Intelligent thermal boiler equipment monitoring control system

Abstract

The invention discloses an intelligent thermal boiler equipment monitoring and control system, relates to the technical field of monitoring and control, and solves the problem that noise source localization and noise pollution degree values are difficult to analyze after the operation state of equipment is identified. A self-attention mechanism is difficult to construct and is combined with a survival analysis model to predict the fault probability; various monitoring data are difficult to analyze so as to comprehensively evaluate a risk state index; the heat is difficult to dynamically distribute and recover; and a closed-loop feedback mechanism is difficult to generate through a digital twin verification platform. Through combination of the modules, comprehensive monitoring of thermal boiler equipment and comprehensive assessment of risk state indexes are realized, so that heat distribution is optimized, remote intelligent control is realized, and equipment operation efficiency and safety are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of monitoring control, and specifically relates to an intelligent heat boiler equipment monitoring control system. BACKGROUND

[0002] The heat boiler equipment is a core component of the heat equipment, is mainly used for converting fuel or electric energy into heat energy, producing hot water, steam or hot air, and is widely applied to the industrial, civil heating and hot water supply fields. With the development of industrial automation and informatization, the heat boiler equipment needs more advanced intelligent control algorithm and optimization technology to realize accurate adjustment of the boiler operation parameters, optimize heat distribution, and improve the overall performance and operation efficiency of the equipment.

[0003] In the existing intelligent heat boiler equipment monitoring control system, the following problems exist, including: it is difficult to identify the equipment operation state, analyze the noise sound source positioning and noise pollution degree value, construct a self-attention mechanism combined with a survival analysis model to predict the fault probability, analyze various monitoring data to comprehensively evaluate the risk state index, and dynamically allocate and recover heat and generate a closed-loop feedback mechanism through a digital twin verification platform. SUMMARY

[0004] To solve the problems existing in the prior art, the first aspect of the present application provides an intelligent heat boiler equipment monitoring control system, comprising the following modules:

[0005] A data acquisition and preprocessing module: real-time monitoring of the heat boiler equipment, acquisition of a plurality of operation data and environment data, and respective preprocessing;

[0006] A data analysis module: according to a deep convolutional neural network model, identifying the equipment operation state, analyzing the noise sound source positioning and calculating the noise pollution degree value; calculating the thermal efficiency, combustion stability coefficient and steam quality index; according to the calculation results and the operation state identification results, constructing a self-attention mechanism combined with a survival analysis model to predict the fault probability; calculating the pollutant emission coefficient by monitoring the discharged pollutants; calculating the corrosion degree value by analyzing the corrosion trend;

[0007] A monitoring and evaluation module: comprehensively evaluating the risk state index of the heat boiler equipment;

[0008] A monitoring and control module: adopting an intelligent control algorithm to dynamically allocate and recover heat; according to the allocation results, generating a closed-loop feedback mechanism through a digital twin verification platform; and through network communication and cloud computing technology, remotely monitoring and controlling the heat boiler equipment.

[0009] Further, according to the deep convolutional neural network model, identifying the equipment operation state, analyzing the noise sound source positioning and calculating the noise pollution degree value, comprising the following steps:

[0010] Collect and label the sound sample data in the historical operation data of the heat boiler equipment under normal operation, fault operation and different load conditions;

[0011] Convert the preprocessed time-domain sound signal into a time-frequency domain image using short-time Fourier transform to generate a spectrogram; at the same time, use wavelet transform method to carry out multi-scale decomposition on the sound signal, extract wavelet coefficients, and rearrange the wavelet coefficients into a pseudo-image, and stack the pseudo-image with the spectrogram to form a multi-channel input;

[0012] Take the multi-channel input as the input data of the deep convolutional neural network model for training, and output the running state of the heat boiler equipment corresponding to the real-time preprocessed running sound data;

[0013] According to the running state recognition result, combined with the physical layout of the heat boiler equipment and the acoustic propagation model, the specific components or positions producing abnormal noise are determined by beamforming or time delay estimation method;

[0014] The formula for calculating the noise pollution degree value is:

[0015] Wherein, σ i is the sound pressure value of the actual measured noise of the i th monitoring point; L max is the maximum noise level; is the actual noise duration of the i th monitoring point; T max is the maximum duration allowed for noise; C is the confidence of the running state recognition result; M is the total number of monitoring points.

[0016] Further: calculate the thermal efficiency, combustion stability coefficient and steam quality index, including the following steps:

[0017] The formula for calculating the thermal efficiency is:

[0018] Wherein, η is the thermal efficiency, Q input is the total heat input into the heat boiler equipment, Q loss is the total energy loss;

[0019] Collect flame image and light intensity time series data by monitoring with optical sensor, preprocess flame image and use U-Net++ architecture to segment flame area at pixel level, calculate flame area and perimeter by combining morphological operation; calculate the combustion stability formula by combining gas composition and boiler operation data:

[0020] Get the combustion stability coefficient R, wherein I(t) is the flame light intensity time series data, A and P are the flame area and perimeter respectively, Nco and are the concentrations of carbon monoxide and carbon dioxide, respectively;

[0021] Calculate the steam quality index by the formula:

[0022] Get the steam quality index Z, where S h is the steam humidity, S c is the salt content of steam, S0 is the impurity content in steam; P 实 is the actual steam pressure, P set is the set steam pressure, P_rated is the rated pressure of the boiler; L represents the energy or mass of steam produced by the boiler.

[0023] Furthermore, based on the above calculation results and the operating status recognition results, a self-attention mechanism is constructed to combine with a survival analysis model to predict the failure probability, including the following steps:

[0024] Collect noise pollution values, thermal efficiency, combustion stability coefficient, and steam quality index of thermal boiler equipment during historical operation. Combine historical operation data with operation status identification results to construct a training dataset and standardize the data.

