Power plant fuel particle size on-line detection and automatic sorting system based on infrared spectrum

Abstract

The present application relates to the technical fields of power plant fuel particle size online detection and automatic sorting based on infrared spectrum, and provides a power plant fuel particle size online detection and automatic sorting system based on infrared spectrum, which comprises a data acquisition module for obtaining a mixed spectrum flow.A difference stripping module performs a space-time difference operation on the mixed spectrum flow, extracts absorption baseline data, and separates out preliminary decoupled scattering data.A photon verification module calculates a residual evaluation matrix using the preliminary decoupled scattering data, and outputs purified scattering data through closed-loop iterative operation.A morphology inversion module performs inversion calculation on the purified scattering data to obtain real particle size characteristic data.A scheduling module drives a sorting device to perform mechanical separation actions according to the real particle size characteristic data, and drives a pulverizing device to perform speed adjustment actions according to the absorption baseline data.The present application solves the problems of spectrum deep coupling and misjudgment caused by the nonlinear changes of moisture and ash concentration under the mixed combustion conditions of multiple coal types.

Description

Technical Field

[0001] This invention relates to the field of online detection and automatic sorting technology for power plant fuel particle size based on infrared spectroscopy, and particularly to an online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy. Background Technology

[0002] Infrared spectroscopy-based online particle size detection technology for power plant fuels employs infrared spectroscopy analysis to characterize the chemical composition and structural information of fuels by analyzing the absorption characteristics of substances to specific wavelengths of infrared light, thereby correlating the size distribution of fuel particles. Infrared spectroscopy technology utilizes the characteristic absorption spectra generated by the vibration of chemical bonds within molecules to achieve non-destructive identification of material components. Power plant fuels broadly refer to solid energy materials such as coal and biomass used in the power generation process. Particle size describes the size, shape, and uniformity of fuel particle distribution, directly affecting the physical properties of the combustion process. Online detection refers to the real-time continuous acquisition of data during the fuel transportation or processing flow in power plants, enabling measurement and monitoring without interrupting production.

[0003] Existing infrared spectroscopy-based online detection and automatic sorting technologies for power plant fuel particle size have the following technical challenges. Specifically, when the probe collects diffuse reflectance signals above the raw coal conveyor belt, the light intensity received by the spectrometer is not only modulated by Mie scattering due to the physical size of the coal particles, but also strongly attenuated by the absorption of moisture and ash chemical bond vibrations within the coal. When thermal power plants implement multi-coal blending strategies, the surface moisture and mineral impurity concentration of the fuel experience drastic and nonlinear fluctuations, causing abrupt changes in the absorption depth of the infrared characteristic bands. This results in irregular dynamic drift of the diffuse reflectance baseline, leading to the scattered light intensity signal originally used for particle size inversion being deeply masked by the attenuated signal generated by chemical absorption. For example, in the application scenario of online monitoring and dynamic screening of fuel in raw coal conveying and pulverizing systems, when a batch of fine coal with a high surface moisture content and a particle size that fully meets the furnace feeding standards flows through the optical detection area, the strong absorption of infrared light by moisture causes a sharp decrease in reflected light intensity. Existing signal processing algorithms cannot differentiate between light intensity reduction caused by chemical absorption and scattering by large particles. Consequently, they misinterpret the spectral distortion of qualified small coal powder particles with high absorption characteristics as scattering features produced by a single coarse particle. As a result, the system instructs the automatic sorting actuator to malfunction frequently, misdirecting a large amount of qualified fine powder into the crushed coal circulation loop, ultimately leading to over-grinding in the coal mill and a surge in plant power consumption. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy. This invention solves the technical problem that, under multi-coal blending conditions, the nonlinear changes in surface moisture and mineral ash concentration of coal particles cause a deep coupling between the chemical absorption effect of infrared light and the Mie scattering effect caused by physical particles. This results in an irregular dynamic drift of the infrared spectral diffuse reflectance baseline, leading the system to misjudge the spectral distortion of qualified small coal powder with high absorption characteristics as the scattering characteristics of a single coarse particle. Consequently, the automatic sorting actuator experiences high-frequency malfunctions, leading to a large amount of qualified fine powder being mistakenly guided into the crushed coal circulation loop, causing excessive grinding in the coal mill and a surge in plant power consumption.

