Hazardous waste collaborative incineration control method and system based on multimode sensing

Abstract

The invention provides a multi-mode sensing hazardous waste collaborative incineration combined heat and power generation regulation and control method and system, and relates to the combustion technology of a combined heat and power generation boiler. The collaborative incineration regulation and control method specifically comprises the steps that working condition data of three flues in a combustion system are obtained, wherein the working condition data at least comprise image data used for representing the flue ash deposition state; extracting at least one ash deposition feature representing the ash deposition state, and performing space-time fusion on the ash deposition features of the three flue positions to generate a system context vector representing the global ash deposition state of the combustion system; based on the system context vector, generating a decision vector for adjusting the hazardous waste ratio; the decision vector is corrected until the predicted temperature corresponding to the corrected decision vector conforms to the temperature constraint interval, and the hazardous waste ratio is adjusted according to the decision vector; and monitoring the actually measured temperature of the combustion system at one or more positions, and if the actually measured temperature deviates from a preset target temperature interval, carrying out feedback adjustment on the current hazardous waste ratio.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of computer vision in artificial intelligence, in particular, to the combustion technology of cogeneration boiler. More specifically, the present application carries out online monitoring, diagnosis and feedforward control on the furnace and flue conditions in the process of hazardous waste and conventional fuel co-incineration, so as to realize stable and efficient combustion process and inhibit key area ash deposition and slagging. BACKGROUND

[0002] Co-incineration of hazardous waste (HW) in large-scale industrial facilities such as power boilers and cement kilns is an important technical route for realizing the reduction, harmlessness and resource utilization of hazardous waste. This technology can utilize the mature high-temperature environment and long flue gas residence time of existing industrial facilities to ensure that the toxic and harmful components in hazardous waste, especially persistent organic pollutants (such as dioxins), are completely decomposed. At the same time, the heat value contained in hazardous waste can be used as a substitute for fossil fuels, replacing part of the fossil energy, thereby reducing energy consumption and carbon emissions, and producing significant economic and environmental benefits. Therefore, hazardous waste co-incineration cogeneration technology has become a green environmental protection technology that is widely concerned and actively developed worldwide.

[0003] However, introducing hazardous waste into the precise and continuous operation of the cogeneration boiler system designed for burning standardized fuel also poses serious technical challenges. Unlike coal powder, which has relatively stable composition and uniform characteristics, hazardous waste has a wide range of sources and exhibits significant volatility and complexity in its physical and chemical properties. Different batches and types of hazardous waste have significant differences in heat value, ash content, moisture, and ash chemical composition (such as the content of alkali metals sodium and potassium, alkaline earth metals calcium and magnesium, and elements such as iron, sulfur, and chlorine). This high uncertainty in fuel characteristics poses a direct threat to the safe, stable, and economic operation of the boiler, with the most prominent and widespread technical problem being the problem of ash deposition and slagging on the furnace and flue heat exchange surfaces.

[0004] Ash deposition and slagging refer to the phenomenon of ash particles produced after fuel combustion softening, melting, and depositing and bonding on key heat exchange surfaces such as boiler water walls, superheaters, reheaters, economizers, and air preheaters at high temperatures. Severe ash deposition and slagging can cause a series of chain negative effects: the ash and slag layer acts as an insulating layer, dramatically reducing heat exchange efficiency, leading to substandard steam parameters, directly affecting the power generation efficiency and heating capacity of the unit; ash deposition can block the flue gas passage, increasing airflow resistance, forcing auxiliary equipment such as induced draft fans and forced draft fans to increase energy consumption; certain ash deposits rich in sulfur, chlorine, and alkali metals have strong corrosiveness at specific temperatures, which can severely erode heat exchanger tubing, shorten equipment life, and even cause major safety accidents such as pipe bursts; severe ash deposition and slagging will eventually force the unit to operate at reduced load or even shut down for manual cleaning, causing significant economic losses and power generation losses.

[0005] To address this problem, existing technologies typically employ relatively passive or indirect control methods. One common approach is offline laboratory analysis. The fundamental drawback of this method is its inherent lag. The composition of hazardous waste can change rapidly, while laboratory analysis has long cycles and cannot provide real-time fuel characteristic data. This leads to deviations between the actual mixed fuel characteristics fed into the furnace and preset values, still causing ash accumulation problems. Another approach relies on traditional boiler operating parameters, such as monitoring the outlet steam temperature, flue gas temperature, or flue pressure drop of each stage of heat exchanger to indirectly determine the ash accumulation status. However, these thermal parameters are results of ash accumulation reaching a significant level, not precursors, and are therefore lagging and indirect indicators. By the time significant anomalies in temperature and pressure are detected, hard, difficult-to-remove deposits have often already formed on the heat exchange surfaces.