[0025] Construct a self-attention mechanism combined with a survival analysis model, and input the training data set into the model training. The input time series feature data enters the multi-head attention layer after position encoding, and the output survival function is: Where h(τ) is the hazard rate function based on the physical degradation equation constraints, which is learned during the model training process;

[0026] The training data set collected by real-time monitoring is input into the trained self-attention mechanism combined with the survival analysis model, the output survival function S(t) is calculated in real time, and the failure probability is derived as G(t) = 1-S(t).

[0027] Further: calculating the pollutant emission coefficient by monitoring the emitted pollutants includes the following steps:

[0028] By monitoring the exhaust gas, wastewater and waste residue emitted by thermal boiler equipment during operation and combining it with environmental data, the pollutant emission coefficient formula is calculated as follows:

[0029] Get the pollutant emission coefficient W, where N i is the emission concentration of the i-th pollutant, R X is the fuel consumption, η is the thermal efficiency, T 实 is the actual temperature of the combustion chamber, T0 is the optimal combustion temperature, H T is the ambient temperature, HS is the ambient humidity, n is the total number of pollutant emission species.

[0030] Further: calculating the corrosion degree value by analyzing the corrosion trend, including the following steps:

[0031] Deploying a MEMS corrosion probe array on the inner wall of the boiler to monitor the local pH value, Cl- ion concentration and redox potential in real time; collecting the pitting corrosion occurrence position data of the inner wall of the boiler as the label for graph neural network training;

[0032] Modeling the electrochemical coupling effect of the corrosion microzone in the boiler using a graph neural network, inputting the local pH value, Cl- ion concentration, redox potential for training, and outputting the predicted pitting corrosion priority occurrence position;

[0033] According to the prediction results of the graph neural network, analyze the corrosion rate, corrosion degree data and corrosion trend, and comprehensively analyze the results to calculate the corrosion degree value by weighted average summation.

[0034] Further: comprehensive evaluation of the risk state index of the thermal boiler equipment, including the following steps:

[0035] According to the fault probability, thermal efficiency, combustion stability coefficient, steam quality index, pollutant emission coefficient, corrosion degree value and noise pollution value calculated by real-time monitoring, perform weighted average summation calculation to comprehensively evaluate the risk state index of the thermal boiler equipment; wherein, the weights are distributed by using entropy weight method or analytic hierarchy process;

[0036] According to the risk state index of the thermal boiler equipment, establish a three-level risk early warning mechanism.

[0037] Further: using intelligent control algorithm, dynamically allocate recovered heat, including the following steps:

[0038] Real-time acquisition of boiler operation parameters, synchronous acquisition of steam demand, pressure threshold and heating time requirements;

[0039] Using a fuzzy PID compound control algorithm to construct a dynamic allocation model, the input layer receives the risk state index of the thermal boiler equipment, the combustion efficiency, the heat storage capacity and the pipe network pressure difference parameters, the influence weights of each factor are quantified by the membership function, and the output layer generates a heat allocation coefficient matrix;

[0040] Establish a multi-objective optimization function, with the constraints of minimizing fuel consumption, maximizing waste heat recovery rate and balancing equipment wear, convert the risk index into equipment reliability constraints, and use NSGA-II genetic algorithm to solve the Pareto optimal solution set;

[0041] The dynamic adjustment is performed through a distributed control system (DCS), and the excess heat is intelligently distributed to the waste heat boiler, the heat storage tank or the heat network circulation system according to the optimal solution set heat distribution scheme obtained.

[0042] Further, a closed-loop feedback mechanism is generated through a digital twin verification platform according to the distribution result, including the following steps:

[0043] A digital twin verification platform is constructed to simulate the running state of the heat supply boiler equipment and heat distribution thereof.

[0044] The life attenuation curve of the heat supply boiler equipment under different heat distribution strategies is predicted by using an LSTM neural network.

[0045] The equipment life attenuation curve predicted by the LSTM neural network is compared and analyzed with actual running data, the control parameters are continuously optimized according to the comparison result, and the optimized parameters are fed back to the intelligent control algorithm to form a closed-loop feedback mechanism.

[0046] Compared with the prior art, the present application has the following advantages:

[0047] The present application can accurately identify the running state of the heat supply boiler equipment and analyze the noise source positioning by combining a deep convolutional neural network model with various signal processing methods.

[0048] The present application can predict the failure probability by comprehensively analyzing various running data by constructing a self-attention mechanism combined with a survival analysis model, thereby effectively improving the reliability and safety of equipment operation; the optical sensor is used to analyze the heat efficiency, combustion stability coefficient and steam quality index to comprehensively evaluate the boiler operation performance.

[0049] The present application dynamically distributes and recovers heat by using an intelligent control algorithm, and optimizes the heat distribution strategy by means of a digital twin verification platform and a closed-loop feedback mechanism, thereby improving the energy utilization rate; remote monitoring and control are realized based on network communication and cloud computing technology, so that the operation is more convenient and efficient, and the intelligent monitoring and control level of the heat supply boiler equipment is improved as a whole, the running cost is reduced, and the service life of the equipment is prolonged. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0051] Fig. 1 The system module diagram of the present application.