[0005] To solve the above-mentioned technical problems, the specific contents of the present invention are as follows: The present invention provides an online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy, comprising a control device and a physical device, wherein the control device and the physical device establish a communication connection; the physical device includes a transmitting device, a receiving device, a sorting device, and a pulverizing device; The control device includes: a data acquisition module, which drives the transmitting device to emit transient thermal pulses and single linearly polarized beams toward the coal flow, controls the receiving device to acquire the mixed spectral stream, and transmits the mixed spectral stream to the differential stripping module; The differential stripping module receives the mixed spectral stream, performs spatiotemporal differential operations on the mixed spectral stream, extracts absorption baseline data and separates preliminary decoupled scattering data, transmits the absorption baseline data to the scheduling module, and transmits the preliminary decoupled scattering data to the photon verification module. The photon verification module receives the preliminary decoupled scattering data, calculates the residual evaluation matrix using the preliminary decoupled scattering data, and sends the residual evaluation matrix back to the differential stripping module to trigger closed-loop iterative calculation until the residual evaluation matrix converges. It then outputs the purified scattering data and transmits the purified scattering data unidirectionally to the topography inversion module. The morphology inversion module receives the purified scattering data, performs inversion calculations on the purified scattering data to obtain the true particle size characteristic data, and transmits the true particle size characteristic data to the scheduling module. The scheduling module receives the actual particle size characteristic data and the absorption baseline data respectively, drives the sorting equipment to perform mechanical separation action according to the actual particle size characteristic data, and drives the powder making equipment to perform speed adjustment action according to the absorption baseline data.

[0006] Furthermore, in the infrared spectroscopy-based online detection and automatic sorting system for power plant fuel particle size described in this invention, the transmitting device includes a high-frequency thermal excitation array and a collimated infrared light source; the receiving device includes an avalanche diode array, a parallel polarization filter component, and a vertical polarization filter component. The data acquisition module sends a high-frequency synchronous trigger level to the high-frequency thermal excitation array; The high-frequency thermal excitation array receives the high-frequency synchronous trigger level and releases the transient thermal pulse to the coal flow, causing the liquid water molecules in the coal flow to transform into gaseous water molecules; The data acquisition module drives the collimated infrared light source to emit a single linearly polarized beam of light to irradiate the coal flow, and the single linearly polarized beam of light generates reflected photons after irradiating the coal flow; The avalanche diode array captures the reflected photons through the parallel polarization filter and the vertical polarization filter; The data acquisition module adds an absolute arrival timestamp sequence to the returned photons and encapsulates them to generate the mixed spectral stream.

[0007] Furthermore, in the infrared spectroscopy-based online detection and automatic sorting system for power plant fuel particle size described in this invention, the differential stripping module is configured as follows: The differential stripping module reads the absolute arrival timestamp sequence embedded in the mixed spectral stream; The differential stripping module divides the mixed spectral stream into a first subset, a second subset, and a third subset according to the absolute arrival timestamp sequence. The differential stripping module performs an operation by subtracting the first subset from the second subset to calculate and derive a first-order time-domain difference matrix. The differential stripping module performs an operation by subtracting the first-order time-domain difference matrix from the third subset to generate a steady-state orthogonal polarization dataset.

[0008] Furthermore, in the infrared spectroscopy-based online detection and automatic sorting system for power plant fuel particle size described in this invention, the differential stripping module is further configured as follows: The steady-state orthogonal polarization dataset includes both vertical polarization branches and parallel polarization branches; The differential stripping module initiates a particle swarm optimization algorithm to perform a global search operation on the hyperparameter solution space of the asymmetric least squares smoothing algorithm, and calculates and outputs the optimal smoothing weights. The differential stripping module applies an asymmetric least squares smoothing algorithm with the optimal smoothing weights to the vertical polarization branch. The differential stripping module performs a fitting operation on the vertical polarization branch to extract the absorption baseline data; The differential stripping module performs a spatial dimension subtraction operation between the parallel polarization branch and the absorption baseline data to derive the preliminary decoupled scattering data.