[0006] Therefore, there is an urgent need in this field for a new technology that can monitor and diagnose the co-incineration process of hazardous waste in real time and online, and that can perform feedforward control of the hazardous waste ratio based on the diagnostic results and in combination with the process constraints of the boiler, so as to ensure the long-term safe, stable and efficient operation of the cogeneration unit when accepting complex and variable hazardous waste. Summary of the Invention

[0007] This invention provides a multi-modal sensing method for the regulation and control of hazardous waste co-incineration combined heat and power, which specifically includes the following steps: Acquire operating condition data for three flues within the combustion system, wherein the operating condition data includes at least image data characterizing the ash accumulation state of the flues; Extract at least one ash accumulation feature to characterize the ash accumulation state, and perform spatiotemporal fusion of the ash accumulation features at the three flue locations to generate a system context vector characterizing the global ash accumulation state of the combustion system. Based on the system context vector, a decision vector for adjusting the hazardous waste ratio is generated; The decision vector is corrected until the predicted temperature corresponding to the corrected decision vector meets the temperature constraint range. The hazardous waste ratio is adjusted according to the decision vector. The measured temperature of the combustion system at one or more locations is monitored. If the measured temperature deviates from the preset target temperature range, the current hazardous waste ratio is adjusted accordingly.

[0008] The aforementioned ash accumulation features include ash accumulation rate features calculated based on temporal differences of image sequences; color features calculated based on statistical moments in the color space; and morphological and texture features extracted by a pre-trained convolutional neural network.

[0009] The three flues are constructed as a directed graph with connections, where each flue is a node and its corresponding ash accumulation feature is a node feature; by aggregating the node features of each node and its upstream neighbor nodes, the node feature representation with fused spatial correlation is updated and generated. The time series of node features with fused spatial correlation is input into a gated recurrent unit network to capture the dynamic evolution of the ash accumulation state over time and output the system context vector.

[0010] The decision vector includes multiple adjustment values ​​corresponding one-to-one with various controllable hazardous materials; wherein the sign of each adjustment value is used to determine the direction of adjustment for increasing or decreasing the proportion of the corresponding hazardous waste.

[0011] The step of correcting the decision vector is a loop-iterative verification process, which specifically includes: (1) determining a candidate hazardous waste and fuel ratio based on the initial or previous corrected decision vector; (2) predicting the temperature at at least one critical location inside the combustion system caused by the candidate ratio; (3) comparing the predicted temperature with a preset temperature constraint range; (4) if the predicted temperature does not conform to the temperature constraint range, correcting the current decision vector and returning to step (1) until the predicted temperature conforms to the constraint range.

[0012] The step of feedback adjustment of the current hazardous waste and fuel ratio is a secondary optimization process to maintain the stability of the thermal conditions of the flue in the medium and low temperature zones. The feedback adjustment is to fine-tune the total calorific value corresponding to the hazardous waste and fuel ratio based on the deviation between the measured temperature of the second-stage flue and / or the third-stage flue and the set target temperature range. The embodiments in this specification also propose a multi-modal sensing hazardous waste co-incineration cogeneration control system, which includes: Acquisition module: Acquires operating condition data of three flues within the combustion system, wherein the operating condition data includes at least image data characterizing the ash accumulation state of the flues; Feature extraction module: Extracts at least one ash accumulation feature that characterizes the ash accumulation state, performs spatiotemporal fusion of the ash accumulation features at the three flue locations, and generates a system context vector that characterizes the global ash accumulation state of the combustion system; Decision generation module: Based on the system context vector, it generates a decision vector for adjusting the hazardous waste ratio; Correction module: Corrects the decision vector until the predicted temperature corresponding to the corrected decision vector meets the temperature constraint range, adjusts the hazardous waste ratio according to the decision vector; and monitors the measured temperature of the combustion system at one or more locations. If the measured temperature deviates from the preset target temperature range, the current hazardous waste ratio is adjusted accordingly.

[0013] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned multi-mode sensing method for controlling the co-incineration of hazardous waste with combined heat and power.

[0014] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned multi-mode sensing method for the control of hazardous waste co-incineration combined heat and power.

[0015] A computer program product, characterized in that the computer program product includes a computer program, which, when executed by a processor, implements the above-mentioned multi-mode sensing method for the control of hazardous waste co-incineration and cogeneration.

[0016] Compared with existing technologies, the multimodal sensing-based hazardous waste co-incineration control method disclosed in this invention achieves a leap from passive response to proactive prevention. Existing technologies rely on lagging thermal parameters such as flue gas temperature and pressure difference to indirectly infer ash accumulation, resulting in slow response and an inability to diagnose the root cause of the problem. This invention, however, directly extracts and analyzes the visual characteristics of ash accumulation, such as growth rate, color, morphology, and texture, online in a multi-level and refined manner by arranging visual sensing modules at each stage of the flue. This direct image detection-based approach allows the system to capture subtle changes in ash accumulation at its nascent stage, with sensitivity and timeliness far exceeding traditional methods. More importantly, through in-depth analysis of ash color and morphology, the system can preliminarily determine the type of ash accumulation (such as high-temperature molten slag or low-temperature corrosive ash) and its possible chemical causes, thus providing a scientific basis for subsequent precise proportioning adjustments.

[0017] This invention proposes a spatiotemporal fusion method for system state characterization, enabling a global understanding of the health status of the entire combustion system. Existing technologies treat each measuring point as an isolated information source, failing to reveal their inherent relationships. This invention, however, abstracts the three flues into a directed graph structure and utilizes spatial information fusion, allowing the state analysis of downstream flue nodes to naturally include the state information of their upstream flue nodes, mapping the physical causal relationships of flue gas flow. Building upon this, this invention further incorporates a gated cyclic unit network to process time-series information, thereby learning and understanding the evolution and dynamic trends of the entire system state over time. Through this deep fusion of spatiotemporal dimensions, the system ultimately generates not scattered data points, but a context vector that comprehensively reflects the spatiotemporal characteristics of the entire combustion system.

