact". The sound wave-steam collaborative deslagging is upgraded from simple independent parameter control to systematic collaborative control level, significantly improving the coke peeling efficiency through the optimal phase matching of wave-steam, reducing the overall energy consumption, and improving the stability and reliability of the control system

[0119] In one specific embodiment, as shown in Figure 2 A boiler targeted deslagging system based on multi-modal sound-heat coupling and intelligent monitoring mainly includes: a data acquisition layer, a data transmission layer, a data processing layer, a model prediction layer, a decision control and execution optimization layer, and an execution layer.

[0120] Data acquisition layer: includes acoustic wave sensors and infrared thermal imaging equipment, respectively collecting acoustic wave reflection signals and temperature distribution data inside the boiler, forming a multi-modal raw data basis.

[0121] Data transmission layer: the collected data is input to the subsequent processing system through the transmission channel, ensuring timely and reliable data flow.

[0122] Data processing layer: contains multi-source data fusion and cloud storage architecture, realizing the unification, synchronization and efficient management of data from different sources and formats. Signal denoising, feature engineering and LSTM network training functions respectively process data signals to improve signal quality, extract key features, and use deep learning models to train and analyze data. That is, the data processing layer is configured to perform fusion processing on the multi-modal coke distribution data to obtain coke distribution characteristic parameters inside the boiler.

[0123] Model prediction layer: predicting the growth trend of coke based on the processed feature data to guide subsequent control actions. That is, configured to perform, based on the coke distribution characteristic parameters, determining a coke growth trend using a coke growth prediction model.

[0124] Decision control and execution optimization layer: according to the prediction results, the efficiency of coke removal and energy consumption are predicted, and intelligent decision and optimization of coke removal control parameters (such as through multi-objective optimization, reinforcement learning, PID parameter tuning, etc.) are performed to improve the adaptive and collaborative control capabilities of the system. That is, configured to perform, based on the coke growth trend, determining acoustic wave control parameters and steam control parameters using digital twin simulation technology.

[0125] Execution layer: control the acoustic wave generator, steam pulse device and other coke removal execution mechanisms to complete the real-time and efficient removal of coke in a closed loop. That is, configured to perform, based on the acoustic wave control parameters and steam control parameters, controlling the acoustic wave generator and steam pulse device to perform targeted coke removal on the boiler.

[0126] The above-mentioned boiler targeted coke removal system based on multi-modal acoustic-thermal coupling and intelligent monitoring realizes the whole process perception, analysis, decision and optimization control of boiler coke removal.

[0127] In one specific example, the above-mentioned boiler coke removal method can specifically include the following steps in actual application:

[0128] Step 1. Real-time acquisition of boiler operating parameters, including acoustic wave reflection signals, infrared radiation signals and coke thickness data, through Internet of Things sensors, to build a multi-dimensional coke removal feature data set. That is, collecting multi-modal coke distribution data inside the boiler.

[0129] Specifically, high-temperature piezoelectric ceramic acoustic wave sensor arrays (pitch 3-4 m) can be deployed in the low-temperature reheater, low-temperature superheater, air preheater, and economizer area of the horizontal flue and vertical flue, and endoscopic high-temperature infrared thermal imagers can be arranged at the locations prone to coking such as water cooling walls to collect three-dimensional acoustic wave reflection signals and infrared radiation signals in real time and monitor the thickness of the cinder synchronously.

[0130] Step 2. The original data is smoothed and denoised using the Savitzky-Golay filtering algorithm, and the acoustic main frequency, steam pulse energy, and cinder distribution density are extracted by Min-Max normalization to unify the dimension. That is, the multi-modal cinder distribution data is fused to obtain the cinder distribution characteristic parameters inside the boiler.

[0131] Specifically, the training data can cover the full operating parameters of 300 MW and above units, and the coking removal effect indicators include the improvement range of coking removal efficiency and energy consumption.