[0052] Fig. 2 The method flowchart of the present application. DETAILED DESCRIPTION

[0053] The technical solutions of the present application will be described clearly and completely in combination with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0054] Referring to Figs. 1-2 The first aspect embodiment of the present application provides an intelligent heat boiler equipment monitoring and control system, which comprises the following modules:

[0055] The data acquisition and preprocessing module: real-time monitoring of the heat boiler equipment, collecting a plurality of operation data and environmental data, and preprocessing respectively;

[0056] The data analysis module: according to the deep convolutional neural network model, identifying the equipment operation state, analyzing the noise source positioning and calculating the noise pollution degree value; calculating the thermal efficiency, combustion stability coefficient and steam quality index; according to the above calculation results combined with the operation state recognition result, constructing a self-attention mechanism combined with a survival analysis model to predict the failure probability; calculating the pollutant emission coefficient by monitoring the discharged pollutants; calculating the corrosion degree value by analyzing the corrosion trend;

[0057] The monitoring and evaluation module: comprehensive evaluation of the risk state index of the heat boiler equipment;

[0058] The monitoring and control module: using intelligent control algorithm, dynamically allocating and recovering heat; according to the allocation result, generating a closed-loop feedback mechanism through a digital twin verification platform; through network communication and cloud computing technology, remotely monitoring and controlling the heat boiler equipment.

[0059] Specifically, at key positions of the thermal boiler equipment, such as the furnace, flue, burner, water supply system, steam system, and other positions, various sensors are installed, including temperature sensors, pressure sensors, flow sensors, vibration sensors, sound sensors, optical sensors, etc. At the same time, environmental data such as ambient temperature, humidity, wind speed, atmospheric pressure, etc. are collected inside and outside the boiler room to determine multiple reasonable data collection points to ensure comprehensive monitoring of equipment operation and the impact of environmental factors on the equipment. Using the data acquisition and preprocessing module, the running data and environmental data of the thermal boiler equipment are collected in real time by the sensors according to the set sampling frequency. The running data includes but is not limited to the internal temperature, pressure, steam flow, water supply flow, fuel flow, combustion flame image, sound signal, etc. of the boiler, and the environmental data covers the temperature, humidity, wind speed, atmospheric pressure, etc. inside and outside the boiler room. The collected data is preprocessed, including data cleaning, filtering, normalization, standardization, etc. According to the deep convolutional neural network model, the noise source positioning and noise pollution degree value are calculated after identifying the equipment running state; the thermal efficiency, combustion stability coefficient, and steam quality index are calculated; according to the calculation results above and the running state identification results, a self-attention mechanism combined with a survival analysis model is constructed to predict the failure probability; the pollutant emission coefficient is calculated by monitoring the discharged pollutants; the corrosion degree value is calculated by analyzing the corrosion trend; according to the analysis results of the data analysis module, the risk state index of the thermal boiler equipment is comprehensively evaluated; an intelligent control algorithm is used to dynamically allocate and recover heat; according to the allocation results, a closed-loop feedback mechanism is generated through the digital twin verification platform. Using network communication technology such as 4G / 5G network or industrial Ethernet, the monitoring data of the thermal boiler equipment is transmitted in real time to the remote monitoring center. In the remote monitoring center, a large amount of data is stored and processed through cloud computing technology, and the staff can remotely access these data through monitoring terminals such as computers, mobile phones, etc. to view the running state of the boiler equipment. At the same time, the staff can also send control instructions through the monitoring terminal, which are transmitted to the control system of the boiler equipment through the network, realizing remote control of the boiler equipment, such as remote start and stop of the equipment, adjustment of operating parameters, etc.

[0060] In this embodiment, according to the deep convolutional neural network model, the noise source positioning and noise pollution degree value are calculated after identifying the equipment running state, including the following steps:

[0061] Collect and label sound sample data in historical running data of the thermal boiler equipment under normal operation, fault operation, and different load conditions;

[0062] The preprocessed time-domain sound signal is converted into a time-frequency domain image using short-time Fourier transform to generate a spectrogram. At the same time, the sound signal is decomposed into multiple scales using wavelet transform method, the wavelet coefficients are extracted, and the wavelet coefficients are rearranged into a pseudo-image, which is stacked with the spectrogram to form a multi-channel input.

[0063] The multi-channel input is trained as input data of a deep convolutional neural network model, and the running state of the heat boiler equipment corresponding to the real-time preprocessed running sound data is output;

[0064] According to the running state recognition result, combined with the physical layout of the heat boiler equipment and the acoustic propagation model, the specific component or position producing abnormal noise is determined through beam forming or time delay estimation method;

[0065] The formula for calculating the noise pollution degree value is:

[0066] Wherein, σ i is the sound pressure value of the actual measured noise of the i th monitoring point; L max is the maximum noise level; is the actual noise duration of the i th monitoring point; T max is the maximum duration allowed for noise; C is the confidence of the running state recognition result; M is the total number of monitoring points.

[0067] Specifically, sound sample data in historical operation data of the heat boiler equipment under normal operation, fault operation and different load conditions are collected and labeled, and the equipment operation state corresponding to each sound sample is determined, such as normal operation state, fault operation state including unstable combustion fault, equipment component loosening fault, etc. The preprocessed time-domain sound signal is converted into a time-frequency domain image by using short-time Fourier transform to generate a spectrogram, so as to intuitively show the energy distribution of the sound signal at different times and frequencies; at the same time, the sound signal is decomposed by using a wavelet transform method to extract wavelet coefficients, and the wavelet coefficients are rearranged into a pseudo-image, which is stacked with the spectrogram to form a multi-channel input, thereby increasing the perception ability of the model to the characteristics of the sound signal. The multi-channel input is used as the input data of the deep convolutional neural network model, the historical sound sample data labeled is used to train the model, the parameters of the convolutional layer, the pooling layer and the full connection layer of the model are adjusted, so that the model can learn the feature mode of the sound signal under different operation states, and when the model training is completed, the real-time preprocessed operation sound data is input into the model to output the operation state recognition result of the heat boiler equipment. According to the physical layout of the heat boiler equipment and the acoustic propagation model, a suitable beamforming or time delay estimation method is selected to determine the specific component or position of the abnormal noise. For example, for a boiler room with multiple noise sources, a beamforming method can be used, a microphone array is arranged at different positions, the time difference or phase difference of sound waves reaching different microphones is calculated by using an algorithm, a beamforming diagram is formed, and thus the direction and position of the noise source are determined; the time delay estimation method is to measure the time delay of sound waves propagating between different sensors, and calculate the position coordinates of the noise source according to the sound velocity. According to the noise source positioning result, the actual measured sound pressure value of the noise at each monitoring point, the noise duration and other data, the noise pollution degree value is calculated according to the noise pollution degree value calculation formula, so as to evaluate the influence degree of the noise generated in the operation process of the heat boiler equipment on the surrounding environment.