[0018] This invention establishes a dual closed-loop control mechanism of verification and feedback correction that balances safety and optimality, ensuring the robustness and reliability of control decisions. After deriving the initial decision vector based on ash accumulation analysis, this invention does not execute it immediately but introduces a verification step. This step pre-estimates the impact of the new formulation on key thermodynamic parameters such as furnace temperature and first-stage flue gas temperature, and uses a preset temperature constraint range for judgment, ensuring that any adjustment aimed at suppressing ash accumulation will not impair the boiler's combustion stability and steam quality. After verification and execution of the new formulation, this invention also designs a correction step, which continuously monitors the actual temperatures of the second and third-stage flues, compares them with the target range, and makes fine-grained feedback adjustments for any possible deviations. This prediction-execution-re-verification closed-loop control not only resolves the potential conflict between the two objectives of suppressing ash accumulation and maintaining thermodynamic conditions but also adapts to deviations in actual operating conditions, enabling the entire control system to exhibit extremely high adaptability and reliability in complex industrial environments. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of the multi-modal sensing-based control of hazardous waste co-incineration according to the present invention. Detailed Implementation

[0021] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0022] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this application, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number and aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0024] Additionally, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that practice can be carried out without these specific details.

[0025] This invention provides a multi-modal sensing method for the regulation and control of hazardous waste co-incineration combined heat and power, which specifically includes the following steps: Acquire operating condition data for three flues within the combustion system, wherein the operating condition data includes at least image data characterizing the ash accumulation state of the flues; Extract at least one ash accumulation feature to characterize the ash accumulation state, and perform spatiotemporal fusion of the ash accumulation features at the three flue locations to generate a system context vector characterizing the global ash accumulation state of the combustion system. Based on the system context vector, a decision vector for adjusting the hazardous waste ratio is generated; The decision vector is corrected until the predicted temperature corresponding to the corrected decision vector meets the temperature constraint range. The hazardous waste ratio is adjusted according to the decision vector. The measured temperature of the combustion system at one or more locations is monitored. If the measured temperature deviates from the preset target temperature range, the current hazardous waste ratio is adjusted accordingly.

[0026] This embodiment provides a combined heat and power (CHP) combustion system for multi-mode sensing hazardous waste co-incineration. This system, primarily serving as part of a large-scale CHP unit, converts the chemical energy contained in a mixed fuel, including hazardous waste and pulverized coal, into thermal energy for power generation and external heating. Structurally, the system mainly comprises a combustion furnace, a series of multi-stage flues, and a coupled water-steam circulation system. The combustion furnace is the core area for energy release, while the multi-stage flues are key locations for cascaded energy recovery. The mixed fuel is combusted within the furnace, and the generated high-temperature flue gas flows sequentially through each stage of the flue, transferring its heat to the working medium in the water-steam circulation system, ultimately achieving CHP.

[0027] The combustion chamber is a large, chamber-like combustion space composed of membrane water-cooled walls, with an overall tower-like structure. Multiple sets of burners are evenly distributed circumferentially in the lower region of the chamber, including primary air burners for injecting finely ground coal powder from a coal mill, and injection devices specifically for conveying pretreated hazardous waste, including atomizing spray guns for liquid hazardous waste and conveying nozzles for solid and semi-solid hazardous waste. To achieve complete and staged combustion of the fuel and suppress the formation of nitrogen oxides, the chamber is also equipped with secondary air burners and burnout air nozzles. During operation, pulverized coal and hazardous waste are fed into the chamber in a specific ratio, rapidly mixing with high-temperature combustion air (primary and secondary air) heated by an air preheater, and undergoing intense combustion, forming a stable flame core with a temperature reaching 1300°C to 1500°C in the central region of the chamber. The flame core transfers most of the heat to the surrounding membrane water-cooled walls through thermal radiation. Inside the pipes of the membrane water-cooled wall, high-pressure feedwater from the economizer flows. After absorbing radiant heat, the feedwater rapidly heats up and undergoes violent vaporization, forming a saturated water-steam mixture.

[0028] After completing the main heat release process in the furnace, the high-temperature flue gas exits from the upper outlet of the furnace and enters a series of three flue ducts. These three flue ducts are structurally continuous channels, functionally divided into areas with different operating temperatures and heat exchange tasks. The first-stage flue duct, namely the uppermost part of the horizontal flue and the tail vertical flue duct, has the highest operating flue gas temperature, between 900℃ and 1200℃. High-temperature superheaters and reheaters are located in this area. The saturated water-steam mixture generated by the furnace water-cooled walls is separated in the steam-water separator. The saturated steam enters the high-temperature superheater, where it is heated by the high-temperature flue gas through radiation and convection heat exchange to become superheated steam. This superheated steam is then directed to the high-pressure cylinder of the turbine to generate electricity. The reheater is used to reheat the exhaust steam from the high-pressure cylinder to its original high temperature before sending it to the intermediate and low-pressure cylinders of the turbine to continue generating power, thereby improving the thermal efficiency of the entire cycle.

[0029] The flue gas, carrying a moderate temperature, continues to flow downwards into the second-stage flue. This area mainly houses the low-temperature superheater and economizer. The low-temperature superheater continues to heat the steam, while the economizer's main function is to utilize the waste heat of the flue gas to preheat the high-pressure feedwater before it enters the steam-water separator. Preheating increases the feedwater temperature, reducing the amount of fuel required to heat the water to its boiling point in the furnace, thereby directly improving the boiler's overall thermal efficiency. After passing through the second-stage flue, the flue gas temperature decreases further.