[0132] Step 3. A cinder growth prediction model is established based on the LSTM neural network to dynamically analyze the influence of acoustic vibration mode, steam pulse intensity, and action time on cinder removal efficiency. That is, the cinder growth trend is determined using the cinder growth prediction model; based on the cinder growth trend, different combinations of acoustic control parameters and steam control parameters are set to perform multiple simulations using the digital twin model of the boiler to obtain the coking removal efficiency and system energy consumption data under each combination.

[0133] Specifically, the number of input layer nodes of the model is set to 12 (corresponding to key features), the hidden layer uses a bidirectional LSTM structure, and the number of output layer nodes is 3 (predicting coking removal efficiency, energy consumption, and emissions.

[0134] Step 4. Simulate the operating conditions of the boiler by combining digital twin technology, and optimize the power and action parameters of the acoustic generator in real time through the PID controller. That is, the acoustic control parameters and steam control parameters are determined through digital twin simulation technology.

[0135] Specifically, the steam nozzle adopts a spiral flow channel structure (spiral angle 30°) to enhance the uniformity of steam coverage, and the nozzle material is selected to be 316L stainless steel (wear-resistant, corrosion resistance ≥10 years).

[0136] Step 5. The high-frequency acoustic wave destroys the microstructure of the cinder, and the pulsed steam is injected synchronously to penetrate the cracks, and the thermal stress peeling effect is used to achieve targeted removal. That is, the working cycle of the acoustic generator and the steam pulse device is synchronized through the PID controller, and the phase difference between the acoustic wave and the steam pulse is controlled to meet the preset range.

[0137] Specifically, the acoustic action timing and steam pulse are triggered synchronously, and the phase difference is controlled within ±5° to maximize the coupling efficiency.

[0138] Step 6. Based on the de-coke efficiency, energy consumption, and emission data output by the prediction model, the system adjusts the frequency of the acoustic wave generator, the steam jet pressure, and the action area in real time.

[0139] When an abnormal increase in the thickness of the coke is monitored, the system automatically triggers the enhanced coupling mode of high-frequency acoustic waves and pulsed steam. During the low-load operation stage of the boiler, the system switches to the energy-saving mode, reduces the power of the acoustic waves, and optimizes the timing of the steam jet. All adjustment strategies are executed through a closed-loop PID controller, while dynamically matching the changes in the operating conditions of the boiler to achieve dual optimization of energy efficiency and environmental performance.

[0140] Specifically, for data with a prediction error > 5%, iterative correction is performed, and the Bayesian optimization algorithm is used to adjust the model hyperparameters to ensure that the output deviation from the measured value is ≤ 2%.

[0141] In this embodiment, by constructing a multi-modal acoustic-thermal coupling and intelligent monitoring boiler targeted de-coke system, the technical bottleneck of traditional de-coke technology, which relies on manual experience and lacks dynamic optimization capability, is broken through. The significant advantages are as follows: (1) Efficiency improvement: the de-coke efficiency is improved to 80%, the action depth reaches 30-40 cm, the radius reaches 1.2-1.5 m, the cleaning speed is 0.5-1 times faster than traditional methods, and the loss caused by non-stop of the unit due to large-area coking is eliminated; (2) Life extension: physical damage to the heating surface caused by mechanical de-coke is avoided, the equipment life is extended by 2-3 years, and the maintenance cost is reduced by 40%; (3) Intelligent decision-making: a coke prediction model based on an LSTM neural network (error < 5%) is combined with digital twin technology to simulate the operating conditions, making the decision-making process scientific and the response real-time, and the production efficiency is improved by more than 30%. This provides a transformation opportunity for enterprises from "passive response" to "active prevention", not only strengthening the stability and thermal efficiency of the boiler, but also significantly enhancing the market competitiveness and sustainable development capability of the enterprise through technological upgrading.

[0142] Further, the present application also provides a boiler targeted de-coke system based on multi-modal acoustic-thermal coupling and intelligent monitoring, which comprises an acoustic wave resonance module, a steam pulse module, an intelligent monitoring and early warning module, and a closed-loop control module.

