Coal blending combustion optimization system and method based on real-time as-fired coal quality data

Abstract

The embodiment of the invention discloses a coal blending combustion optimization system and method based on real-time in-furnace coal quality data, and the method comprises the steps: receiving a data set of boiler combustion conditions (a combustion temperature value, an air door opening value, a fan rotating speed value and a combustion chamber pressure value) collected by a sensor assembly, and obtaining the real-time in-furnace coal quality data; carrying out time domain aggregation and multi-scale semantic coding on the boiler combustion condition by adopting an AI-based data processing and coding algorithm, and carrying out embedded coding on the real-time boiler entering coal quality data; therefore, the coal mixing proportion is intelligently recommended according to prompt response interactive representation between the real-time in-furnace coal quality data embedding features and the boiler combustion condition multi-scale time sequence semantic coding features. In this way, real-time monitoring and intelligent control over the combustion process of the coal-fired boiler are achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent coal blending optimization, and more specifically, relates to a coal blending and combustion optimization system and method based on real-time coal quality data entering a furnace. Background Art

[0002] Coal-fired power plants often face reduced efficiency and environmental problems due to unstable coal quality during operation. Coal blending refers to the process of mixing different types and qualities of coal in a certain proportion for boiler combustion. This practice is very common in coal-fired power plants and aims to improve combustion efficiency, reduce pollutant emissions, and reduce costs by optimizing the blending ratio of coal. Different types of coal have different combustion characteristics. Through scientific and reasonable coal blending, coal can reach the best state during the combustion process, thereby improving thermal efficiency and combustion stability while reducing environmental pollution.

[0003] However, changes in coal quality have seriously affected combustion efficiency and pollutant emissions. The traditional fixed-ratio blending method has obvious defects and cannot adapt to such changes. Specifically, fixed-ratio blending lacks flexibility and cannot adjust the mixing ratio according to the real-time coal quality, resulting in low combustion efficiency. Different batches of coal have significant differences in ash, sulfur, volatile matter and moisture. The fixed ratio cannot fully optimize the combustion process, which affects thermal efficiency and combustion stability. In addition, this static method is difficult to deal with the problem of combustion instability caused by coal quality fluctuations, increasing the risk of equipment failure. Changes in coal quality may cause fluctuations in combustion conditions such as combustion temperature and pressure, resulting in incomplete combustion and affecting the stability of the combustion process.

[0004] Therefore, an optimization scheme for coal blending and combustion based on real-time coal quality data entering the furnace is desired. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art, and discloses a coal blending and combustion optimization system and method based on real-time coal quality data entering the furnace.

[0006] In a first aspect, an embodiment of the present invention provides a coal blending and combustion optimization system based on real-time coal quality data entering a furnace, comprising:

[0007] A combustion condition data acquisition module, used to receive a data set of boiler combustion conditions collected by the sensor assembly;

[0008] A combustion condition data sorting module, used for sorting the data set of the boiler combustion condition to obtain a boiler combustion condition time domain aggregation matrix;

[0009] A combustion condition analysis module, used for inputting the boiler combustion condition time domain aggregation matrix into a boiler combustion condition multi-scale feature extractor to obtain a boiler combustion condition multi-scale temporal semantic coding feature map;

[0010] The coal quality data acquisition module is used to obtain real-time coal quality data.

[0011] The coal quality embedding coding module is used to perform structured coding on the real-time coal quality data to obtain the real-time coal quality data embedding coding vector;

[0012] A coal quality combustion condition response coding module is used to perform semantic response coding of modal interactive prompts on the real-time coal quality data embedded coding vector and the boiler combustion condition multi-scale temporal semantic coding feature map to obtain a coal quality-combustion condition semantic response coding feature map;

[0013] The optimization module is used to obtain an optimization instruction based on the coal quality-combustion condition semantic response coding feature map entering the furnace, and the optimization instruction includes a decoded value of a recommended coal mixing ratio.

[0014] In some possible embodiments, the boiler combustion conditions include a combustion temperature value, a damper opening value, a fan speed value, and a combustion chamber pressure value.

[0015] In some possible embodiments, the boiler combustion condition multi-scale feature extractor includes a first convolutional neural network model and a second convolutional neural network model in parallel, and the first convolutional neural network model and the second convolutional neural network model have different convolution kernel sizes.

[0016] In some possible embodiments, the incoming coal quality embedding coding module is used to: use the incoming coal quality data embedding coding matrix to perform structured coding on the real-time incoming coal quality data to obtain the real-time incoming coal quality data embedding coding vector.

[0017] In some possible embodiments, the coal quality combustion condition response coding module includes:

[0018] A combustion condition feature local decomposition unit, used for performing local feature decomposition along the channel dimension on the boiler combustion condition multi-scale temporal semantic coding feature map to obtain a set of boiler combustion condition multi-scale temporal semantic local feature matrices;

[0019] A coal quality combustion condition local query prompt coding unit, used for performing local query prompt coding on the set of the real-time incoming coal quality data embedding coding vector and the boiler combustion condition multi-scale temporal semantic local feature matrix to obtain a set of incoming coal quality-combustion condition local query prompt semantic coding vectors;

[0020] The coal quality and combustion condition mask unit is used to embed the real-time coal quality data into the coding vector and each coal quality and combustion condition local query prompt semantic coding vector in the set of coal quality and combustion condition local query prompt semantic coding vector into a cross-modal mask weaving network based on prompt information to obtain a set of coal quality and combustion condition local feature mask weight matrices based on prompt information;

[0021] The coal quality and combustion condition response aggregation unit is used to perform significant semantic response aggregation on the set of multi-scale temporal semantic local feature matrices of the boiler combustion conditions based on the set of coal quality and combustion condition local feature mask weight matrices based on the prompt information to obtain the semantic response encoding feature map of the coal quality and combustion condition.

[0022] In some possible embodiments, the coal quality and combustion condition local query prompt encoding unit is used to: use the real-time entering coal quality data embedded coding vector as the query vector and each boiler combustion condition multi-scale temporal semantic local feature matrix in the set of boiler combustion condition multi-scale temporal semantic local feature matrices as the key matrix, and input the query vector and the key matrix into a prompt learning network based on a converter structure to obtain the set of the entering coal quality-combustion condition local query prompt semantic coding vectors.