[0030] The flue gas terminates in the third-stage flue, which houses an air preheater. At this point, the flue gas temperature is relatively low, and the remaining heat it carries is recovered by the air preheater to heat the combustion air that will be sent into the furnace. This process recovers the low-temperature heat energy that would otherwise be lost, and the hot air sent into the furnace helps improve fuel ignition and combustion stability. Finally, the flue gas, with a temperature reduced to 130°C to 150°C, leaves the air preheater and enters the subsequent flue gas purification system for treatment. After meeting the standards, it is discharged through the chimney. Throughout the process, the steam turbine drives the generator to produce electricity, while a portion of steam is extracted from a specific stage of the turbine for industrial or residential heating.

[0031] This embodiment, based on the cogeneration combustion system described in the previous embodiment, further discloses an online monitoring and data acquisition subsystem for obtaining key operating parameters within the system. This subsystem, through the fusion arrangement of multi-source, heterogeneous sensors, provides real-time, high-dimensional data input for subsequent combustion process control. The subsystem mainly consists of a multimodal visual perception module and a set of distributed sensing modules.

[0032] The multimodal visual sensing module consists of multiple high-temperature industrial vision probes arranged within the multi-stage flue of the combustion system. Each vision probe comprises a high-temperature resistant industrial camera, an active broadband illumination source, and an integrated protective cover with cooling and purging functions. Forced cooling utilizes instrument air to ensure stable operation of the camera's electronic components in a high-temperature environment; purging uses high-pressure instrument air to form an air curtain in front of the lens, continuously removing adhering fly ash particles and ensuring long-term cleanliness of the optical window and image clarity. Specifically, in the first-stage flue, i.e., the area containing the high-temperature superheater and reheater, multiple vision probes are installed on the observation holes of the furnace wall, with their field of view facing the windward side of the superheater or reheater tube bank. This area is where high-temperature molten slagging is most likely to occur. In the second-stage flue, i.e., the area containing the economizer, vision probes are used to monitor the ash accumulation on the tube bundles in the medium-temperature zone. In the third-stage flue, vision probes are positioned to observe the heat exchange elements of the air preheater to monitor low-temperature viscous or corrosive ash accumulation that is prone to occur there. The vision probe continuously acquires images of the heat exchanger tube array within its field of view at a preset frequency, forming a high-resolution image sequence data stream. The acquired image data can be used for subsequent processing to analyze characteristics such as the ash growth rate.

[0033] To effectively complement and verify visual information, this embodiment also incorporates sensing modules within the combustion system and flue. Multiple K-type or N-type thermocouples are arranged in a matrix at the inlet and outlet sections of each flue stage to measure the average flue gas temperature and the uniformity of the temperature field at that section. By calculating the flue gas temperature drop at the inlet and outlet of each heat exchanger stage, the actual heat exchange efficiency of the heat exchanger can be quantified in real time. Differential pressure transmitters are installed before and after the tube bundles of each flue heat exchanger stage to monitor the pressure loss of flue gas flowing through the tube bundles in real time. The differential pressure value is a direct quantitative indicator of the degree of flue blockage. Simultaneously, temperature and flow rate sensors are installed at the cold air main entering the air preheater and the hot air main leaving the air preheater to accurately calculate the heat absorption on the air side and the supply status of the combustion air.

[0034] All digital video signals acquired by the vision probes, as well as analog or digital signals output from thermocouples, differential pressure transmitters, flow meters, etc., are transmitted in real time via industrial Ethernet to a central data acquisition and preprocessing server. This server performs timestamp alignment, data cleaning, and storage on the received multi-source heterogeneous data, forming a structured, multi-dimensional time-series database that is synchronized with the boiler's operating status in real time.

[0035] Based on the online monitoring and data acquisition subsystem, the acquired image sequence data is processed. This process is completed by the feature extraction module, which targets the data deployed in the first... One flue ( The visual probe at all times Images acquired Perform analysis and output a feature vector describing the ash accumulation status of the flue. The feature vector consists of three parts: ash accumulation rate feature, color feature, and morphology and texture feature.

[0036] The ash accumulation rate is an indicator of the speed at which the ash accumulation problem develops. In this embodiment, the ash accumulation rate feature is calculated by combining image segmentation and temporal difference.

[0037] For the input raw image Preprocessing is performed, including grayscale conversion and Gaussian filtering to remove noise. The image is then binarized using the Otsu adaptive thresholding algorithm to obtain a binary mask image containing only the gray areas. Based on this mask image, the average thickness of the ash layer in the direction perpendicular to the heat exchanger tube surface was calculated. Ash accumulation rate Within a unit time window The change in average thickness within the interior. The calculation formula is as follows: in, Indicates the first Each flue is constantly The characteristics of ash accumulation rate; Indicates at time The calculated average thickness of the ash layer; Indicates the previous moment The calculated average thickness of the ash layer; It is the length of the time window used to calculate the rate.