[0143] The acoustic wave resonance module uses frequency conversion control to generate directional resonance waves to destroy the structure of the coke. The steam pulse module nozzle is designed with a spiral flow channel structure to enhance the steam penetration ability. The intelligent monitoring and early warning module integrates an LSTM neural network algorithm with a prediction error ≤ 5% and supports a three-level early warning mechanism for predicting the growth rate of the coke. The closed-loop control module uses digital twin technology to dynamically optimize parameters with an optimization period ≤ 5 seconds to determine the acoustic wave control parameters and the steam control parameters. The acoustic wave generator uses high-temperature-resistant piezoelectric ceramic materials to adapt to the high-temperature environment of the boiler.

[0144] The sound wave generator is arranged in the low-temperature reheater, low-temperature superheater, air preheater and coal economizer area of the horizontal flue and vertical flue, the steam nozzle is installed in the boiler water wall area, the action range covers the whole furnace coking area, the sound wave soot blower is responsible for large-area preventive soot removal, the steam soot blower is used for deep removal of local serious coking area, and a collaborative mode of "comprehensive coverage + key breakthrough" is formed.

[0145] The application of the system in the coal-fired power plant boiler supports online decoking of units with a load of 50% or more, and the decoking efficiency is stable at more than 80%.

[0146] In another specific embodiment, the above-mentioned boiler decoking method can specifically include the following steps:

[0147] Step 1, collecting multi-modal coking residue distribution data.

[0148] When the proportion of high-alkali coal mixed with the boiler of a unit is more than 60%, a steam soot blower is arranged in the water wall of the boiler, a high-temperature-resistant piezoelectric ceramic sound wave sensor array is arranged in the low-temperature reheater, low-temperature superheater, air preheater and coal economizer area of the horizontal flue and vertical flue, and an endoscopic high-temperature infrared thermal imager is arranged in the coking prone position such as the water wall, so as to collect three-dimensional sound wave reflection signals and infrared radiation signals in real time and synchronously monitor the thickness of the coking residue. The sensor data is subjected to preliminary filtering treatment through an edge gateway, is transmitted to a data center through an MQTT protocol, and the original data sampling frequency is ensured to be greater than or equal to 1 kHz.

[0149] Step 2, data fusion and storage architecture.

[0150] A multi-element data fusion-cloud storage architecture is adopted, the sound wave, infrared and other multi-source data are processed through a Kalman filtering algorithm, the synchronization problem of different sampling rates is solved, a multi-dimensional data set aligned in time and space is constructed, and the prediction accuracy of the model is improved. The cloud storage architecture adopts an InfluxDB time series database to store the original data (retention policy 7 days), and establishes a Redis hot data cache area (memory allocation 8 GB), uses columnar storage to optimize the time series query efficiency, and uses an in-memory database to accelerate real-time calculation. The system requires a write throughput of greater than or equal to 10,000 points / second, a query response time of less than 100 ms, a data storage reliability of greater than or equal to 99.9%, and a cache hit rate of greater than or equal to 80%, so as to support large-scale historical data analysis and realize continuous iteration and optimization of the model.

[0151] Step 3, signal denoising processing and feature extraction.

[0152] Savitzky-Golay filter is used for signal denoising processing. Local polynomial fitting is used to retain the high frequency characteristics of the signal and suppress white noise, aiming to eliminate the background noise of the boiler (mechanical vibration, fluid turbulence interference) and improve the signal quality. At the same time, when constructing the feature engineering, the main frequency of the sound wave, the steam pulse energy, the distribution density of the slag, etc. are extracted. Principal component analysis (PCA) is used for dimensionality reduction processing, retaining 95% of the variance contribution rate and eliminating feature redundancy, in order to establish a feature set that is strongly related to the deslagging efficiency and improve the model generalization ability.

[0153] Step 4, establish the LSTM slag growth prediction model.

[0154] Based on LSTM neural network, a slag growth prediction model is established to dynamically analyze the influence of sound wave vibration mode (sine wave / square wave), steam pulse intensity and action time on the slag removal efficiency, and to simultaneously predict the deslagging efficiency.

[0155] Step 5, model construction based on digital twinning technology.