[0023] In some possible embodiments, the coal quality combustion condition response polymerization unit is used to:

[0024] Calculate the set of the set of the local feature mask weight matrices of the incoming coal quality-combustion condition based on the prompt information and the set of the multi-scale temporal semantic local feature matrices of the boiler combustion conditions, and obtain the set of the local granularity significant interaction matrices of the incoming coal quality-combustion condition by multiplying the position points between each corresponding set of the local feature mask weight matrices of the incoming coal quality-combustion condition based on the prompt information and the multi-scale temporal semantic local feature matrices of the boiler combustion conditions;

[0025] The set of the local granularity significant interaction matrices of the coal quality entering the furnace and the combustion conditions is feature aggregated to obtain the semantic response coding feature map of the coal quality entering the furnace and the combustion conditions.

[0026] In some possible embodiments, the optimization module is used to: input the coal quality-combustion condition semantic response encoding feature map into the decoder-based coal blending optimization module to obtain the optimization instruction, and the optimization instruction includes the decoded value of the recommended coal mixing ratio.

[0027] In a second aspect, an embodiment of the present invention provides a method for optimizing coal blending and combustion based on real-time coal quality data entering a furnace, comprising:

[0028] receiving a data set of boiler combustion conditions collected by a sensor assembly;

[0029] Arranging the data set of the boiler combustion condition to obtain a boiler combustion condition time domain aggregation matrix;

[0030] Inputting the boiler combustion condition time domain aggregation matrix into a boiler combustion condition multi-scale feature extractor to obtain a boiler combustion condition multi-scale temporal semantic coding feature map;

[0031] Obtain real-time coal quality data entering the furnace;

[0032] Performing structured coding on the real-time coal quality data entering the furnace to obtain an embedded coding vector of the real-time coal quality data entering the furnace;

[0033] The real-time coal quality data embedding coding vector and the multi-scale temporal semantic coding feature map of boiler combustion conditions are subjected to semantic response coding of modal interaction prompts to obtain a semantic response coding feature map of coal quality-combustion conditions;

[0034] Based on the coal quality-combustion condition semantic response coding feature map, an optimization instruction is obtained, and the optimization instruction includes a decoded value of a recommended coal mixing ratio.

[0035] In some possible embodiments, the boiler combustion conditions include a combustion temperature value, a damper opening value, a fan speed value, and a combustion chamber pressure value.

[0036] Compared with the prior art, the embodiment of the present invention provides a coal blending optimization system and method based on real-time coal quality data entering the furnace, which receives a data set of boiler combustion conditions (combustion temperature value, air door opening value, fan speed value and combustion chamber pressure value) collected by the sensor component, and obtains real-time coal quality data entering the furnace, and uses AI-based data processing and coding algorithms to perform time domain aggregation and multi-scale semantic coding on the boiler combustion conditions, and embeds the real-time coal quality data entering the furnace, so as to intelligently recommend the coal mixing ratio based on the prompt response interaction between the embedded features of the real-time coal quality data entering the furnace and the multi-scale temporal semantic coding features of the boiler combustion conditions. In this way, the coal mixing ratio can be adjusted in real time and dynamically according to the characteristics of different coal qualities and in combination with the combustion conditions, ensuring that the combustion process is always in the best state, improving thermal efficiency and combustion stability, thereby realizing real-time monitoring and intelligent control of the combustion process of coal-fired boilers. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The above and other features, advantages and aspects of the embodiments of the present invention will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0038] Figure 1 A block diagram of a coal blending and combustion optimization system based on real-time coal quality data according to an embodiment of the present invention;

[0039] Figure 2 Schematic diagram of data flow of a coal blending and combustion optimization system based on real-time coal quality data according to an embodiment of the present invention;

[0040] Figure 3 It is a block diagram of a coal quality combustion condition response encoding module in a coal blending and combustion optimization system based on real-time coal quality data entering the furnace according to an embodiment of the present invention;

[0041] Figure 4 The present invention is a flowchart of a method for optimizing coal blending and combustion based on real-time coal quality data entering the furnace according to an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work are within the scope of protection of the present invention.

[0043] Unless otherwise specified, the technical terms or scientific terms used in the embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. "Including" or "comprising" used in the embodiments of the present invention neither limit the shapes, numbers, steps, actions, operations, components, originals and / or their groups mentioned, nor exclude the appearance or addition of one or more other different shapes, numbers, steps, actions, operations, components, originals and / or their groups, or the addition of these. In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number and order of the indicated technical features. Thus, the features defined as "first" and "second" can explicitly or implicitly include one or more of the features. In the description of the embodiments of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0044] Unless otherwise specifically stated, the relative arrangement of the components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention. At the same time, it should be understood that for ease of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship, and the techniques, methods and devices known to ordinary technicians in the relevant fields may not be discussed in detail, but where appropriate, the techniques, methods and devices shown should be considered as part of the authorized specification. In all examples shown and discussed here, any specific other examples may have different values. It should be noted that similar symbols and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0045] In the description of the embodiments of the present invention, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In the embodiments of the present invention, the schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in the embodiments of the present invention and the features of the different embodiments or examples, without contradiction.

[0046] Coal-fired power plants often face reduced efficiency and environmental problems due to unstable coal quality during operation. Coal blending refers to the process of mixing different types and qualities of coal in a certain proportion for boiler combustion. This practice is very common in coal-fired power plants and aims to improve combustion efficiency, reduce pollutant emissions, and reduce costs by optimizing the blending ratio of coal. Different types of coal have different combustion characteristics. Through scientific and reasonable coal blending, coal can reach the best state during the combustion process, thereby improving thermal efficiency and combustion stability while reducing environmental pollution.

[0047] However, changes in coal quality have seriously affected combustion efficiency and pollutant emissions. The traditional fixed-ratio blending method has obvious defects and cannot adapt to such changes. Specifically, fixed-ratio blending lacks flexibility and cannot adjust the mixing ratio according to the real-time coal quality, resulting in low combustion efficiency. Different batches of coal have significant differences in ash, sulfur, volatile matter and moisture. The fixed ratio cannot fully optimize the combustion process, which affects thermal efficiency and combustion stability. In addition, this static method is difficult to deal with the problem of combustion instability caused by coal quality fluctuations, increasing the risk of equipment failure. Changes in coal quality may cause fluctuations in combustion conditions such as combustion temperature and pressure, resulting in incomplete combustion and affecting the stability of the combustion process.