[0009] Furthermore, in the infrared spectroscopy-based online detection and automatic sorting system for power plant fuel particle size described in this invention, the photon verification module is configured as follows: The photon verification module extracts the first flight time set and the second flight time set covered by the absolute arrival timestamp sequence. The photon verification module maps the first time-of-flight set to a Gaussian probability density function. The photon verification module maps the second time-of-flight set to an asymmetric gamma distribution probability density function. The photon verification module inputs the Gaussian probability density function and the asymmetric gamma probability density function into the spatiotemporal joint Monte Carlo photon tracing physics calculation model. The photon verification module performs integral calculations using the spatiotemporal joint Monte Carlo photon tracing physics calculation model and outputs theoretical scattering characteristic data.

[0010] Furthermore, in the infrared spectroscopy-based online detection and automatic sorting system for power plant fuel particle size described in this invention, the photon verification module is further configured to: perform matrix differentiation subtraction on the theoretical scattering characteristic data and the preliminary decoupled scattering data to calculate the residual evaluation matrix; The photon verification module extracts the element with the largest absolute value within the residual evaluation matrix; When the maximum absolute value element is determined to be greater than the preset tolerance threshold, the photon verification module will send the residual evaluation matrix back to the differential stripping module. The differential stripping module uses a built-in adaptive Kalman filter algorithm to update the state covariance matrix based on the residual evaluation matrix and calculates the Kalman gain. The differential stripping module uses the Kalman gain to perform scaling correction on the particle flight velocity of the particle swarm optimization algorithm; The differential stripping module recalculates the optimal smoothing weights and re-extracts the absorption baseline data using the modified particle swarm optimization algorithm. When the maximum absolute value element is determined to be less than or equal to the preset tolerance threshold, the photon verification module outputs the purified scattering data.

[0011] Furthermore, in the infrared spectroscopy-based online detection and automatic sorting system for power plant fuel particle size described in this invention, the morphology inversion module is configured as follows: The topography inversion module extracts the scattering phase function and the degree of polarization physical quantity from the purified scattering data. The topography inversion module imports the scattering phase function and the polarization degree physical quantity into the light scattering topography numerical analysis model, which includes geometric topology vectors. The topography inversion module performs an inner product matching operation between the geometric topology vector and the scattering phase function; The topography inversion module retrieves the second time-of-flight set and limits the aspect ratio extrema of the geometric topology vector based on the second time-of-flight set; The topography inversion module uses the constrained geometric topology vector to perform cross-sectional extinction inverse operation on the purified scattering data to calculate the equivalent spherical volume distribution probability function. The morphology inversion module performs matrix inverse solving on the equivalent spherical volume distribution probability function to calculate the true particle size characteristic data.

[0012] Furthermore, in the infrared spectroscopy-based online detection and automatic sorting system for power plant fuel particle size described in this invention, the scheduling module is configured as follows: The scheduling module receives the actual particle size characteristic data; The scheduling module performs a Boolean logic judgment operation on the actual particle size feature data and the preset particle size threshold built into the system. When the actual particle size feature data is greater than the preset particle size threshold, the scheduling module generates a separation instruction; The scheduling module sends the separation command to the sorting device; The sorting equipment performs the mechanical separation action by opening the high-pressure air valve or deflecting the drive motor according to the separation command.

[0013] Furthermore, in the infrared spectroscopy-based online detection and automatic sorting system for power plant fuel particle size described in this invention, the scheduling module is further configured as follows: The scheduling module receives the absorption baseline data; The scheduling module extracts the absorption peak height and absorption peak width features from the absorption baseline data; The scheduling module constructs a component correlation matrix based on the absorption peak height feature and the absorption peak width feature; The scheduling module retrieves the internal features of the component correlation matrix, performs calculations, and generates a duty cycle adjustment electrical signal. The scheduling module sends the duty cycle adjustment electrical signal to the pulverizing equipment; The pulverizing equipment receives the duty cycle adjustment electrical signal and executes the speed adjustment action.