[0156] Based on digital twinning

[0157] Technology builds a boiler thermal-structure coupling model. Through the fusion of finite element analysis and thermodynamic equations, the temperature field, flow field and component thermal stress distribution in the furnace are simulated. Real-time sensor data is used to ensure that the virtual and physical system states are consistent. A multi-objective optimization function including deslagging efficiency and energy consumption is designed. NSGA-II algorithm is used to generate a Pareto optimal solution set. Q-Learning reinforcement learning framework is used to dynamically select control strategies. Q table is updated every 10 minutes to balance exploration and utilization. Finally, particle swarm algorithm is used to online tune PID parameters, realizing multi-objective collaborative optimization and adaptive control.

[0158] Step 6, control of multi-modal actuators, including parameter adjustment and closed-loop control.

[0159] The multi-modal coupling actuator is adopted, the frequency modulation technology is adopted to dynamically adjust the sound wave intensity, the spiral flow channel nozzle is configured for the steam pulse device to enhance the coverage uniformity, the PID control is adopted to form the phase difference of less than 5 degrees of the steam pressure and the sound wave time sequence, the data with the prediction error greater than 5% is iteratively corrected, the Bayesian optimization algorithm is adopted to adjust the model hyperparameters, and the deviation between the output and the measured value is ensured to be less than or equal to 2%. The Ziegler-Nichols critical proportion method is adopted to initialize the parameters in the PID parameter self-tuning, the particle swarm optimization algorithm is adopted to iteratively adjust the comprehensive error index as the fitness function, the particle swarm size is 20, the inertia weight is 0.729, and the iteration is performed until convergence or the maximum number of iterations is 100; the Bayesian optimization establishes a Gaussian process proxy model for parameters such as sound wave frequency and steam pressure, uses the expected improvement function to balance exploration and utilization, and completes the parameter adjustment within 2 hours by using a 10-core CPU in parallel; in the closed-loop execution, when the prediction error is greater than 5%, the Bayesian module is triggered to retrain, the sound wave power is reduced by 50% at low load and the frequency is increased at high load to dynamically match the working conditions, and the output deviation is ensured to be less than 2%;

[0160] Step 7, mode switching for different working conditions.

[0161] When the abnormal increase of the coke thickness is monitored, the reinforced coupling mode of the high-frequency sound wave and the pulse steam is automatically triggered; in the low-load operation stage of the boiler, the energy-saving mode is switched to, the sound wave power is reduced, and the steam injection time sequence is optimized. All adjustment strategies are executed in a closed loop by a PID controller to ensure that the coke removal efficiency is stably above 80%, while dynamically matching the changes of the boiler working conditions to realize the dual optimization of energy efficiency and environmental protection performance.

[0162] Embodiment 2:

[0163] As shown in Figure 3 The present embodiment provides a boiler decoking device for executing the above-mentioned boiler decoking method, comprising:

[0164] The acquisition module 301 is configured to acquire multi-modal coke distribution data inside the boiler; the multi-modal coke distribution data is original state information of the coke directly obtained by a multi-source sensor;

[0165] The acquisition module 301 is configured to acquire multi-modal coke distribution data inside the boiler; the multi-modal coke distribution data is original state information of the coke directly obtained by a multi-source sensor;

[0166] The first determination module 303 is connected with the acquisition module 302 and is configured to determine the growth trend of the coke based on the coke distribution characteristic parameters and by using a coke growth prediction model; the coke growth prediction model is configured to predict the dynamic change trend of the coke;

[0167] The second determining module 304 is connected with the first determining module 303, and determines the sound wave control parameter and the steam control parameter according to the focus slag growth trend through a digital twin simulation technology; the digital twin technology is a simulation technology for simulating the physical process of the action of sound waves and steam on focus slag and optimizing parameters through a digital twin model reflecting the thermal-structure coupling relationship of the boiler;

[0168] The control module 305 is connected with the second determining module 304, and is configured to control the sound wave generator and the steam pulse device to perform targeted deslagging on the boiler based on the sound wave control parameter and the steam control parameter.