[0068] In the embodiment, the thermal efficiency, the combustion stability coefficient and the steam quality index are calculated, including the following steps:

[0069] The calculation formula of the thermal efficiency is:

[0070] Wherein, η is the thermal efficiency, Q input is the total heat input into the heat boiler equipment, Q loss is the total energy loss;

[0071] The flame image and light intensity time series data are collected by using an optical sensor, the flame image is preprocessed, the U-Net++ architecture is used for pixel-level segmentation of the flame area, and the flame area and perimeter are calculated by combining morphological operations; the combustion stability formula is calculated by combining the gas composition and the boiler operation data:

[0072] The combustion stability coefficient R is obtained, where I(t) is the flame intensity time series data, A and P are the flame area and perimeter respectively, and N co and are the concentrations of carbon monoxide and carbon dioxide, respectively;

[0073] Calculate the steam quality index by the formula:

[0074] Get the steam quality index Z, where S h is the steam humidity, S c is the salt content of steam, S0 is the impurity content in steam; P 实 is the actual steam pressure, P set is the set steam pressure, P_rated is the rated pressure of the boiler; L represents the energy or mass of steam produced by the boiler.

[0075] Specifically, by monitoring the boiler's heat input (e.g., fuel calorific value and fuel flow rate) and heat output (e.g., steam enthalpy and hot water temperature), while also accounting for heat losses (e.g., flue gas heat loss, heat dissipation loss, and incomplete combustion heat loss), the thermal efficiency of the thermal boiler equipment is calculated according to the thermal efficiency formula to assess the boiler's energy efficiency. Optical sensors are used to monitor and collect flame images and light intensity time series data. Flame images are preprocessed, and pixel-level segmentation of the flame region is performed using the U-Net++ architecture. Morphological operations are then used to calculate the flame area and perimeter. Furthermore, by combining gas composition data (e.g., carbon monoxide and carbon dioxide concentrations) with boiler operating data (e.g., fuel flow rate and air flow rate), a combustion stability coefficient is derived according to the combustion stability formula to assess the stability and completeness of the boiler's combustion process. Steam parameters such as humidity, salt content, impurity content, and steam pressure are monitored, and the steam quality index (SQI) is calculated according to the SQI formula to measure whether the steam quality generated by the boiler meets production process requirements. Weighting coefficients of 0.5, 0.3, and 0.2 are adjustable based on actual application and historical data.

[0076] In this embodiment, based on the above calculation results and the operating status recognition results, a self-attention mechanism is constructed in combination with a survival analysis model to predict the failure probability, including the following steps:

[0077] Collect noise pollution values, thermal efficiency, combustion stability coefficient, and steam quality index of thermal boiler equipment during historical operation. Combine historical operation data with operation status identification results to construct a training dataset and standardize the data.

[0078] The self-attention mechanism is combined with the survival analysis model, and the training data set is input into the model training. The input time sequence feature data enters the multi-head attention layer after position encoding, and the output survival function is: where h(τ) is a hazard rate function based on the constraint of the physical degradation equation, which is learned during model training;

[0079] The real-time monitoring collected training data set is input into the trained self-attention mechanism combined with the survival analysis model, the survival function S(t) is calculated in real time, and the failure probability G(t) = 1-S(t) is derived.

[0080] Specifically, the noise pollution degree value, thermal efficiency, combustion stability coefficient and steam quality index of the thermal boiler equipment in the historical running process are collected. These data can be obtained from the sensors, monitoring systems and historical running records of the equipment. At the same time, the historical running data of the thermal boiler are collected, including the start-up time, shutdown time, maintenance record, running parameter adjustment record and other information of the equipment, which helps to understand the running state change history of the equipment. The existing running state recognition model or method is used to analyze the collected historical running data, and the running state of the equipment at different time points is recognized, such as normal running, fault running and other states. The recognized running state results are combined with the above collected various data to form a data set containing complete running state information. The noise pollution degree value, thermal efficiency, combustion stability coefficient, steam quality index and corresponding running state recognition results collected are sorted to form a data set for model training. Each row in the data set represents a time point or a boiler running data sample of a running period, and each column corresponds to different features such as noise pollution degree value, thermal efficiency and running state label. The data set is standardized.

[0081] A model architecture is constructed by fusing self-attention mechanism and survival analysis. The input layer of the model is used to receive processed time series feature data, including standardized noise pollution level, thermal efficiency, combustion stability coefficient, steam quality index, etc. After inputting the time series feature data, position encoding is performed. The role of position encoding is to provide the position information of each data point in the sequence for the model, because the order of time series data is very important for capturing the trend of device running state and other information. For example, for a time series data sequence of length T, a position encoding vector is generated for each time position t, where 1≤t≤T. The vector is concatenated or added to the input feature data, so that the model can perceive the position of the data in the time sequence. The data after position encoding is input into the multi-head attention layer. The multi-head attention mechanism enables the model to learn the dependency between data in different representation subspaces in parallel, thereby better capturing long-term dependencies and key features in time series data. For example, set multiple attention heads such as 8 or 16, each attention head independently calculates the query, key and value matrices, calculates the correlation weight between different time steps through the scaled dot product attention method, and then concatenates or sums the outputs of multiple attention heads to obtain the output result of the multi-head attention layer. Some feedforward neural network layers and residual connections can be added in the middle of the model to further perform nonlinear transformation and feature extraction on the data, while maintaining the training stability of the model. The output layer outputs the survival function S(t). According to the survival analysis theory, the survival function S(t) represents the probability that the device has not failed at time t, where h(τ) is the hazard rate function constrained based on the physical degradation equation. The hazard rate function h(τ) reflects the instantaneous probability of device failure at time τ, which is determined by learning the rules in the data and the constraints based on the physical degradation equation during model training.