[0048] Therefore, in response to the above technical problems, the technical concept of the present invention is to receive a data set of boiler combustion conditions (combustion temperature value, damper opening value, fan speed value and combustion chamber pressure value) collected by the sensor assembly, and obtain real-time coal quality data entering the furnace, and use AI-based data processing and encoding algorithms to perform time domain aggregation and multi-scale semantic encoding on the boiler combustion conditions, and embed the real-time coal quality data entering the furnace to perform encoding, so as to intelligently recommend the coal mixing ratio based on the prompt response interaction between the embedded features of the real-time coal quality data entering the furnace and the multi-scale temporal semantic coding features of the boiler combustion conditions. In this way, the coal mixing ratio can be adjusted in real time and dynamically according to the characteristics of different coal qualities and in combination with the combustion conditions, ensuring that the combustion process is always in the best state, improving thermal efficiency and combustion stability, thereby realizing real-time monitoring and intelligent control of the combustion process of coal-fired boilers.

[0049] Figure 1 4 is a block diagram of a coal blending and combustion optimization system based on real-time coal quality data according to an embodiment of the present invention. Figure 2 This is a data flow diagram of a coal blending and combustion optimization system based on real-time coal quality data according to an embodiment of the present invention. Figure 1 and Figure 2As shown, according to an embodiment of the present invention, a coal blending optimization system 100 based on real-time in-furnace coal quality data includes: a combustion condition data acquisition module 110, which is used to receive a data set of boiler combustion conditions collected by a sensor component; a combustion condition data sorting module 120, which is used to sort the data set of the boiler combustion conditions to obtain a boiler combustion condition time domain aggregation matrix; a combustion condition analysis module 130, which is used to input the boiler combustion condition time domain aggregation matrix into a boiler combustion condition multi-scale feature extractor to obtain a boiler combustion condition multi-scale temporal semantic coding feature map; an in-furnace coal quality data acquisition module 140, which is used to obtain real-time Coal quality data entering the furnace; an embedded coding module 150 for the coal quality entering the furnace, which is used to perform structured coding on the real-time embedded coding vector of the coal quality data entering the furnace; a coal quality combustion condition response coding module 160, which is used to perform semantic response coding of the real-time embedded coding vector of the coal quality data entering the furnace and the multi-scale temporal semantic coding feature map of the boiler combustion conditions with modal interactive prompts to obtain a semantic response coding feature map of the coal quality entering the furnace-combustion condition semantic response coding feature map; an optimization module 170, which is used to obtain an optimization instruction based on the semantic response coding feature map of the coal quality entering the furnace-combustion condition, wherein the optimization instruction includes a decoded value of a recommended coal mixing ratio.

[0050] In an embodiment of the present invention, the combustion condition data acquisition module 110 is used to receive a data set of boiler combustion conditions collected by a sensor assembly. In particular, the boiler combustion conditions described herein include a combustion temperature value, a damper opening value, a fan speed value, and a combustion chamber pressure value. It should be understood that the boiler combustion conditions refer to various parameters that affect the combustion efficiency and safety during the operation of the boiler, and these parameters are of great significance for improving the overall performance of the boiler. Specifically, the combustion temperature is an important indicator for evaluating the degree of fuel combustion. Appropriate high temperatures can ensure that the combustible substances in the fuel are completely burned, thereby improving the energy conversion efficiency and reducing the emission of incomplete combustion products. Too high a temperature may cause equipment damage or increase the generation of nitrogen oxides, while too low a temperature may cause incomplete combustion, resulting in excessive smoke and harmful gases. The damper opening determines the amount of air entering the combustion chamber. A suitable damper opening can ensure a good mixing of fuel and air and promote the smooth progress of the combustion reaction. If the damper opening is too large, too much cold air will be introduced, resulting in a decrease in combustion efficiency; conversely, if the opening is too small, it may cause insufficient oxygen supply, which also affects the combustion effect. The fan speed directly affects the oxygen supply, which in turn affects the combustion efficiency. By adjusting the fan speed, the oxygen supply during the combustion process can be controlled to ensure the full combustion of the fuel. The correct fan speed setting is very important for maintaining a stable combustion process and improving thermal efficiency. The pressure level in the combustion chamber reflects the stability and safety of the combustion process. Appropriate pressure helps to maintain the stability of the flame and avoid the occurrence of flashback. In addition, a reasonable pressure in the combustion chamber can also help improve combustion efficiency and reduce pollutant emissions. Based on this, in the technical solution of the present invention, receiving a data set of boiler combustion conditions collected by the sensor assembly, and analyzing and processing it can provide data support for the subsequent accurate recommendation of the coal mixing ratio. That is, by reasonably allocating the proportions of different types of coal, the combustion characteristics of the coal can be optimized, the stability of the combustion process can be ensured, and combustion instability caused by fluctuations in coal quality can be avoided.

[0051] The deployment of sensor components is the basis of the entire data collection. In the boiler system of a coal-fired power plant, various types of sensors need to be installed to monitor and collect data on boiler combustion conditions in real time. These sensors include temperature sensors, pressure sensors, flow sensors, and position sensors. The temperature sensor is used to measure the temperature distribution in the combustion chamber to ensure that the temperature during the combustion process is controlled within a reasonable range; the pressure sensor is used to monitor the pressure difference inside and outside the combustion chamber to prevent safety accidents caused by excessive or low pressure; the flow sensor is used to detect the air flow entering the combustion chamber to ensure sufficient oxygen supply; the position sensor is used to monitor the opening of the damper to ensure the accuracy of air volume regulation. These sensors need to be installed in key parts of the boiler, such as the combustion chamber, air duct, fan, etc., to ensure that the required combustion condition data can be fully and accurately collected.

[0052] After the sensor components are installed, it is very important to ensure that these sensors can work stably and reliably. This includes regular calibration of sensors to ensure the accuracy of their measurements; and maintenance of sensors to prevent environmental factors such as dust and high temperature from affecting sensor performance. For example, temperature sensors may drift due to long-term exposure to high temperature environments and need to be calibrated regularly. In addition, a fault detection and alarm system needs to be established to issue an alarm in time once abnormal sensor data is found so that technicians can check and repair. For example, if the reading of a temperature sensor suddenly jumps, it may be a sensor failure or poor contact. The system should immediately issue an alarm to remind technicians to check.

[0053] The process of data acquisition is to convert physical signals into electrical signals through sensors, and then convert the electrical signals into digital signals through data acquisition cards or data acquisition modules, and finally transmit them to the central control system. In this process, data can be transmitted by wired transmission or wireless transmission. Wired transmission usually adopts industrial standard communication protocols, such as Modbus, Profibus, etc., to ensure the stability and reliability of data transmission. Wireless transmission can use wireless communication technologies such as Wi-Fi and Zigbee, which are suitable for occasions where wiring is difficult, but signal interference and transmission distance issues need to be considered. For example, in large coal-fired power plants, sensors are widely distributed, and wired transmission may be costly and difficult to wire. At this time, the use of wireless transmission technology can effectively solve this problem.