[0169] Optionally, the multi-modal focus slag distribution data comprises: sound wave reflection signals, infrared thermal imaging signals, steam parameter signals and boiler operating condition signals,

[0170] The sound wave reflection signals are collected by a high-temperature piezoelectric ceramic sound wave sensor array arranged in the horizontal flue and the vertical flue area; the infrared thermal imaging signals are collected by an endoscopic high-temperature infrared thermal imager arranged at the water-cooled wall and other easy-coking parts; the steam parameter signals are collected by pressure, flow and temperature sensors arranged on the steam pipeline; and the boiler operating condition signals are collected by a distributed control system.

[0171] Optionally, the acquisition module 302 comprises:

[0172] The first processing unit is configured to perform time-space alignment on the multi-modal focus slag distribution data by using a Kalman filtering algorithm to obtain first distribution data;

[0173] The second processing unit is configured to perform filtering and denoising processing on the first distribution data to obtain second distribution data;

[0174] The third processing unit is configured to perform principal component analysis dimension reduction processing on the second distribution data to obtain third distribution data;

[0175] The fourth processing unit is configured to perform normalization processing on the third distribution data by using a deviation standardization method to obtain target distribution data;

[0176] The first extraction unit is configured to extract a focus slag distribution characteristic parameter from the target distribution data.

[0177] Optionally, the first extraction unit comprises:

[0178] The first extraction sub-unit is configured to determine the focus slag thickness by time-frequency of the sound wave reflection signals in the target distribution data;

[0179] The second extraction sub-unit is configured to determine the temperature gradient data of the focus slag by the infrared thermal imaging signals in the target distribution data;

[0180] a third extraction subunit configured to determine a cinder distribution density according to the temperature gradient data;

[0181] a fourth extraction subunit configured to determine a cinder growth rate according to the change data of the cinder thickness in the continuous time period;

[0182] a fifth extraction subunit configured to determine a cinder thermal resistance value according to the steam parameter signal in the target distribution data and the temperature gradient data,

[0183] wherein the cinder distribution characteristic parameters include the cinder thickness, the cinder distribution density, the cinder growth rate and the cinder thermal resistance value.

[0184] Optionally, the device further comprises:

[0185] an acquisition unit configured to acquire boiler operation historical data, the historical data including historical collected acoustic wave reflection signals, infrared thermal imaging signals, steam parameters and boiler working condition parameters;

[0186] a second extraction unit configured to extract historical cinder distribution characteristic parameters based on the boiler operation historical data;

[0187] a first creation unit configured to construct a neural network model according to the historical cinder distribution characteristic parameters;

[0188] an adjustment unit configured to adjust hyperparameters of the neural network model until a prediction error meets a preset condition, to obtain a cinder growth prediction model.

[0189] Optionally, the second determination module 304 comprises:

[0190] a simulation unit configured to set different combinations of the acoustic wave control parameters and the steam control parameters for multiple rounds of simulation by using a digital twin model of the boiler according to the cinder growth trend, to obtain descaling efficiency and system energy consumption data under each combination, wherein the digital twin model is a digital twin model reflecting a thermal-structure coupling relationship of the boiler, which is constructed based on structure data and operation working condition parameters of the boiler;

[0191] a second creation unit configured to create a multi-objective optimization function aiming at maximizing descaling efficiency and minimizing energy consumption based on the data obtained from the multiple rounds of simulation;

[0192] a generation unit configured to process the multi-objective optimization function by using a non-dominated sorting genetic algorithm II to generate a Pareto optimal solution set of the acoustic wave control parameters and the steam control parameters;

[0193] a solution unit configured to determine an initial acoustic wave and steam control parameter combination matching a current working condition from the Pareto optimal solution set by using a model-free reinforcement learning algorithm based on an action value function;

[0194] The determining unit is configured to apply a Bayesian optimization algorithm to adjust the initial combination of the acoustic wave control parameter and the steam control parameter to obtain the acoustic wave control parameter and the steam control parameter.