[0014] Furthermore, in the infrared spectroscopy-based online detection and automatic sorting system for power plant fuel particle size described in this invention, the scheduling module is further configured as follows: The scheduling module extracts the nonlinear weight parameters within the component correlation matrix; The scheduling module inputs the nonlinear weight parameter and the real particle size characteristic data into the built-in combustion efficiency evaluation model; The scheduling module performs analytical operations on the nonlinear weight parameters and the real particle size characteristic data through the combustion efficiency evaluation model to calculate the dynamic compensation coefficient. The scheduling module uses the dynamic compensation coefficient to perform a multiplication correction operation on the preset particle size threshold to calculate the updated particle size threshold. The scheduling module replaces the preset particle size threshold with the updated particle size threshold.

[0015] Beneficial effects of this invention; This invention provides an online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy. It utilizes a transient thermal pulse to induce a transient gasification phase transition in the liquid water molecules on the surface of coal particles. Combined with reflected photons generated by a single linearly polarized beam of light and an absolute arrival timestamp sequence, this system achieves physical property stripping of the optical signal in both the temporal and spatial domains. This fundamentally eliminates the deep coupling interference between chemical absorption and Mie scattering effects on the infrared spectral signal. The differential stripping module effectively filters out dynamic drift in the absorption baseline data caused by nonlinear concentration fluctuations of moisture and mineral impurities through a first-order temporal difference matrix and an asymmetric least-squares smoothing algorithm. Furthermore, the photon verification module, with its built-in spatiotemporal joint Monte Carlo photon tracing physical calculation model, performs reverse feedback verification of the residual evaluation matrix, forcing the initially decoupled scattering data to converge towards purified scattering data that conforms to the photon propagation laws. This eliminates the distortion masking of particle size detection results caused by high humidity and high ash content conditions. The morphology inversion module introduces a light scattering morphology numerical analysis model based on T-matrix calculation theory and performs cross-sectional extinction inverse operation using constrained geometric topological vectors. This overcomes the bias in Mie scattering theory caused by irregular fuel particle morphology, and the calculated true particle size characteristic data accurately reproduces the fuel size distribution. The scheduling module, based on the accurately reproduced true particle size characteristic data and the updated particle size threshold corrected by dynamic compensation coefficients, achieves a deep closed-loop linkage between the mechanical separation action of the sorting equipment for substandard particles and the speed adjustment action of the pulverizing equipment to address fluctuations in fuel chemical composition. This avoids the problem of excessive grinding in the coal mill and a surge in plant power consumption caused by misdirecting qualified fine powder into the crushed coal circulation loop, significantly improving the automation monitoring level and operational efficiency of the thermal power plant's fuel preparation system. Attached Figure Description

[0016] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on the drawings without creative effort.

[0017] Figure 1 This is a system architecture diagram of the online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy, according to the present invention. Detailed Implementation

[0018] To make the technical solution of the present invention clearer, the present invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. The present invention provided by various embodiments will be described in detail below with reference to the accompanying drawings. To better understand the purpose of the present invention, the present invention will be described in further detail below.

[0019] Please see Figure 1 The present invention provides an online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy, comprising a control device and a physical device, wherein the control device and the physical device establish a communication connection; the physical device includes a transmitting device, a receiving device, a sorting device, and a pulverizing device; The control device includes: a data acquisition module, which drives the transmitting device to emit transient thermal pulses and single linearly polarized beams toward the coal flow, controls the receiving device to acquire the mixed spectral stream, and transmits the mixed spectral stream to the differential stripping module; The differential stripping module receives the mixed spectral stream, performs spatiotemporal differential operations on the mixed spectral stream, extracts absorption baseline data and separates preliminary decoupled scattering data, transmits the absorption baseline data to the scheduling module, and transmits the preliminary decoupled scattering data to the photon verification module. The photon verification module receives the preliminary decoupled scattering data, calculates the residual evaluation matrix using the preliminary decoupled scattering data, and sends the residual evaluation matrix back to the differential stripping module to trigger closed-loop iterative calculation until the residual evaluation matrix converges. It then outputs the purified scattering data and transmits the purified scattering data unidirectionally to the topography inversion module. The morphology inversion module receives the purified scattering data, performs inversion calculations on the purified scattering data to obtain the true particle size characteristic data, and transmits the true particle size characteristic data to the scheduling module. The scheduling module receives the actual particle size characteristic data and the absorption baseline data respectively, drives the sorting equipment to perform mechanical separation action according to the actual particle size characteristic data, and drives the powder making equipment to perform speed adjustment action according to the absorption baseline data.