[0082] The training data set is input into the model for training. During the training process, a suitable loss function is defined, such as a negative log-likelihood loss function, which measures the difference between the predicted survival function of the model and the actual failure condition according to the nature of the survival analysis model. By optimizing the algorithm to minimize the loss function, the parameters in the model are adjusted, including the query, key, and value matrix parameters in the multi-head attention layer, the weights and biases in the feedforward neural network layer, etc., so that the model can accurately learn the rules in the data and the form of the hazard rate function h(τ). During the training process, cross-validation and other methods can be used to evaluate the performance of the model to prevent overfitting. For example, the training data set is divided into multiple subsets, one of which is used as the validation set and the remaining subsets are used as the training set for model training and validation. The structure and training parameters of the model are adjusted according to the performance indicators on the validation set, such as the accuracy of the model in predicting the probability of failure, the fitting degree of the survival function, etc.

[0083] Real-time monitoring of the running state of the thermal boiler equipment, continuously collecting noise pollution degree value, thermal efficiency, combustion stability coefficient and steam quality index and other data through sensors and other equipment. At the same time, the running state recognition model is used to recognize the real-time collected data to obtain the real-time running state result. The real-time collected data is standardized according to the same standardization method as the training data set to ensure that the data format and dimension input into the model are consistent with the training.

[0084] The real-time monitoring collected and processed data is input into the trained self-attention mechanism combined with the survival analysis model. The model calculates the survival function S(t) using the previously learned hazard rate function h(τ) according to the input data, and then derives the failure probability G(t) through the formula G(t) = 1-S(t), thereby realizing real-time prediction of the failure probability of the thermal boiler equipment.

[0085] In this embodiment, the pollutant emission coefficient is calculated by monitoring the emitted pollutants, including the following steps:

[0086] By monitoring the exhaust gas, wastewater and waste residue emitted by the thermal boiler equipment during operation, combined with environmental data, the pollutant emission coefficient formula is calculated as:

[0087] The pollutant emission coefficient W is obtained, where N i is the emission concentration of the i-th pollutant, R X is the fuel consumption, η is the thermal efficiency, T 实 is the actual temperature of the combustion chamber, T0 is the optimal combustion temperature, H T is the environmental temperature, H S is the environmental humidity, and n is the total number of pollutant emission types.

[0088] Specifically, by monitoring the types and quantities of exhaust gas such as sulfur dioxide, nitrogen oxides, particulate matter, wastewater such as chemical oxygen demand, ammonia nitrogen, and waste residue such as ash and slag, combined with environmental data such as environmental temperature and humidity, the pollutant emission coefficient is calculated according to the pollutant emission coefficient calculation formula, which is used to evaluate the pollution degree of the boiler operation to the environment and provide a basis for the development of environmental protection measures.

[0089] In this embodiment, the corrosion degree value is calculated by analyzing the corrosion trend, including the following steps:

[0090] Deploy MEMS corrosion probe array on the inner wall of the boiler to monitor the local pH value, Cl- ion concentration and oxidation-reduction potential in real time; collect the pitting corrosion occurrence position data of the inner wall of the boiler as the label for graph neural network training;

[0091] Modeling the electrochemical coupling effect of the corrosion micro-zone in the boiler by using a graph neural network, inputting local pH value, Cl- ion concentration, oxidation-reduction potential for training, and outputting the predicted pitting corrosion preferential occurrence position;

[0092] According to the prediction result of the graph neural network, analyze the corrosion rate, corrosion degree data and corrosion trend, and comprehensively analyze the results to calculate the corrosion degree value by weighted average summation.

[0093] Specifically, MEMS corrosion probe arrays are deployed reasonably on the inner wall of the boiler, ensuring that the probe arrays can cover different areas of the inner wall of the boiler, especially the parts prone to corrosion, such as the vicinity of the water level line, the entrance of flue gas, etc. Real-time monitoring of local pH value, Cl- ion concentration, and redox potential and other corrosion-related parameters, and transmitting these data to the data acquisition and processing system. At the same time, regular non-destructive testing of the inner wall of the boiler is carried out, such as using an ultrasonic detector to measure the corrosion depth, using an endoscope to take pictures of the inner wall to determine the corrosion area and pitting situation, etc., and the detection results are digitized, extracting corrosion depth data, corrosion area ratio data, and pitting density data, etc. The collected various corrosion data are preprocessed, including data cleaning, filtering, and normalization, etc. to improve the accuracy and reliability of subsequent analysis. Collect the location data of pitting on the inner wall of the boiler, which is used as the label for training the graph neural network. At the same time, the preprocessed local pH value, Cl- ion concentration, redox potential, etc. are used as input features to construct the training data set. A graph neural network model is constructed, which can model the electrochemical coupling effect of the inner corrosion microzone of the boiler, and capture the complex relationships and interactions in the corrosion process by learning the correlations and influences between different corrosion factors. For example, the model can learn how a decrease in pH value promotes the anodic dissolution reaction of the metal, and how an increase in Cl- ion concentration exacerbates the formation of pitting, etc. The graph neural network model is trained using the training data set, and the parameters and structure of the model are adjusted so that the output of the model can accurately predict the location where pitting occurs first. In the training process, cross-validation and other methods can be used to evaluate the performance of the model to avoid overfitting or underfitting. According to the prediction results of the graph neural network, the corrosion risk level of different areas of the inner wall of the boiler is determined, and the corrosion situation around the predicted pitting location is focused on. Combined with the actually collected corrosion rate data, corrosion depth data, corrosion area ratio data, and pitting density data, etc., the corrosion degree of these areas is analyzed in detail, including corrosion rate, corrosion degree data, and corrosion trend, where the corrosion trend mainly refers to the trend of the corrosion situation of the inner wall of the boiler over time, including the evolution of corrosion-related indicators such as corrosion rate, corrosion depth, corrosion area ratio, and pitting density over time. The weights are determined using expert scoring method, analytic hierarchy process, etc. The corrosion rate, corrosion degree data, and corrosion trend are multiplied by their corresponding weights and then weighted average summed to obtain the corrosion degree value.