[0195] Optionally, the acoustic wave control parameter includes an acoustic wave frequency, an acoustic wave intensity and an acoustic wave coverage angle of the acoustic wave generator, and the steam control parameter includes a steam pressure, a pulse frequency and a jet angle of the steam pulse device.

[0196] Optionally, the control module 305 is specifically configured to: based on the acoustic wave control parameter and the steam control parameter, control a working period of the acoustic wave generator and the steam pulse device to be synchronized by a PID controller, and control a phase difference between the acoustic wave and the steam pulse to satisfy a preset range.

[0197] In the embodiments of the present application, first, multi-modal acquisition means is used to comprehensively acquire the boiler coke distribution information, overcoming the limitations of single-point measurement; then, multi-source heterogeneous data is fused and processed to extract characteristic parameters that can truly reflect the coking state; based on the characteristic parameters, a prediction model is used to predict the coke growth trend; then, digital twin technology is used to optimize the combination of acoustic wave and steam control parameters in a virtual environment; finally, precise targeted collaborative control of the acoustic wave generator and the steam pulse device is realized. The efficiency of boiler decoking is improved.

[0198] Embodiment 3

[0199] Figure 4 A hardware structure schematic diagram of an electronic device provided by an embodiment of the present application is shown.

[0200] The electronic device can include a processor 401 and a memory 402 having stored computer program instructions.

[0201] Specifically, the processor 401 can include a central processing unit (CPU), or a specific integrated circuit (Application Specific Integrated Circuit, ASIC), or can be configured to implement one or more integrated circuits of the embodiments of the present application.

[0202] The memory 402 can include a mass storage for data or instructions. By way of example and not limitation, the memory 402 can include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (Universal Serial Bus, USB) drive, or a combination of two or more of these. Where appropriate, the memory 402 can include removable or non-removable (or fixed) media. Where appropriate, the memory 402 can be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, the memory 402 is a non-volatile solid-state memory.

[0203] The memory can include read-only memory (ROM), random access memory (RAM), magnetic disk storage mediums devices, optical storage mediums devices, flash memory devices, electrical, optical, or other physical / tangible memory storage devices. Thus, generally, the memory includes one or more tangible (non-transitory) computer-readable storage mediums (e.g., memory devices) encoded with software that, when executed (e.g., by one or more processors), is operable to perform the operations described with reference to the methods according to an aspect of the present disclosure.

[0204] The processor 401 implements the above-described any one of the methods of the boiler decoking by reading and executing computer program instructions stored in the memory 402.

[0205] In one example, the electronic device can further include a communication interface 403 and a bus 404. Wherein, as shown, the processor 401, the memory 402, the communication interface 403 are connected through the bus 404 and complete the communication between each other. Figure 4

[0206] The communication interface 403 is mainly used to realize the communication between each module, device, unit and / or equipment in the embodiments of the present application.

[0207] The bus 404 includes hardware, software or both to couple components of the online data traffic billing device to each other. By way of example, and not limitation, the bus can include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand (IB) interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association local (VLB) bus, or another suitable bus or a combination of two or more of these. Where suitable, the bus 404 can include one or more buses. Although specific buses are described and illustrated in the embodiments of the present application, the present application contemplates any suitable bus or interconnect.

[0208] In addition, in combination with the above-described method of the boiler decoking, the embodiments of the present application can provide a computer storage medium to implement. The computer storage medium has computer program instructions stored thereon; the computer program instructions are executed by the processor to implement any one of the above-described methods of the boiler decoking.

[0209] ​It is to be understood that the application is not limited to particular configurations and processes described herein and shown in the drawings. The detailed description is not to be taken as limiting the application. In the above embodiments, several specific steps are described and illustrated in order to provide a thorough understanding of the application. However, the application can be practiced with fewer or additional steps, and in a different order. The application is not limited to the described and illustrated steps.