[0020] An infrared spectroscopy-based online fuel particle size detection and automatic sorting system has been deployed at the raw coal conveying site of a thermal power plant. The system includes control and physical devices, which communicate with each other for bidirectional command exchange. The physical devices include transmitting and receiving equipment positioned above the coal conveyor belt, as well as sorting and pulverizing equipment deployed downstream of the system.

[0021] The data acquisition module built into the control device serves as the initiator of the entire closed-loop ecosystem. The data acquisition module drives the transmitting device to emit transient thermal pulses and a single linearly polarized beam of light towards the coal flow on the conveyor belt. The emission of the transient thermal pulses causes instantaneous vaporization of the liquid water on the surface of the coal flow within a very short time, resulting in a sudden change in the infrared spectral absorption characteristics. The synchronously irradiated single linearly polarized beam undergoes surface scattering and diffuse reflection after passing through the vaporization region. The data acquisition module controls the receiving device to synchronously capture the reflected photons modulated by the coal flow, thereby obtaining a mixed spectral stream. This mixed spectral stream contains a deep superposition of surface polarization-maintaining scattering information and deep depolarization absorption attenuation information. The data acquisition module transmits the mixed spectral stream to the differential stripping module for subsequent calculations.

[0022] After receiving the mixed spectral stream, the differential stripping module initiates a mechanism to separate the underlying physical properties. It utilizes a time-domain difference matrix to eliminate disturbances caused by the instantaneous phase transition of water molecules. Following time-domain differentiation, the module performs spatiotemporal difference operations on the mixed spectral stream, thoroughly removing the chemical absorption attenuation background deep within the spectral signal. Through these operations, the differential stripping module extracts absorption baseline data characterizing the nonlinear concentration fluctuations of water and mineral impurities, and separates preliminary decoupled scattering data freed from chemical absorption masking. The differential stripping module independently transmits the absorption baseline data to the system's end-of-line scheduling module for macroscopic energy consumption adjustment, while simultaneously transmitting the preliminary decoupled scattering data to the photon verification module for physical boundary checks.

[0023] The photon verification module receives preliminary decoupled scattering data. Built-in Monte Carlo photon tracing physics model, the module converts the time-series characteristics of photons' random walk within the micropores of coal into a probability density distribution. The module forward extrapolates the theoretical scattering characteristics, performing matrix differentiation on the preliminary decoupled scattering data and the extrapolated theoretical characteristics to calculate the residual evaluation matrix. This residual evaluation matrix reflects the degree of deviation between the data stripping process and the actual physical propagation laws. The module sends the residual evaluation matrix back as an error feedback signal to the differential stripping module, triggering a closed-loop iterative operation that forces adaptive adjustments to the weights of the underlying smoothing algorithm within the differential stripping module. This closed-loop iterative operation continues until the residual evaluation matrix converges to within the physical tolerance range. After convergence, the module outputs purified scattering data, completely free of absorption interference, and transmits this purified scattering data unidirectionally to the topography inversion module.