[0038] Color is the basis for determining the chemical composition of accumulated dust. To eliminate the interference of lighting changes on color analysis, this embodiment first converts the pixels of the segmented dust accumulation area from the RGB color space to CIE L. a b Color space, where L The component represents brightness, a The component represents the range from green to red, b The component represents a range from blue to yellow. Calculate the L value for all pixels within this region. a b The statistical moments across the three channels, including the mean and standard deviation, are used to construct a six-dimensional color feature vector. .

[0039] ; in, ; ; Here, It is the first Each flue is constantly The six-dimensional color feature column vector, L , a , b The pixel value of one of the three color channels. At any moment The set of coordinates of gray pixels segmented from the image. It is a set The total number of pixels in the middle. It is a passage The pixel mean value represents the overall tone of the dust accumulation. It is a passage The pixel standard deviation characterizes the uniformity or heterogeneity of the gray color. and In this case, c can be any one of L, a, and b.

[0040] The morphology of loose, sintered, and molten ash, as well as its rough or smooth texture, are key factors in determining its bonding properties. This embodiment employs a pre-trained convolutional neural network as a deep feature extractor to simultaneously acquire morphological and textural information in an end-to-end manner. This network... Based on a backbone architecture (such as ResNet-50 or EfficientNet), and trained on the ImageNet dataset, fine-tuned on acquired grayscale images. At runtime, image patches cropped from the original images are used... Instead of using its final classification output as input to the network, its activation output before the global average pooling layer or the last fully connected layer is extracted as a high-dimensional feature vector for deep semantic meaning of the image. .

[0041] in, This represents the feature extraction function of a pre-trained convolutional neural network. The calculated ash accumulation rate, color features, morphology, and texture features are concatenated to output a feature vector describing the ash accumulation status of the flue. : .

[0042] This embodiment details the mid-level module that performs spatiotemporal information fusion based on the underlying feature extraction module. This module fuses multiple feature vectors from different flues and discrete at different time points into a unified context vector that can characterize the internal state correlation and dynamic evolution trend of the entire combustion system. This module is implemented through three steps: dynamic graph construction, spatial information fusion, and temporal information fusion.

[0043] To explicitly represent the physical connections and causal relationships between different levels of flue gas in the model, at each time step... The flue gas path of the combustion system is abstracted as a directed graph. .

[0044] The image Define a set of nodes Representing three flues, namely ,in Corresponding to the One flue, total number of nodes edge set The physical flow direction of the flue gas is represented by a set of directed edges. This indicates that the flue gas comes from Flow direction , and then from Flow direction Adjacency matrix It is a graph structure The mathematical representation of , where elements This indicates the existence of a slave node. point to The edge, j is the variable index ( ),otherwise In this embodiment, the adjacency matrix is ​​an upper triangular matrix. Node feature matrix .

[0045] To achieve information exchange between nodes in the flue, the purpose of each graph convolutional layer is to allow each node to aggregate information from its neighbors, thereby updating its own feature representation. For a standard graph convolutional layer, the propagation rule between layers can be expressed as: in, It is the first The node activation matrix feature representation of the layer, and the initial input , It is the first The weight matrix of the layer, It adds a self-loop adjacency matrix, where It is an identity matrix. Adding self-loops allows the aggregation of neighbor information while retaining the node's own information. yes The diagonal matrix, its diagonal elements . It is the result of symmetric normalization of the adjacency matrix, which can prevent drastic changes in the numerical scale of node features during the aggregation process and play a role in stabilizing training. It is a non-linear activation function. By stacking multiple such graph convolutional layers, the final output features of each node contain the state information of its upstream nodes. For example, a node... The output features will not only reflect its own condition, but also incorporate those from... and The information. After processing by the GCN module, at time... We obtain a feature matrix after spatial information fusion. ,in It is the output feature dimension of the GCN module.

[0046] To capture the evolution of ash accumulation over time, this embodiment outputs a sequence of feature matrices. The input is fed into a gated recurrent unit (GRU) network. The GRU unit at each time step... Update its internal hidden state This hidden state ( (For the hidden layer dimension) it encodes all spatiotemporal information up to the current moment. Reset gate This determines the extent to which previous hidden states should be ignored. Update Gate It controls the extent to which hidden state information from the previous moment is incorporated into the current state. Candidate hidden state Calculate the candidate hidden information at the current time step. Ultimately hidden state Combine the update gate with the candidate state and the state from the previous time step: ; and Both are parameter matrices. It is the bias vector. It is the Sigmoid activation function. It represents the Hadamardi (or Hadama) stack.

[0047] The final output of the middle-layer module is in a hidden state. It distributes and encodes the spatiotemporal fusion information of all flue nodes. A global average pooling operation is then performed on this hidden state to generate the system context vector. .

[0048] in, At any moment The system context vector.

[0049] By averaging the hidden state vectors of all nodes, we can obtain a representation that summarizes the overall state of the entire graph (i.e., the entire flue system) at the current moment. This vector... It does not favor any single node; it is a description of the overall situational balance and will serve as the sole input to the top-level decision-making network.

[0050] The top-level decision network is implemented using a multilayer perceptron, with the system context vector as the input layer. . It is the final output decision vector, where It is the number of controllable hazardous waste types. Decision vector. Each element ( The range of values ​​for ) is strictly limited to This value does not directly represent the absolute proportion of hazardous waste, but rather a normalized adjustment amount. A positive value indicates a recommendation to increase the proportion of this type of hazardous waste, while a negative value indicates a recommendation to decrease it; the absolute value represents the magnitude of the adjustment.