[0210] The functions indicated in the structural block diagrams above can be implemented in hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application specific integrated circuits (ASICs), appropriate firmware, plug-ins, functional cards, and the like. When implemented in software, the elements of the application are program or code segments that are used to perform the required tasks. The program or code segments can be stored in a machine-readable medium or transmitted through a data signal carried in a carrier wave over a transmission medium or a boiler decoking link. The "machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, and the like. The code segments can be downloaded via a computer network such as the Internet, an intranet, and the like.

[0211] It is also to be understood that the example embodiments described in this application are based on a series of steps or devices to describe some methods or systems. However, the application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiments, or in an order different from the embodiments, or several steps can be performed simultaneously.

[0212] The computer program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other processing devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other processing devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer program instructions can also be stored in a computer readable medium that can direct a computer, other programmable data processing apparatus, or other processing devices to operate in a particular manner, such that the instructions stored in the computer readable medium produce an article of manufacture including instructions which implement the function / act specified in the flowchart and / or block diagram block or blocks. The computer program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other processing devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other processing devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer program instructions can also be stored in a computer readable medium that can direct a computer, other programmable data processing apparatus, or other processing devices to operate in a particular manner, such that the instructions stored in the computer readable medium produce an article of manufacture including instructions which implement the function / act specified in the flowchart and / or block diagram block or blocks.

[0213] The above is merely specific implementation of the present application, and those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the foregoing method embodiments, which will not be described herein again. It should be understood that the protection scope of the present application is not limited to this, and any person skilled in the art can easily think of various equivalent modifications or replacements within the technical range disclosed in the present application, and these modifications or replacements should be covered within the protection scope of the present application.

Claims

1. A method of decoking a boiler, characterized by, The method comprises: Collecting multi-modal coke distribution data inside the boiler; the multi-modal coke distribution data is coke original state information obtained by a multi-source sensor; Fusing the multi-modal coke distribution data to obtain coke distribution characteristic parameters inside the boiler; the coke distribution characteristic parameters are used to represent the physical state of the coke and its evolution law; Based on the coke distribution characteristic parameters, a coke growth prediction model is used to determine the coke growth trend; the coke growth prediction model is used to predict the dynamic change trend of the coke; According to the coke growth trend, a digital twin simulation technology is used to determine the sound wave control parameter and the steam control parameter; the digital twin technology is a simulation technology that reflects the digital twin model of the boiler thermal-structure coupling relationship, simulates the physical process of the sound wave and the steam on the coke and optimizes the parameters; Based on the sound wave control parameter and the steam control parameter, a sound wave generator and a steam pulse device are controlled to target the coke removal of the boiler.

2. The method of claim 1, wherein, The multi-modal coke distribution data includes: sound wave reflection signals, infrared thermal imaging signals, and steam parameter signals, Wherein, the sound wave reflection signals are collected by a high-temperature piezoelectric ceramic sound wave sensor array arranged in the horizontal flue and the vertical flue area; the infrared thermal imaging signals are collected by an endoscopic high-temperature infrared thermal imager arranged at the easy coking position; the steam parameter signals are collected by pressure, flow and temperature sensors arranged on the steam pipeline.

3. The method of claim 2, wherein, The multi-modal coke distribution data is fused to obtain coke distribution characteristic parameters inside the boiler, specifically including: A Kalman filtering algorithm is used to align the multi-modal coke distribution data in time and space to obtain first distribution data; The first distribution data is filtered and denoised to obtain second distribution data; The second distribution data is processed by principal component analysis dimension reduction to obtain third distribution data; The third distribution data is normalized by a dispersion standardization method to obtain target distribution data; Coke distribution characteristic parameters are extracted from the target distribution data.

4. The method of claim 3, wherein, The coke distribution characteristic parameters are extracted from the target distribution data, specifically including: The thickness of the coke is determined by the time-frequency of the sound wave reflection signals in the target distribution data; The temperature gradient data of the coke is determined by the infrared thermal imaging signals in the target distribution data; The coke distribution density is determined according to the temperature gradient data; The growth rate of the coke is determined by the change data of the coke thickness in a continuous time period; The thermal resistance value of the coke is determined by the steam parameter signals in the target distribution data and the temperature gradient data, Wherein, the coke distribution characteristic parameters include coke thickness, coke distribution density, coke growth rate, and coke thermal resistance value.