[0054] After data collection is completed, the collected data needs to be preliminarily processed to ensure the quality of the data. Data preprocessing mainly includes steps such as data cleaning, data correction and data labeling. Data cleaning is mainly to remove outliers and missing values ​​in the data to ensure the integrity of the data. For example, if a sensor fails, it may cause data loss or abnormality. At this time, interpolation, smoothing and other methods can be used to fill in missing values ​​or correct abnormal values. Data correction is to calibrate the data collected by the sensor to eliminate the error of the sensor itself. For example, the temperature sensor may have a certain measurement deviation, which needs to be corrected by the correction coefficient. Data labeling is to mark the data to facilitate subsequent data analysis and processing. For example, data in different time periods can be classified and labeled to facilitate the analysis of combustion conditions in different time periods. Through this process, the collected boiler combustion condition data can be ensured to be accurate and complete, providing a solid data foundation for the subsequent intelligent coal blending optimization.

[0055] In an embodiment of the present invention, the combustion condition data sorting module 120 is used to sort the data set of the boiler combustion condition to obtain a boiler combustion condition time domain aggregation matrix. Accordingly, considering that the combustion state in different time periods in the boiler combustion condition is different, and there is a correlation between the various combustion parameters on a time scale, therefore, in order to observe the change trend and periodic characteristics of the boiler combustion condition over time, it is helpful to identify abnormal conditions in the combustion process, such as temperature mutations, pressure fluctuations, etc., in the technical solution of the present invention, the data set of the boiler combustion condition is sorted to obtain a boiler combustion condition time domain aggregation matrix. The specific implementation process of this step is as follows:

[0056] First, after the sensor component collects the data set of boiler combustion conditions, these data need to be preliminarily processed to ensure the quality and integrity of the data.

[0057] After completing the data preprocessing, the next step is to normalize the data set of boiler combustion conditions. Normalization is to convert data of different units and magnitudes to the same dimension to facilitate subsequent feature extraction and analysis. Boiler combustion conditions include combustion temperature values, damper opening values, fan speed values, and combustion chamber pressure values. The units and magnitudes of these parameters are different. For example, the combustion temperature value is usually in degrees Celsius (℃), the damper opening value is in percentage (%), the fan speed value is in speed (rpm), and the combustion chamber pressure value is in Pascal (Pa). Normalization can convert these data of different units to a numerical range between 0-1. The specific method can use Min-Max Normalization or Z-Score Standardization.

[0058] After the normalization process is completed, the next step is feature extraction. Feature extraction is to extract features that are important to the combustion process from the original data. These features can be the mean and variance of the temperature, the rate of change of the damper opening, the fluctuation of the fan speed, etc. The purpose of feature extraction is to reduce the dimension of the data and extract the information that best reflects the characteristics of the combustion process. For example, the mean and variance of the combustion temperature in each time window can be calculated to reflect the stability and fluctuation of the temperature; the rate of change of the damper opening can be calculated to reflect the frequency and amplitude of the air volume adjustment; the fluctuation of the fan speed can be calculated to reflect the stability of the oxygen supply.

[0059] Time series analysis is an important part of feature extraction. Through time series analysis, periodic and trend characteristics in data can be identified. For example, methods such as Fourier transform or wavelet transform can be used to analyze the trend of combustion temperature over time. Fourier transform can convert time domain data to frequency domain and identify periodic components in the data; wavelet transform can simultaneously analyze the time and frequency characteristics of data and identify features on different time scales. Through time series analysis, the dynamic changes of the combustion process can be more fully understood.

[0060] After completing feature extraction and time series analysis, the next step is to perform time series aggregation on the boiler combustion condition data to form a boiler combustion condition time domain aggregation matrix. Time series aggregation is to aggregate the combustion condition data at different time points into a matrix, where each row represents a time point and each column represents a combustion condition parameter. In this way, the changes in boiler combustion conditions over time can be intuitively displayed. For example, the combustion temperature values, damper opening values, fan speed values, and combustion chamber pressure values ​​collected every 1 minute in a day can be combined into a time domain aggregation matrix. The number of rows in the matrix is ​​1440 (the number of minutes in a day) and the number of columns is 4 (four combustion condition parameters). In this way, the combustion condition data at each time point is recorded to form a complete time series data set.

[0061] Specifically, suppose there are the following four combustion condition parameter data: combustion temperature value T(t), damper opening value v(t), fan speed value v(t), combustion chamber pressure value P(t), these data are arranged by time point to form a matrix S, each row represents the combustion condition data at a time point, and each column represents a combustion condition parameter. In this way, the changing trend of each combustion condition parameter over time can be clearly seen, providing a basis for subsequent multi-scale feature extraction.

[0062] In an embodiment of the present invention, the combustion condition analysis module 130 is used to input the boiler combustion condition time domain aggregation matrix into the boiler combustion condition multi-scale feature extractor to obtain a boiler combustion condition multi-scale temporal semantic coding feature map. In particular, in an embodiment of the present invention, the boiler combustion condition multi-scale feature extractor includes a first convolutional neural network model and a second convolutional neural network model in parallel, and the first convolutional neural network model and the second convolutional neural network model have different convolution kernel sizes. It should be understood that considering that there are rich features of boiler combustion conditions extracted at different time scales and spatial scales in the boiler combustion condition time domain aggregation matrix. Therefore, in order to capture and mine the different levels of combustion condition temporal feature information displayed between different time scales, in the technical solution of the present invention, the boiler combustion condition time domain aggregation matrix is ​​input into the boiler combustion condition multi-scale feature extractor to obtain a boiler combustion condition multi-scale temporal semantic coding feature map. It should be understood that the boiler combustion condition multi-scale feature extractor uses the first convolutional neural network model and the second convolutional neural network model with different convolution kernel sizes to capture and refine. Specifically, a small convolution kernel (such as 3x3 or 5x5) can capture local, fine-grained features, such as temperature fluctuations in a short period of time, small changes in the damper opening, etc. A large convolution kernel (such as 7x7 or 9x9) can capture global, coarse-grained features, such as combustion trends over a long period of time, changes in the overall combustion state, etc. In this way, through multi-scale feature extraction, more comprehensive combustion condition time series feature information can be provided, enhancing the model's ability to comprehensively understand boiler combustion conditions.