[0094] In the present embodiment, the risk state index of the thermal boiler equipment is comprehensively evaluated, including the following steps:

[0095] According to the fault probability, thermal efficiency, combustion stability coefficient, steam quality index, pollutant emission coefficient, corrosion degree value and noise pollution value calculated by real-time monitoring, weighted average summation calculation is performed to comprehensively evaluate the thermal boiler equipment risk state index; wherein, the weight distribution is performed by using entropy weight method or analytic hierarchy process;

[0096] According to the thermal boiler equipment risk state index, a three-level risk early warning mechanism is established.

[0097] Specifically, the index data such as fault probability, thermal efficiency, combustion stability coefficient, steam quality index, pollutant emission coefficient, corrosion degree value and noise pollution value calculated by real-time monitoring are obtained from the data acquisition and preprocessing module and the data analysis module to ensure the accuracy and timeliness of the data, and are normalized or standardized respectively. According to the actual demand and evaluation purpose, the weight distribution method such as entropy weight method or analytic hierarchy process is selected to determine the weight of each index in the comprehensive evaluation of risk state index. For example, if the combustion stability has a greater impact on the safe operation of the boiler, the proportion of the combustion stability coefficient in the weight distribution can be appropriately increased; if the pollutant emission has a more critical impact on the environment, the weight of the pollutant emission coefficient can be increased. Through the entropy weight method, the weight can be objectively distributed according to the dispersion degree and information entropy of each index; while the analytic hierarchy process is a method combining qualitative and quantitative analysis, which establishes a hierarchical structure model to determine the relative importance relationship between each index and calculates the corresponding weight vector. After multiplying each index data by its corresponding weight, weighted average summation calculation is performed to obtain the comprehensive risk state index of the thermal boiler equipment, which can comprehensively reflect the overall risk level of the boiler equipment under the current operating state. According to the size range of the comprehensive risk state index, three different risk levels are divided to establish a three-level risk early warning mechanism. For example, when the risk state index is below the threshold value 0.3, it is determined as a low risk state; when the risk state index is in the threshold value range of 0.3 and 0.6, it is determined as a medium risk state; when the risk state index exceeds the threshold value 0.6, it is determined as a high risk state. The threshold value should be adjusted and optimized according to the specific application, and at the same time, the threshold value should be re-evaluated and calibrated regularly as the operation of the boiler equipment and the external environment change, so as to ensure the effectiveness and reliability of the three-level risk early warning mechanism.

[0098] In this embodiment, an intelligent control algorithm is used to dynamically allocate and recover heat, including the following steps:

[0099] Real-time acquisition of boiler operating parameters, synchronous acquisition of steam demand, pressure threshold and heating time requirement;

[0100] The dynamic allocation model is constructed by using a fuzzy PID compound control algorithm. The input layer receives the risk state index of the heat supply boiler equipment, the combustion efficiency, the heat storage capacity and the pipe network pressure difference parameters. The influence weight of each factor is quantified by using a membership function. The output layer generates a heat allocation coefficient matrix.

[0101] A multi-objective optimization function is established. The fuel consumption is minimized, the waste heat recovery rate is maximized, and the equipment loss is balanced as a constraint condition. The risk index is converted into an equipment reliability constraint. The NSGA-II genetic algorithm is used to solve a Pareto optimal solution set.

[0102] The dynamic adjustment is executed by using a distributed control system (DCS). According to the optimal solution set of the heat allocation scheme, the dynamic adjustment operation is executed. The excess heat is intelligently allocated to the waste heat boiler, the heat storage tank or the heat network circulating system.

[0103] Specifically, the data acquisition system is used to collect real-time boiler operating parameters, including but not limited to internal temperature, pressure, steam flow, fuel flow, and water flow, and simultaneously obtain production demand information such as steam demand, pressure threshold, and heating time requirements, thereby providing basic data support for heat distribution. The input layer of the fuzzy PID compound control algorithm receives parameters such as risk state index, combustion efficiency, heat storage capacity, and pipe network pressure difference of the thermal boiler equipment, which can comprehensively reflect the operating state of the boiler and the influencing factors of heat distribution. For each input parameter, a corresponding membership function is determined to quantify the influence weight of each factor on heat distribution. For example, for the risk state index, different membership functions can be set to represent the different influence degrees of low risk, medium risk, and high risk on heat distribution; for the combustion efficiency, the membership function can be set according to its level to reflect its priority in heat distribution, etc. The output layer generates a heat distribution coefficient matrix, which clearly defines the heat distribution proportion and coefficient corresponding to each heat distribution destination such as the waste heat boiler, heat storage tank, or heat network circulation system, so as to facilitate subsequent heat distribution execution operations. A multi-objective optimization function is established to minimize fuel consumption, maximize waste heat recovery rate, and balance equipment wear and tear as constraint conditions, while converting the risk state index into a device reliability constraint to ensure that the optimization of heat distribution is achieved under the premise of meeting production demand and safe and reliable operation of the equipment. The NSGA-II genetic algorithm is used to solve the multi-objective optimization function, and through genetic operations such as initialization of population, selection, crossover, and mutation, the Pareto optimal solution set is obtained through continuous iteration and optimization, which represents the optimal balance scheme between different objectives, providing multiple feasible alternative schemes for actual heat distribution decision-making. The optimal solution set of heat distribution scheme obtained by the NSGA-II genetic algorithm is sent to the distributed control system (DCS), which executes dynamic adjustment operations according to the set control logic and strategy, intelligently distributes excess heat to different heat utilization links such as the waste heat boiler, heat storage tank, or heat network circulation system according to the optimized distribution scheme, realizes the rational recovery and effective utilization of heat, and improves the energy utilization efficiency and operating economy of the entire thermal system.