5. The method according to any one of claims 1 to 4, characterized in that, The method further comprises: Obtaining historical data of the boiler operation, the historical data including historically collected sound wave reflection signals, infrared thermal imaging signals, and steam parameters; Based on the historical data of the boiler operation, historical coke distribution characteristic parameters are extracted; According to the historical coke distribution characteristic parameters, a neural network model is constructed; Adjusting hyperparameters of the neural network model until a prediction error meets a preset condition to obtain the cinder growth prediction model.

6. The method according to any one of claims 1-4, characterized in that, The acoustic wave control parameters and the steam control parameters are determined based on the cinder growth trend through a digital twin simulation technology, and specifically include: According to the cinder growth trend, different combinations of acoustic wave control parameters and steam control parameters are set for multiple rounds of simulation using a digital twin model of the boiler, and cinder removal efficiency and system energy consumption data under each combination are obtained; the digital twin model is a digital twin model reflecting the thermal-structure coupling relationship of the boiler, which is constructed based on structural data and operating condition parameters of the boiler; Based on the cinder removal efficiency and system energy consumption data under each combination, a multi-objective optimization function is created with the goal of maximizing cinder removal efficiency and minimizing energy consumption; The multi-objective optimization function is processed using a non-dominated sorting genetic algorithm II to generate a Pareto optimal solution set of acoustic wave control parameters and steam control parameters; An action value function-based model-free reinforcement learning algorithm is used to determine an initial acoustic wave and steam control parameter combination from the Pareto optimal solution set that matches the current operating condition; A Bayesian optimization algorithm is applied to adjust the initial acoustic wave and steam control parameter combination to obtain acoustic wave control parameters and steam control parameters.

7. The method according to any one of claims 1-4, characterized in that, The acoustic wave control parameters include acoustic wave frequency, acoustic wave intensity, and acoustic wave coverage angle of the acoustic wave generator, and the steam control parameters include steam pressure, pulse frequency, and jet angle of the steam pulse device.

8. The method of claim 7, wherein, The acoustic wave control parameters and the steam control parameters are used to control the acoustic wave generator and the steam pulse device, and specifically include: Based on the acoustic wave control parameters and the steam control parameters, a PID controller is used to synchronize the working periods of the acoustic wave generator and the steam pulse device and control the phase difference between the acoustic wave and the steam pulse to meet a preset range.

9. A boiler decoking apparatus characterized by, The device includes: A collection module is configured to collect multi-modal cinder distribution data inside the boiler; the multi-modal cinder distribution data is original state information of the cinder directly obtained by a multi-source sensor; An acquisition module is connected to the collection module and is configured to fuse and process the multi-modal cinder distribution data to obtain cinder distribution characteristic parameters inside the boiler; the cinder distribution characteristic parameters are used to represent the physical state of the cinder and its evolution law; A first determination module is connected to the acquisition module and is configured to determine a cinder growth trend based on the cinder distribution characteristic parameters using a cinder growth prediction model; the cinder growth prediction model is used to predict the dynamic change trend of the cinder; A second determination module is connected to the first determination module and is configured to determine acoustic wave control parameters and steam control parameters based on the cinder growth trend through a digital twin simulation technology; the digital twin technology is a simulation technology that simulates the physical process of the action of acoustic waves and steam on cinder and optimizes parameters through a digital twin model reflecting the thermal-structure coupling relationship of the boiler; A control module is connected to the second determination module and is configured to control an acoustic wave generator and a steam pulse device based on the acoustic wave control parameters and the steam control parameters to perform targeted cinder removal on the boiler.

10. An electronic device, comprising: A computer program product comprising a computer readable storage medium having computer program instructions embodied therewith, the computer program instructions, when executed by a processor, implement the method of any of claims 1-8.

11. A computer readable storage medium, characterized in that, A computer program product comprising a computer readable storage medium having computer program instructions embodied therewith, the computer program instructions, when executed by a processor, implement the method of any of claims 1-8.

Images