[0063] In an embodiment of the present invention, the module 140 for acquiring the quality data of the coal entering the furnace is used to acquire the real-time quality data of the coal entering the furnace. Accordingly, the real-time quality data of the coal entering the furnace refers to the data on the various physical and chemical properties of the coal entering the furnace collected in real time during the operation of the boiler. For example, volatile matter: the percentage of gas and light liquid released by the coal during the heating process. The level of volatile matter directly affects the ignition and combustion characteristics of the coal. Ash: the inorganic residue remaining after the coal is burned, which affects the thermal efficiency of the boiler and the wear of the equipment. Moisture: the moisture content in the coal affects the calorific value and combustion efficiency of the coal. Sulfur: the sulfur content in the coal directly affects the generation of sulfur dioxide during the combustion process and has an important impact on environmental protection requirements. Therefore, in order to adjust the mixed combustion ratio of coal more accurately and in real time to improve the combustion efficiency, in the technical solution of the present invention, real-time quality data of the coal entering the furnace is acquired.

[0064] In coal-fired power plants, it is necessary to install various types of sensors and detection equipment to monitor and collect various physical and chemical properties of coal entering the furnace in real time. These data include but are not limited to volatile matter, ash content, moisture, sulfur content and calorific value. For example, volatile matter refers to the percentage of gas and light liquid released by coal during the heating process. The level of volatile matter directly affects the ignition and combustion characteristics of coal; ash refers to the inorganic residue remaining after coal combustion, which affects the thermal efficiency of the boiler and the wear of equipment; moisture refers to the moisture content in coal, which affects the calorific value and combustion efficiency of coal; sulfur refers to the sulfur content in coal, which directly affects the generation of sulfur dioxide during the combustion process and has an important impact on environmental protection requirements; calorific value refers to the heat released when a unit mass of coal is completely burned, which is an important indicator for evaluating the calorific value of coal.

[0065] During the data collection process, it is necessary to ensure that the installation positions of sensors and detection equipment are reasonable and can fully and accurately reflect the characteristics of the coal entering the furnace. For example, online detection equipment can be installed on the coal conveyor belt to monitor the physical and chemical properties of coal in real time; sensors can be installed in coal bunkers and coal pipelines to monitor changes in coal storage and transportation. These devices need to have high precision and high reliability to ensure that the collected data is accurate. For example, volatile matter sensors need to be calibrated regularly to ensure the accuracy of their measurement results; moisture sensors need to be dustproof and waterproof to prevent environmental factors from affecting the measurement results. Through this process, it can ensure that the real-time coal quality data collected entering the furnace is accurate and complete, providing a solid data foundation for subsequent intelligent coal blending and combustion optimization.

[0066] In the embodiment of the present invention, the coal quality embedding coding module 150 is used to perform structured coding on the real-time coal quality data to obtain the real-time coal quality data embedding coding vector. Specifically, in the embodiment of the present invention, the coal quality embedding coding module is used to: use the coal quality data embedding coding matrix to perform structured coding on the real-time coal quality data to obtain the real-time coal quality data embedding coding vector. It should be understood that, considering that the real-time coal quality data contains key parameter information about coal quality, and in order to convert the original, unstructured coal quality data (such as volatile matter, ash, moisture, etc.) into a structured, standardized vector form, in the technical solution of the present invention, the coal quality data embedding coding matrix is ​​used to perform structured coding on the real-time coal quality data to obtain the real-time coal quality data embedding coding vector, so that the various features of the coal quality data can be effectively represented, and the features that are important to the combustion process can be extracted, while capturing the implicit features in the coal quality data, and enhancing the model's ability to understand the coal quality data.

[0067] In particular, in a specific embodiment of the present invention, the structured encoding of the real-time coal quality data entering the furnace using the coal quality data embedding coding matrix can be achieved by the following steps: 1. Preprocessing the real-time coal quality data entering the furnace to ensure that the format and quality of the data meet the requirements of subsequent processing. Data cleaning includes removing outliers and missing values ​​in the data. For example, interpolation methods can be used to fill missing values, or data records containing a large number of missing values ​​can be deleted. Data standardization is to normalize the data to the same dimension to facilitate subsequent feature extraction and encoding. Common methods include minimum-maximum normalization and Z-Score normalization. 2. Extract useful features from the preprocessed data, which will be used for subsequent embedded coding. Feature selection is to select features related to coal quality according to actual application scenarios. Common coal quality features include ash, volatile matter, fixed carbon, sulfur, calorific value, etc. Feature conversion is to convert non-numerical features (such as coal type categories) into numerical features. For example, unique hot encoding can be used to convert category features into binary vectors. 3. Then, construct an embedding coding matrix for the coal quality data entering the furnace, which is used to map the coal quality features to a low-dimensional embedding space. The embedding coding matrix for the coal quality data entering the furnace is usually obtained by training a neural network or other machine learning model. First, randomly initialize an embedding coding matrix E for the coal quality data entering the furnace, whose size is d×k, where d is the dimension of the feature and k is the dimension of the embedding space. Then, use historical coal quality data to train a neural network model, the input of which is the coal quality feature and the output is some target variable, such as combustion efficiency, emissions, etc. During the training process, the embedding coding matrix E for the coal quality data entering the furnace will be optimized according to the loss function to minimize the prediction error. Common training methods include gradient descent and stochastic gradient descent. 4. Use the trained embedding coding matrix for the coal quality data entering the furnace to perform structured encoding on the real-time coal quality data entering the furnace to obtain an embedded coding vector. The real-time coal quality data entering the furnace is converted into a real-time coal quality data feature vector X, whose dimension is d. For example, assuming that five features of ash, volatile matter, fixed carbon, sulfur, and calorific value are selected, the real-time coal quality data feature vector X = [x1, x2, x3, x4, x5], where x1, x2, x3, x4, and x5 are ash, volatile matter, fixed carbon, sulfur, and calorific value, respectively. Encoding is performed through the coal quality data embedding coding matrix E, that is, each row of the coal quality data embedding coding matrix E is dot-product-added with the real-time coal quality data feature vector X to obtain the real-time coal quality data embedding coding vector v. The calculation formula is v = E·X, where e is the coal quality data embedding coding matrix, X is the real-time coal quality data feature vector, and v is the real-time coal quality data embedding coding vector.