[0104] In this embodiment, a closed-loop feedback mechanism is generated through the digital twin verification platform according to the distribution results, including the following steps:

[0105] A digital twin verification platform is constructed to simulate the operating state of the thermal boiler equipment and its heat distribution;

[0106] An LSTM neural network is used to predict the life attenuation curve of the thermal boiler equipment under different heat distribution strategies;

[0107] The equipment life attenuation curve predicted by the LSTM neural network is compared with the actual operation data, and the control parameters are continuously optimized according to the comparison result, and the optimized parameters are fed back to the intelligent control algorithm to form a closed-loop feedback mechanism.

[0108] Specifically, a digital twin model corresponding to the actual thermal boiler equipment and its heat distribution system is constructed by using digital modeling technology. The model can accurately simulate the physical structure, operation principle, heat distribution process and interaction relationship between components of the boiler equipment, realize the simulation of the equipment running state in the virtual environment and the verification of the heat distribution strategy. The historical operation data and life attenuation data of the thermal boiler equipment under different heat distribution strategies are collected for training the LSTM neural network model. The LSTM neural network can effectively model the sequence data, predict the life attenuation curve of the thermal boiler equipment under different heat distribution strategies by learning the rules and patterns in the historical data, and the prediction result can reflect the life change trend of the equipment under different operating conditions and heat distribution schemes. The equipment life attenuation curve predicted by the LSTM neural network is compared with the actual operation data, and the differences and deviations between the two are found out. According to the comparison result, the related parameters in the intelligent control algorithm are continuously optimized and adjusted, such as adjusting the membership function parameters in the fuzzy PID compound control algorithm, the population size and mutation probability in the NSGA-II genetic algorithm, etc., to improve the rationality and optimization effect of the heat distribution strategy. The optimized parameters are fed back to the intelligent control algorithm to realize the self-optimization and iterative update of the control algorithm, thereby forming a closed-loop feedback mechanism. Through this closed-loop feedback mechanism, the thermal boiler equipment monitoring and control system can continuously optimize the heat distribution strategy and control parameters according to the actual operation situation and the prediction and evaluation result of the digital twin verification platform, improve the operation performance and reliability of the equipment, reduce the operation risk and energy consumption, and ensure the continuous and effective operation and optimization of the intelligent monitoring and control system of the thermal boiler equipment.

[0109] The above embodiments are only used to illustrate the technical method of the present application and not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. An intelligent thermal boiler equipment monitoring and control system, characterized in that: Includes the following modules: Data acquisition and preprocessing module: real-time monitoring of thermal boiler equipment, collection of several operating data and environmental data, and preprocessing of them separately; Data Analysis Module: Based on a deep convolutional neural network model, the system identifies the equipment's operating status, locates noise sources, and calculates the noise pollution level. It also calculates thermal efficiency, combustion stability coefficient, and steam quality index. Based on these calculations and the operating status identification results, it constructs a self-attention mechanism and a survival analysis model to predict failure probability. It also calculates pollutant emission coefficients by monitoring emitted pollutants. Calculate the corrosion degree value by analyzing the corrosion trend; Monitoring and evaluation module: Comprehensively evaluate the risk status index of thermal boiler equipment; Monitoring and control module: uses intelligent control algorithm to dynamically distribute recovered heat; Based on the allocation results, a closed-loop feedback mechanism is generated through the digital twin verification platform; through network communication and cloud computing technology, thermal boiler equipment is remotely monitored and controlled.

2. The intelligent thermal boiler equipment monitoring and control system according to claim 1 is characterized in that: Based on the deep convolutional neural network model, the noise source is located and the noise pollution level is calculated after the equipment operating status is identified. The following steps are included: Collect and annotate sound sample data from historical operating data of thermal boiler equipment under normal operation, fault operation, and different load conditions; The preprocessed time-domain sound signal is converted into a time-frequency domain image using short-time Fourier transform to generate a spectrogram. Meanwhile, the wavelet transform method is used to perform multi-scale decomposition on the sound signal, extracting wavelet coefficients. These wavelet coefficients are then rearranged into a pseudo-image and stacked with the spectrogram to form a multi-channel input. The multi-channel input is used as the input data for training the deep convolutional neural network model, and the operating status of the thermal boiler equipment corresponding to the real-time pre-processed operating sound data is output; Based on the operating status identification results, combined with the physical layout of the thermal boiler equipment and the acoustic propagation model, the specific component or location generating the abnormal noise is determined through beamforming or time delay estimation methods. Formula for calculating noise pollution level: Among them, σ i is the actual sound pressure value of the noise measured at the i-th monitoring point; L max is the maximum noise level; is the actual noise duration of the ith monitoring point; T max is the maximum duration allowed by noise; C is the confidence level of the operating status identification result; M is the total number of monitoring points.