[0051] After the top-level decision-making network outputs a suggested adjustment to the hazardous waste mix ratio to optimize ash accumulation, the system does not execute it immediately. Instead, it uses the adjusted ratio as input to quickly estimate the flue gas temperature change at key sections of the combustion system caused by the new ratio. This ensures that the new ratio meets ash accumulation suppression requirements without negatively impacting the boiler's thermal conditions and operational safety.

[0052] This invention is based on the first law of thermodynamics and the fundamental equations of heat transfer, and solves the energy balance of the entire combustion system in different regions.

[0053] Heat balance in the furnace area and flue gas temperature at the furnace outlet The estimation is performed using a two-step method: the theoretical combustion temperature is calculated, and then the actual flue gas temperature at the furnace outlet is calculated based on the furnace thermal balance characteristics.

[0054] The theoretical combustion temperature is the highest temperature achievable under adiabatic conditions for complete combustion of fuel. It is obtained by solving the enthalpy balance equation, which equates the total enthalpy of reactants to the total enthalpy of products. Its expression is: ; The Reactant on the left side of the equation represents the total enthalpy of all substances involved in the combustion process. In the application scenario of this invention, the reactants mainly include fuel (a mixture of pulverized coal and various hazardous wastes) and combustion air.

[0055] The first reactant The molar amount (mol) of a chemical component (such as C, H, and S elements in fuel, and O2 and N2 in air). This refers to the initial temperature (K) of each reactant when it enters the furnace. It can be further subdivided into the fuel's inlet temperature. and the temperature of combustion air entering the furnace . Refers to the first The standard molar enthalpy of formation (J / mol) of the reactants under standard conditions (298.15 K, 1 atm). For stable elements such as O2 and N2, this value is 0. In engineering calculations, the chemical energy of the fuel is expressed using its lower heating value. The heat is directly incorporated into the heat balance in the form of the reactants. In this case, the left side of the equation can be simplified to the sum of the physical heat brought in by all reactants and the heat of combustion of fuel.

[0056] Refers to the first The reactant is heated from its standard state temperature to its initial inlet temperature. The increment of the enthalpy value. Its calculation method is as follows: ,in It is the molar isobaric specific heat capacity of this component.

[0057] The Products on the right side of the equation represents the total enthalpy of all substances produced after the combustion process. In the application scenario of this invention, : refers to the first in the product The molar amount (mol) of a chemical component (such as CO2, H2O, etc.). It is the theoretical combustion temperature (K) to be solved, that is, the highest temperature that the combustion products can reach under adiabatic conditions. Refers to the first Standard molar enthalpy of formation (J / mol) of the product under standard conditions. Refers to the first The product is heated from its standard state temperature to its theoretical combustion temperature. The physical enthalpy increase (J / mol) at that time.

[0058] Furnace outlet flue gas temperature Always below the theoretical combustion temperature This is because the water-cooled walls inside the furnace absorb a large amount of heat through radiation. This embodiment uses a semi-empirical formula, widely used in boiler engineering and based on the radiative heat transfer characteristics of the furnace, for calculation: ; This formula describes the dimensionless furnace outlet temperature. Boltzmann number, which characterizes furnace properties The relationship between them. Among them: and Thermodynamic temperature is used in all cases. Boltzmann number is the dimensionless ratio of the effective heat of fuel combustion to the radiative energy of the furnace at the theoretical combustion temperature, and is defined as: This is the total fuel feed rate (kg / s). It is the weighted average lower heating value (J / kg) of the blended fuel calculated based on the new proportions. It refers to the boiler combustion efficiency. It is the total effective heat exchange area (m²) of the furnace water-cooled walls. It is the Stefan-Boltzmann constant. and These are empirical coefficients related to furnace geometry, burner arrangement, and fuel type. For Π-type boilers, Between 0.4 and 0.6, (between 0.6 and 0.7), and can be accurately calibrated by regression analysis of historical boiler operating data.

[0059] For a specific hazardous waste co-incineration scenario, an exemplary new hazardous waste mix is ​​given: {Pulverized coal: 80%, Class A hazardous waste: 5%, Class B hazardous waste: 10%, Class C hazardous waste: 5%}. The lower heating values ​​of each component are as follows: MJ / kg, MJ / kg, MJ / kg, MJ / kg ; ; The combined lower heating value of the blended fuel is 24.15 MJ / kg.

[0060] Elemental analysis: C: 66.0%, H: 4.5%, O: 8.5%, N: 1.2%, S: 1.8%, Ash: 15.0%, Moisture: 3.0%.

[0061] Operating parameters: Total fuel feed rate kg / s, boiler combustion efficiency Excess air coefficient Combustion air temperature ℃ (573.15 K). Boiler structure and characteristic parameters: Total effective heat exchange area of ​​the furnace water-cooled wall. m². Empirical coefficients for furnace characteristics: , Average isobaric specific heat capacity of flue gas (approximate value): kJ / (kg·K).

[0062] Based on the aforementioned enthalpy balance principle, the total effective heat on the reactant side is calculated, and then the temperature on the product side is determined using an iterative method. After detailed enthalpy balance iterative calculations, the convergent theoretical combustion temperature under this operating condition is obtained as follows: Calculate the flue gas temperature at the furnace outlet According to the calculation of the Boltzmann number : Then, the furnace temperature characteristic formula in the application is used for calculation. : Converted to Celsius: ℃. This is the inlet temperature of the first-stage flue.