[0068] In an embodiment of the present invention, the coal quality combustion condition response coding module 160 is used to perform semantic response coding of the real-time coal quality data entering the furnace embedded coding vector and the boiler combustion condition multi-scale temporal semantic coding feature map for modal interactive prompting to obtain the coal quality entering the furnace-combustion condition semantic response coding feature map. Furthermore, considering that the real-time coal quality data entering the furnace embedded coding vector contains a structured representation of various physical and chemical properties of coal, and captures the complex relationship between coal quality characteristics, it provides dynamic state information of coal quality. The boiler combustion condition multi-scale temporal semantic coding feature map contains a multi-scale feature representation of the changing trends of various combustion conditions over time during boiler operation. These two types of data information belong to different types, but there is an intrinsic connection in time series and a dynamic change impact association between the two. Therefore, in order to capture and integrate the complex relationship between coal quality data and combustion conditions, so as to more accurately recommend the optimal coal mixing ratio and provide detailed data support for optimizing the combustion process, in the technical solution of the present invention, the real-time coal quality data entering the furnace is embedded in the coding vector and the multi-scale temporal semantic coding feature map of the boiler combustion conditions is semantically coded with modal interactive prompts to obtain the semantic response coding feature map of coal quality entering the furnace and combustion conditions.

[0069] Figure 3 : is a block diagram of a coal quality combustion condition response coding module in a coal blending and combustion optimization system based on real-time coal quality data according to an embodiment of the present invention. Figure 3 As shown, the coal quality combustion condition response coding module 160 includes: a combustion condition feature local decomposition unit 161, which is used to perform local feature decomposition along the channel dimension on the boiler combustion condition multi-scale temporal semantic coding feature map to obtain a set of boiler combustion condition multi-scale temporal semantic local feature matrices; a coal quality combustion condition local query prompt coding unit 162, which is used to perform local query prompt coding on the real-time coal quality data embedded coding vector and the set of boiler combustion condition multi-scale temporal semantic local feature matrix to obtain a set of coal quality-combustion condition local query prompt semantic coding vectors; a coal quality combustion condition mask unit 163, which is used to convert the real-time coal quality data embedded coding vector into the local query prompt coding vector. The coal quality data embedded coding vector and each coal quality-combustion condition local query prompt semantic coding vector in the set of coal quality-combustion condition local query prompt semantic coding vectors are input into a cross-modal mask weaving network based on prompt information to obtain a set of coal quality-combustion condition local feature mask weight matrices based on prompt information; the coal quality and combustion condition response aggregation unit 164 is used to perform significant semantic response aggregation on the set of multi-scale temporal semantic local feature matrices of boiler combustion conditions based on the set of coal quality-combustion condition local feature mask weight matrices based on prompt information to obtain the coal quality-combustion condition semantic response coding feature map.

[0070] The processing process of the combustion condition characteristic local decomposition unit 161 can be expressed by the formula:

[0071] Decompose(F2)={M1,M2,...,M i ,...,M n}

[0072] Among them, F2 is the multi-scale temporal semantic coding feature map of the boiler combustion condition, Decompose(F2) is the local feature decomposition operation of F2, M1, M2, M i and M n They are the first, second, i-th and n-th boiler combustion condition multi-scale temporal semantic local feature matrices in the set of the boiler combustion condition multi-scale temporal semantic local feature matrices. That is, through local feature decomposition, the local changes of boiler combustion conditions can be captured more finely, which can help the model identify and understand subtle dynamic changes in the combustion process, such as temperature fluctuations, pressure changes and fuel mixing. This not only enhances the model's ability to understand complex combustion processes, but also lays the foundation for subsequent cross-modal interaction.

[0073] Specifically, in an embodiment of the present invention, the coal quality and combustion condition local query prompt encoding unit 162 is used to: use the real-time entering coal quality data embedded coding vector as the query vector and each boiler combustion condition multi-scale temporal semantic local feature matrix in the set of boiler combustion condition multi-scale temporal semantic local feature matrices as the key matrix, and input the query vector and the key matrix into a prompt learning network based on a converter structure to obtain the set of the entering coal quality-combustion condition local query prompt semantic coding vectors.

[0074] The processing process of the above-mentioned coal quality combustion condition local query prompt coding unit 162 can be expressed by the formula:

[0075] M i = {v i1 ,v i2 ,...,v ij ,...,v im}

[0076]

[0077] Among them, v i1 、v i2 、v ij and v im M i The first, second, jth and mth row vectors in , Transformer(·,·) is the Transformer code, v1 is the real-time coal quality data embedding code vector, vij T Yes ij The transposed vector of is the matrix multiplication, ‖·‖ is the bi-norm of the vector, s j For v1 and v ij The similarity score between them, exp(·) represents the exponential function value with the natural constant e as the base, v ik It is M i The kth row vector in s k For v1 and v ik The similarity score between j v ij The corresponding coal quality-combustion condition weight coefficient, m is M i The number of row vectors in v ti It is M i The corresponding coal quality-combustion condition local query prompt semantic encoding vector. In particular, the key to the converter architecture lies in its self-attention mechanism, which enables the model to adaptively identify important features in the coal quality data and boiler combustion conditions, and assign different attention weights accordingly. The construction of the prompt learning network is to deepen the model's understanding of the interaction between different information modalities, help understand the dynamic changes of the combustion process, and ensure that these interactions can be fully utilized in subsequent analysis and processing.

[0078] The processing process of the above-mentioned coal quality combustion condition mask unit 163 can be expressed by the formula:

[0079]

[0080] Among them, v1 is the real-time coal quality data embedding coding vector, v ti T v ti The transposed vector of ti T The scale of the coal quality-combustion condition correlation matrix obtained by multiplying it with v1 is the length of the matrix multiplied by the width of the matrix. softmax(·) is the softmax function, S ti It is M i The corresponding local feature mask weight matrix of coal quality-combustion conditions based on prompt information. In this way, by generating the local feature mask weight matrix, different features can be selected and weighted, focusing on the features that have a greater impact on the combustion process, suppressing unimportant features, and improving the prediction accuracy of the model.

[0081] Specifically, in an embodiment of the present invention, the coal quality entering the furnace and combustion condition response aggregation unit 164 is used to: calculate the set of the set of the coal quality entering the furnace and combustion condition local feature mask weight matrices based on the prompt information and the set of the boiler combustion condition multi-scale temporal semantic local feature matrices, and obtain the set of coal quality entering the furnace and combustion condition local granularity significant interaction matrices by multiplying the position points between each corresponding group of the coal quality entering the furnace and combustion condition local feature mask weight matrices based on the prompt information and the boiler combustion condition multi-scale temporal semantic local feature matrices; and perform feature aggregation on the set of the coal quality entering the furnace and combustion condition local granularity significant interaction matrices to obtain the coal quality entering the furnace and combustion condition semantic response encoding feature map.