3. The intelligent thermal boiler equipment monitoring and control system according to claim 1 is characterized in that: Calculation of thermal efficiency, combustion stability coefficient, and steam quality index includes the following steps: The formula for calculating thermal efficiency is: Where η is the thermal efficiency, Q input is the total heat input to the thermal boiler equipment, Q loss is the total energy loss; Optical sensors are used to monitor and collect flame images and light intensity time series data. Flame images are preprocessed and the U-Net++ architecture is used to segment the flame area at the pixel level. Morphological operations are then used to calculate the flame area and perimeter. Combining gas composition and boiler operating data, the combustion stability formula is calculated: The combustion stability coefficient R is obtained, where I(t) is the flame intensity time series data, A and P are the flame area and perimeter respectively, and N co and are the concentrations of carbon monoxide and carbon dioxide, respectively; Calculate the steam quality index by the formula: Get the steam quality index Z, where S h is the steam humidity, S c is the salt content of steam, S0 is the impurity content in steam; P 实 is the actual steam pressure, P set is the set steam pressure, P_rated is the rated pressure of the boiler; L represents the energy or mass of steam produced by the boiler.

4. The intelligent thermal boiler equipment monitoring and control system according to claim 3 is characterized in that: Based on the above calculation results and the operating status recognition results, a self-attention mechanism combined with a survival analysis model is constructed to predict the failure probability, including the following steps: Collect noise pollution values, thermal efficiency, combustion stability coefficient, and steam quality index of thermal boiler equipment during historical operation. Combine historical operation data with operation status identification results to construct a training dataset and standardize the data. Construct a self-attention mechanism combined with a survival analysis model, and input the training data set into the model training. The input time series feature data enters the multi-head attention layer after position encoding, and the output survival function is: Where h(τ) is the hazard rate function based on the physical degradation equation constraints, which is learned during the model training process; The training data set collected by real-time monitoring is input into the trained self-attention mechanism combined with the survival analysis model, the output survival function S(t) is calculated in real time, and the failure probability is derived as G(t) = 1-S(t).

5. The intelligent thermal boiler equipment monitoring and control system according to claim 1 is characterized in that: The pollutant emission coefficient is calculated by monitoring the emitted pollutants, including the following steps: By monitoring the exhaust gas, waste water and waste residue emitted by thermal boiler equipment during operation and combining it with environmental data, the pollutant emission coefficient formula is calculated as follows: Get the pollutant emission coefficient W, where N i is the emission concentration of the i-th pollutant, R X is the fuel consumption, η is the thermal efficiency, T 实 is the actual temperature of the combustion chamber, T0 is the optimal combustion temperature, H T is the ambient temperature, H S is the ambient humidity, and n is the total number of pollutant emission types.

6. The intelligent thermal boiler equipment monitoring and control system according to claim 1 is characterized in that: The corrosion degree value is calculated by analyzing the corrosion trend, including the following steps: A MEMS corrosion probe array is deployed on the inner wall of the boiler to monitor the local pH value, Cl- ion concentration, and redox potential in real time. Data on the location of pitting corrosion on the inner wall of the boiler is collected and used as labels for graph neural network training. A graph neural network is used to model the electrochemical coupling effect of corrosion micro-regions in boilers. The local pH value, Cl- ion concentration, and redox potential are input for training, and the output is the predicted location where pitting corrosion will occur first. According to the prediction results of the graph neural network, the corrosion rate, corrosion degree data and corrosion trend are analyzed, the results are comprehensively analyzed, and the weighted average summation is performed to calculate the corrosion degree value.

7. The intelligent thermal boiler equipment monitoring and control system according to claim 1 is characterized in that: Comprehensively assess the risk status index of thermal boiler equipment, including the following steps: Based on the failure probability, thermal efficiency, combustion stability coefficient, steam quality index, pollutant emission coefficient, corrosion degree value, and noise pollution value calculated through real-time monitoring, a weighted average summation calculation is performed to comprehensively evaluate the risk status index of thermal boiler equipment; the weights are assigned by using the entropy weight method or the hierarchical analysis method; A three-level risk warning mechanism is established based on the risk status index of thermal boiler equipment.

8. The intelligent thermal boiler equipment monitoring and control system according to claim 1 is characterized in that: Adopting intelligent control algorithm to dynamically distribute recovered heat, including the following steps: Collect boiler operating parameters in real time and simultaneously obtain steam demand, pressure threshold, and heating time requirements; A dynamic allocation model is constructed using a fuzzy PID composite control algorithm. The input layer receives the risk status index of thermal boiler equipment, combustion efficiency, heat storage capacity, and pipe network pressure difference parameters. The influence weight of each factor is quantified through the membership function, and the output layer generates a heat allocation coefficient matrix. A multi-objective optimization function was established with minimizing fuel consumption, maximizing waste heat recovery, and balancing equipment losses as constraints. The risk index was converted into equipment reliability constraints, and the NSGA-II genetic algorithm was used to solve the Pareto optimal solution set. Dynamic adjustment is performed through the distributed control system DCS. According to the optimal solution for heat distribution, dynamic adjustment operations are performed to intelligently distribute excess heat to waste heat boilers, heat storage tanks or heat network circulation systems.

9. The intelligent thermal boiler equipment monitoring and control system according to claim 8, characterized in that: Based on the allocation results, a closed-loop feedback mechanism is generated through the digital twin verification platform, which includes the following steps: Build a digital twin verification platform to simulate the operating status of thermal boiler equipment and its heat distribution; Use LSTM neural network to predict the life attenuation curve of thermal boiler equipment under different heat distribution strategies; The equipment life attenuation curve predicted by the LSTM neural network is compared and analyzed with the actual operating data. The control parameters are continuously optimized based on the comparison results, and the optimized parameters are fed back to the intelligent control algorithm to form a closed-loop feedback mechanism.

Images