[0063] The intelligent control system receives the normalized ratio adjustment amount output by the top-level decision-making network. Then, a pre-execution verification process is initiated. The system performs verification based on the current hazardous waste ratio. and decision vector Generate a candidate new ratio to be verified. Subsequently, the system calculated the outlet flue gas temperature and the inlet temperature of the first-stage flue gas duct for the candidate new furnace configuration. The system compares the two key temperature values ​​obtained with preset safety and efficiency constraint ranges. These constraint ranges are determined jointly by process requirements and equipment safety limits. The preferred furnace temperature constraint range is: [ , [1100℃, 1450℃] This is to ensure stable and complete fuel combustion, suppress CO generation, and ensure the lower limit of heat exchange efficiency in the high-temperature zone; This is the upper limit to prevent excessive NOx formation due to overheating and to protect the safety of the water-cooled walls. First-stage flue temperature constraint range: [ , [850℃, 1150℃] This is to ensure that the parameters of the superheated steam meet the lower limit of the turbine's power requirements; This is the upper limit to prevent overheating creep and malignant slagging in the tubes of the high-temperature superheater.

[0064] Based on the comparison results, the system executes the following logic: Scenario 1: Fully complies with the constraints, if and Then determine the candidate new ratio. The formula is safe and feasible. The system has officially adopted this formula and has issued it to the hazardous waste feeding control system for execution.

[0065] Scenario 2: Constraints Not Met. If any predicted temperature exceeds the constraint range, the candidate new mix ratio is determined to be infeasible, and a decision vector correction procedure is initiated. The goal of this procedure is to correct the adjustment magnitude while maintaining the original adjustment direction, so that the temperature returns to the allowable range. The correction scheme adopts an adjustment strategy based on total calorific value feedback: if the predicted temperature is too low, it indicates that the total calorific value of the candidate mix ratio is insufficient. The system will adjust the original decision vector... Based on this, an adjustment term positively correlated with the total calorific value is applied. For example, the system will slightly increase the proportion of Class B hazardous waste with the highest known calorific value in the database, and slightly decrease the proportion of Class C hazardous waste with the lowest calorific value, while maintaining the suppressive trend against Class A hazardous waste that causes ash accumulation. If the predicted temperature is too high: it indicates that the total calorific value of the candidate mix is ​​too high. The system then performs the opposite operation, for example, slightly decreasing the proportion of high-calorific-value Class B hazardous waste and / or slightly increasing the proportion of low-calorific-value Class C hazardous waste. After correction, a new decision vector is obtained. The system will then use the correction vector again to generate new candidate ratios and re-predict the temperature. This process will be repeated until a ratio scheme that satisfies both the ash accumulation suppression trend and the temperature constraint is found, and then it will be officially implemented.

[0066] When a new formula After execution begins, the system enters a continuous experimental feedback adjustment phase to address potential deviations between the theoretical model and actual operating conditions.

[0067] In the proportioning scheme After a period of operation, the system continuously monitors the measured temperature at the outlet of the second-stage flue through the sensor module. Measured temperature at the outlet of the third-stage flue .

[0068] The system compares these two measured temperatures with the target temperature ranges set for the medium and low temperature zones. For example: Target range for the second-stage flue: [ , ], Target section of the third-level flue: [ , These range settings ensure optimal heat exchange efficiency and operational safety for the economizer and air preheater.

[0069] Deviation-based fine-tuning: If the measured temperature is within the target range: this indicates the current mixing ratio scheme... It performed well under all operating conditions, and the system continued to maintain this ratio while continuously performing visual monitoring.

[0070] If the measured temperature deviates from the target range, it indicates that the theoretical model's prediction of heat transfer in the medium and low temperature zones is flawed, or that the actual impact of dust accumulation on this area exceeds expectations. In this case, the system will initiate a fine-tuning program based on PID control principles.

[0071] like or A persistently low temperature indicates excessive heat exchange in the medium and low temperature zones or that the upstream flue gas temperature is lower than expected. The system will make minor adjustments to the mix ratio, such as slightly increasing the total calorific value (by slightly increasing the amount of Class B hazardous waste) to raise the overall flue gas temperature.

[0072] like or A persistently high temperature indicates insufficient heat exchange in the medium and low temperature zones, possibly due to ash accumulation in these areas. In addition to instructing the vision module to strengthen monitoring of these areas, the system will also fine-tune the proportions, such as slightly reducing the total calorific value or adjusting the hazardous waste components related to the ash accumulation characteristics at medium and low temperatures (e.g., slightly reducing Class B hazardous waste if sulfur corrosion is suspected).

[0073] The embodiments in this specification also propose a multi-modal sensing hazardous waste co-incineration cogeneration control system, which includes: Acquisition module: Acquires operating condition data of three flues within the combustion system, including at least image data characterizing the ash accumulation state of the flues. Feature extraction module: Extracts at least one ash accumulation feature that characterizes the ash accumulation state, performs spatiotemporal fusion of the ash accumulation features at the three flue locations, and generates a system context vector that characterizes the global ash accumulation state of the combustion system; Decision generation module: Based on the system context vector, generates a decision vector for adjusting the hazardous waste ratio. Correction module: Corrects the decision vector until the predicted temperature corresponding to the corrected decision vector meets the temperature constraint range, adjusts the hazardous waste ratio according to the decision vector; and monitors the measured temperature of the combustion system at one or more locations. If the measured temperature deviates from the preset target temperature range, the current hazardous waste ratio is adjusted accordingly.