[0082] The processing process of the above-mentioned coal quality combustion condition response aggregation unit 164 can be expressed by the formula:

[0083] F 1-2 =Concat{M1⊙S t1 ,M2⊙S t2 ,...,M i ⊙S ti ...,M n ⊙S tn}

[0084] Among them, ⊙ is the point product by position, Concat{·,·...,·} is the feature concatenation along the channel dimension, and F 1-2 It is the semantic response encoding feature map of the coal quality entering the furnace and the combustion conditions. Through the significant interaction of local granularity, the fusion of different modal data of the coal quality entering the furnace and the combustion conditions is strengthened, so that the interaction matrix finally constructed can accurately reveal the correlation and importance between the modal features of the two modalities. Finally, the set of the local granularity significant interaction matrix of the coal quality entering the furnace and the combustion conditions is subjected to feature aggregation. In this way, various local features in the combustion process can be fully reflected to capture higher-level features, thereby providing a more comprehensive description of the coal quality entering the furnace and the combustion state, and improving the accuracy of the subsequent recommended coal mixing ratio.

[0085] In an embodiment of the present invention, the optimization module 170 is used to obtain an optimization instruction based on the semantic response coding feature map of the coal quality entering the furnace-combustion condition, and the optimization instruction includes a decoded value of the recommended coal mixing ratio. Specifically, in an embodiment of the present invention, the optimization module is used to: input the semantic response coding feature map of the coal quality entering the furnace-combustion condition into the coal blending optimization module based on the decoder to obtain the optimization instruction, and the optimization instruction includes a decoded value of the recommended coal mixing ratio. That is, the semantic response coding feature map of the coal quality entering the furnace-combustion condition obtained by interactive prompt response coding using the real-time coal quality data embedded coding vector and the multi-scale temporal semantic coding feature map of the boiler combustion condition is decoded and processed, so as to intelligently recommend the coal mixing ratio. In this way, the coal mixing ratio can be adjusted in real time and dynamically according to the characteristics of different coal qualities and in combination with the combustion conditions, ensuring that the combustion process is always in the best state, improving the thermal efficiency and combustion stability, thereby realizing real-time monitoring and intelligent control of the combustion process of the coal-fired boiler.

[0086] In particular, in a specific embodiment of the present invention, the coal quality-combustion condition semantic response encoding feature map is input into a decoder-based coal blending optimization module to obtain the optimization instruction, which includes a decoded value of a recommended coal blending ratio, which can be achieved by the following steps:

[0087] 1. Before the coal quality-combustion condition semantic response coding feature map is input into the decoder, some preprocessing work needs to be done, such as standardization, to ensure that all input data are on the same scale, which helps the model to better learn the patterns in the data. In addition, the coal quality-combustion condition semantic response coding feature map may also need to be appropriately converted to meet the input requirements of the decoder. 2. Input decoder: The decoder is a specially designed model that can extract useful information from complex inputs (such as the coal quality-combustion condition semantic response coding feature map) and convert it into output that is easy to understand and execute. In the case of the present invention, the task of the decoder is to decode the optimal coal mixing ratio from the coal quality-combustion condition semantic response coding feature map. In order to achieve this goal, the decoder will consider how to effectively capture and utilize the information in the feature map when designing. 3. Model training: Before the decoder is put into use, it needs to be trained with a large amount of historical data. These historical data include the actual coal mixing ratio and its effects under different coal qualities and combustion conditions. Through repeated iterations, the decoder learns how to predict the most suitable coal mixing ratio based on given coal quality and combustion conditions. 4. Generate optimization instructions: After the decoder training is completed, the pre-processed coal quality-combustion condition semantic response encoding feature map can be input into it. The decoder will output one or more recommended coal mixing ratios based on the learned knowledge. These recommended values ​​are the so-called "optimization instructions".

[0088] In summary, the coal blending and combustion optimization system 100 based on the real-time coal quality data entering the furnace according to the embodiment of the present invention is explained, which receives the data set of boiler combustion conditions (combustion temperature value, air door opening value, fan speed value and combustion chamber pressure value) collected by the sensor component, and obtains the real-time coal quality data entering the furnace, and uses the AI-based data processing and encoding algorithm to perform time domain aggregation and multi-scale semantic encoding on the boiler combustion conditions, and embeds the real-time coal quality data entering the furnace to perform encoding, so as to intelligently recommend the coal mixing ratio according to the prompt response interaction between the embedded features of the real-time coal quality data entering the furnace and the multi-scale temporal semantic encoding features of the boiler combustion conditions. In this way, the coal mixing ratio can be adjusted in real time and dynamically according to the characteristics of different coal qualities and in combination with the combustion conditions, ensuring that the combustion process is always in the best state, improving the thermal efficiency and combustion stability, thereby realizing real-time monitoring and intelligent control of the combustion process of the coal-fired boiler.

[0089] As described above, the coal blending and blending optimization system 100 based on real-time coal quality data entering the furnace according to an embodiment of the present invention can be implemented in various wireless terminals, such as a server having a coal blending and blending optimization algorithm based on real-time coal quality data entering the furnace. In a possible implementation, the coal blending and blending optimization system 100 based on real-time coal quality data entering the furnace according to an embodiment of the present invention can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the coal blending and blending optimization system 100 based on real-time coal quality data entering the furnace can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the coal blending and blending optimization system 100 based on real-time coal quality data entering the furnace can also be one of the many hardware modules of the wireless terminal.

[0090] Alternatively, in another example, the coal blending and combustion optimization system 100 based on real-time coal quality data entering the furnace and the wireless terminal may also be separate devices, and the coal blending and combustion optimization system 100 based on real-time coal quality data entering the furnace may be connected to the wireless terminal through a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0091] Figure 4 Flow chart of the coal blending optimization method based on real-time coal quality data according to an embodiment of the present invention. Figure 4 As shown, according to an embodiment of the present invention, the coal blending optimization method based on real-time coal quality data entering the furnace includes: S110, receiving a data set of boiler combustion conditions collected by a sensor component; S120, data sorting the data set of boiler combustion conditions to obtain a boiler combustion condition time domain aggregation matrix; S130, inputting the boiler combustion condition time domain aggregation matrix into a boiler combustion condition multi-scale feature extractor to obtain a boiler combustion condition multi-scale temporal semantic coding feature map; S140, acquiring real-time coal quality data entering the furnace; S150, structured coding the real-time coal quality data entering the furnace to obtain a real-time coal quality data entering the furnace embedded coding vector; S160, modal interactive prompting semantic response coding of the real-time coal quality data entering the furnace embedded coding vector and the boiler combustion condition multi-scale temporal semantic coding feature map to obtain a coal quality entering the furnace-combustion condition semantic response coding feature map; S170, based on the coal quality entering the furnace-combustion condition semantic response coding feature map, obtaining an optimization instruction, the optimization instruction including a decoded value of a recommended coal mixing ratio.