[0074] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned multi-mode sensing method for controlling the co-incineration of hazardous waste with combined heat and power.

[0075] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned multi-mode sensing method for the control of hazardous waste co-incineration combined heat and power.

[0076] A computer program product, characterized in that the computer program product includes a computer program, which, when executed by a processor, implements the above-mentioned multi-mode sensing method for the control of hazardous waste co-incineration and cogeneration.

[0077] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0078] In this specification, the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the descriptions of the embodiments described later are relatively simple, and relevant parts can be referred to the descriptions of the foregoing embodiments.

[0079] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A multi-modal sensing method for the regulation and control of hazardous waste co-incineration combined heat and power, characterized in that, The method includes: acquiring operating condition data of three flues in the combustion system, wherein the operating condition data includes at least image data characterizing the ash accumulation state of the flues; Extract at least one ash accumulation feature to characterize the ash accumulation state, and perform spatiotemporal fusion of the ash accumulation features at the three flue locations to generate a system context vector characterizing the global ash accumulation state of the combustion system. Based on the system context vector, a decision vector for adjusting the hazardous waste ratio is generated; The decision vector is corrected until the predicted temperature corresponding to the corrected decision vector meets the temperature constraint range. The hazardous waste ratio is adjusted according to the decision vector. The measured temperature of the combustion system at one or more locations is monitored. If the measured temperature deviates from the preset target temperature range, the current hazardous waste ratio is adjusted accordingly.

2. The method for multi-mode sensing-based control of hazardous waste co-incineration combined heat and power generation according to claim 1, characterized in that, The aforementioned ash accumulation features include ash accumulation rate features calculated based on temporal differences of image sequences; color features calculated based on statistical moments in the color space; and morphological and texture features extracted by a pre-trained convolutional neural network.

3. The method for multi-mode sensing-based co-incineration and cogeneration control of hazardous waste according to claim 2, characterized in that, The three flues are constructed as a directed graph with connections, where each flue is a node and its corresponding ash accumulation feature is a node feature. By aggregating the node features of each node and its upstream neighbor nodes, a node feature representation with fused spatial correlation is updated and generated. The time series of node features with fused spatial correlation is input into a gated recurrent unit network to capture the dynamic evolution of the ash accumulation state over time and output the system context vector.

4. The method for multi-mode sensing-based co-incineration and cogeneration control of hazardous waste according to claim 3, characterized in that: The decision vector includes multiple adjustment values ​​corresponding one-to-one with various controllable hazardous materials; wherein, the sign of each adjustment value is used to determine the direction of adjustment for increasing or decreasing the proportion of the corresponding hazardous waste, and the absolute value is used to determine the relative adjustment magnitude in the direction of adjustment.

5. The method for multi-mode sensing-based co-incineration and cogeneration control of hazardous waste according to claim 4, characterized in that: The step of correcting the decision vector is a loop-iterative verification process, which specifically includes: (1) determining a candidate hazardous waste to fuel ratio based on the initial or previous corrected decision vector; (2) predicting the temperature at at least one critical location inside the combustion system caused by the candidate ratio; (3) comparing the predicted temperature with a preset temperature constraint range; (4) if the predicted temperature does not conform to the temperature constraint range, correcting the current decision vector and returning to step (1) until the predicted temperature conforms to the constraint range.

6. The method for multi-mode sensing-based control of hazardous waste co-incineration combined heat and power generation according to claim 5, characterized in that: The step of feedback adjustment of the current hazardous waste and fuel ratio is a secondary optimization process to maintain the stability of the thermal conditions of the flue in the medium and low temperature zones. The feedback adjustment is to fine-tune the total calorific value corresponding to the hazardous waste and fuel ratio based on the deviation between the measured temperature of the second-stage flue and / or the third-stage flue and the set target temperature range.

7. A multi-mode sensing hazardous waste co-incineration cogeneration control system, the system being used to execute a multi-mode sensing hazardous waste co-incineration cogeneration control method as described in any one of claims 1-6, characterized in that, The system includes: Acquisition module: Acquires operating condition data of three flues within the combustion system, including at least image data characterizing the ash accumulation state of the flues. Feature extraction module: Extracts at least one ash accumulation feature that characterizes the ash accumulation state, performs spatiotemporal fusion of the ash accumulation features at the three flue locations, and generates a system context vector that characterizes the global ash accumulation state of the combustion system; Decision generation module: Based on the system context vector, generates a decision vector for adjusting the hazardous waste ratio. Correction module: Corrects the decision vector until the predicted temperature corresponding to the corrected decision vector meets the temperature constraint range, adjusts the hazardous waste ratio according to the decision vector; and monitors the measured temperature of the combustion system at one or more locations. If the measured temperature deviates from the preset target temperature range, the current hazardous waste ratio is adjusted accordingly.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements a multi-mode sensing method for controlling hazardous waste co-incineration combined heat and power as described in any one of claims 1-6.

9. A computer-readable storage medium storing a computer program that, when executed by a processor, implements a multi-mode sensing method for controlling hazardous waste co-incineration combined heat and power as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the multi-mode sensing method for controlling hazardous waste co-incineration combined heat and power generation as described in any one of claims 1-6.

Images