[0092] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned coal blending optimization method based on real-time coal quality data have been described in detail above. Figures 1 to 3 The invention has been introduced in detail in the description of the coal blending and combustion optimization system based on real-time coal quality data entering the furnace, and therefore, its repeated description will be omitted.

[0093] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A coal blending and combustion optimization system based on real-time coal quality data, characterized in that: include: A combustion condition data acquisition module, used to receive a data set of boiler combustion conditions collected by the sensor assembly; A combustion condition data sorting module, used for sorting the data set of the boiler combustion condition to obtain a boiler combustion condition time domain aggregation matrix; A combustion condition analysis module, used for inputting the boiler combustion condition time domain aggregation matrix into a boiler combustion condition multi-scale feature extractor to obtain a boiler combustion condition multi-scale temporal semantic coding feature map; The coal quality data acquisition module is used to obtain real-time coal quality data. The coal quality embedding coding module is used to perform structured coding on the real-time coal quality data to obtain the real-time coal quality data embedding coding vector; A coal quality combustion condition response coding module is used to perform semantic response coding of modal interactive prompts on the real-time coal quality data embedded coding vector and the boiler combustion condition multi-scale temporal semantic coding feature map to obtain a coal quality-combustion condition semantic response coding feature map; The optimization module is used to obtain an optimization instruction based on the coal quality-combustion condition semantic response coding feature map entering the furnace, and the optimization instruction includes a decoded value of a recommended coal mixing ratio.

2. The coal blending and combustion optimization system based on real-time coal quality data according to claim 1 is characterized in that: The boiler combustion conditions include a combustion temperature value, a damper opening value, a fan speed value and a combustion chamber pressure value.

3. The coal blending and combustion optimization system based on real-time coal quality data according to claim 2 is characterized in that: The boiler combustion condition multi-scale feature extractor includes a first convolutional neural network model and a second convolutional neural network model in parallel, and the first convolutional neural network model and the second convolutional neural network model have different convolution kernel sizes.

4. The coal blending and combustion optimization system based on real-time coal quality data according to claim 3 is characterized in that: The coal quality embedding coding module is used to: use the coal quality data embedding coding matrix to perform structured coding on the real-time coal quality data to obtain the real-time coal quality data embedding coding vector.

5. The coal blending and combustion optimization system based on real-time coal quality data according to claim 4 is characterized in that: The coal quality combustion condition response coding module includes: A combustion condition feature local decomposition unit, used for performing local feature decomposition along the channel dimension on the boiler combustion condition multi-scale temporal semantic coding feature map to obtain a set of boiler combustion condition multi-scale temporal semantic local feature matrices; A coal quality combustion condition local query prompt coding unit, used for performing local query prompt coding on the set of the real-time incoming coal quality data embedding coding vector and the boiler combustion condition multi-scale temporal semantic local feature matrix to obtain a set of incoming coal quality-combustion condition local query prompt semantic coding vectors; The coal quality and combustion condition mask unit is used to embed the real-time coal quality data into the coding vector and each coal quality and combustion condition local query prompt semantic coding vector in the set of coal quality and combustion condition local query prompt semantic coding vector into a cross-modal mask weaving network based on prompt information to obtain a set of coal quality and combustion condition local feature mask weight matrices based on prompt information; The coal quality and combustion condition response aggregation unit is used to perform significant semantic response aggregation on the set of multi-scale temporal semantic local feature matrices of the boiler combustion conditions based on the set of coal quality and combustion condition local feature mask weight matrices based on the prompt information to obtain the semantic response encoding feature map of the coal quality and combustion condition.

6. The coal blending and combustion optimization system based on real-time coal quality data according to claim 5 is characterized in that: The coal quality and combustion condition local query prompt encoding unit is used to: use the real-time entering coal quality data embedded coding vector as the query vector and each boiler combustion condition multi-scale temporal semantic local feature matrix in the set of boiler combustion condition multi-scale temporal semantic local feature matrices as the key matrix, input the query vector and the key matrix into the prompt learning network based on the converter structure to obtain the set of the entering coal quality-combustion condition local query prompt semantic coding vectors.

7. The coal blending and combustion optimization system based on real-time coal quality data according to claim 6 is characterized in that: The coal quality combustion condition response polymerization unit is used to: Calculate the set of the set of the local feature mask weight matrices of the incoming coal quality-combustion condition based on the prompt information and the set of the multi-scale temporal semantic local feature matrices of the boiler combustion conditions, and obtain the set of the local granularity significant interaction matrices of the incoming coal quality-combustion condition by multiplying the position points between each corresponding set of the local feature mask weight matrices of the incoming coal quality-combustion condition based on the prompt information and the multi-scale temporal semantic local feature matrices of the boiler combustion conditions; The set of the local granularity significant interaction matrices of the coal quality entering the furnace and the combustion conditions is feature aggregated to obtain the semantic response coding feature map of the coal quality entering the furnace and the combustion conditions.

8. The coal blending and combustion optimization system based on real-time coal quality data according to claim 7 is characterized in that: The optimization module is used to input the coal quality-combustion condition semantic response coding feature map into the decoder-based coal blending optimization module to obtain the optimization instruction, and the optimization instruction includes the decoded value of the recommended coal mixing ratio.

9. A coal blending optimization method based on real-time coal quality data entering the furnace, characterized in that: include: receiving a data set of boiler combustion conditions collected by a sensor assembly; Arranging the data set of the boiler combustion condition to obtain a boiler combustion condition time domain aggregation matrix; Inputting the boiler combustion condition time domain aggregation matrix into a boiler combustion condition multi-scale feature extractor to obtain a boiler combustion condition multi-scale temporal semantic coding feature map; Obtain real-time coal quality data entering the furnace; Performing structured coding on the real-time coal quality data entering the furnace to obtain an embedded coding vector of the real-time coal quality data entering the furnace; The real-time coal quality data embedding coding vector and the multi-scale temporal semantic coding feature map of boiler combustion conditions are subjected to semantic response coding of modal interaction prompts to obtain a semantic response coding feature map of coal quality-combustion conditions; Based on the coal quality-combustion condition semantic response coding feature map, an optimization instruction is obtained, and the optimization instruction includes a decoded value of a recommended coal mixing ratio.

10. The coal blending and combustion optimization method based on real-time coal quality data according to claim 9, characterized in that: The boiler combustion conditions include a combustion temperature value, a damper opening value, a fan speed value and a combustion chamber pressure value.

Images