[0024] The topography inversion module receives the purified scattering data. The particles of the coal blended in actual thermal power plants exhibit extremely irregular geometric topologies. The topography inversion module abandons the traditional standard spherical analytical model and extracts the scattering phase function within the purified scattering data for non-spherical numerical analysis. The module combines the hysteresis time parameter of photons traveling within irregular pores with the equivalent volume distribution probability to perform a cross-sectional extinction inverse solution. After rigorous matrix inverse calculations, the topography inversion module performs inversion calculations on the purified scattering data to obtain the true particle size characteristic data. This true particle size characteristic data reconstructs the geometric size distribution of the mixed coal flow in three-dimensional space. The topography inversion module transmits this true particle size characteristic data to the scheduling module to execute the final mechanical linkage.

[0025] The scheduling module receives the actual particle size characteristic data transmitted by the morphology inversion module and the absorption baseline data transmitted across stages by the differential stripping module. The scheduling module constructs two parallel hard real-time control links. It logically compares the actual particle size characteristic data with the system's preset physical critical threshold for coal particle size entering the furnace, and drives the sorting equipment to perform mechanical separation actions according to the actual particle size characteristic data, intercepting unqualified large coal lumps from the conveyor belt coal flow. Simultaneously, the scheduling module analyzes the absorption peak height and width variables included in the absorption baseline data, constructing a dynamic correlation matrix of moisture and ash concentration within the coal flow. The scheduling module converts the matrix characteristics into frequency conversion adjustment signals, driving the pulverizing equipment to adjust its speed according to the absorption baseline data.

[0026] Under the harsh conditions of raw coal transportation at thermal power plants, the transmitting equipment integrates a high-frequency thermal excitation array and a collimated infrared light source, while the receiving equipment integrates an avalanche diode array, a parallel polarization filter, and a vertical polarization filter. The data acquisition module, acting as the underlying control core, sends a high-frequency synchronous trigger level to the high-frequency thermal excitation array. Upon receiving the high-frequency synchronous trigger level, the high-frequency thermal excitation array releases a transient thermal pulse onto the coal flow. This transient thermal pulse, carrying high-density thermal energy, directly acts on the surface of the coal, causing the liquid water molecules in the coal flow to absorb a large amount of heat energy and rapidly cross the latent heat of vaporization threshold, transforming into gaseous water molecules. These gaseous water molecules form a transient vapor layer on the surface of the coal particles. The data acquisition module synchronously drives the collimated infrared light source to emit a single linearly polarized beam that irradiates the coal flow. This single linearly polarized beam penetrates the transient vapor layer and irradiates the coal flow, resulting in complex physical collisions and generating reflected photons. The avalanche diode array, distributed at the periphery of the space, captures these reflected photons through the parallel and vertical polarization filters. The parallel polarization filter and the vertical polarization filter perform polarization state filtering on the returned photons, separating optical signals with different polarization properties. The data acquisition module appends an absolute arrival timestamp sequence to the returned photons. The absolute arrival timestamp sequence records the hysteresis path of the photons in the microscopic aperture. The data acquisition module encapsulates the signal with the absolute arrival timestamp sequence to generate a mixed spectral stream.

[0027] The differential stripping module is responsible for the data preprocessing task of eliminating phase transition interference. It reads the absolute arrival timestamp sequence embedded within the mixed spectral stream. Based on the time nodes set by the absolute arrival timestamp sequence, the module divides the mixed spectral stream into three subsets in the time dimension: the first subset corresponds to the background state before the arrival of the thermal pulse, the second subset corresponds to the instantaneous vaporization state of water, and the third subset corresponds to the steady-state state after the thermal disturbance dissipates. The module subtracts the first subset from the second subset. This subtraction extracts the independent absorption features caused by the water vaporization phase transition, deriving a first-order time-domain difference matrix. The module then subtracts the first-order time-domain difference matrix from the third subset. This subtraction process removes potential residual phase transition thermal disturbance components from the steady-state signal, and the module generates a steady-state orthogonal polarization dataset.

[0028] The steady-state orthogonal polarization dataset physically comprises both vertical and parallel polarization branches. The vertical polarization branch primarily carries deep depolarization absorption information, while the parallel polarization branch superimposes shallow polarization-preserving scattering information and some absorption attenuation information. The differential stripping module initiates a particle swarm optimization algorithm to perform a global search operation on the hyperparameter solution space of the asymmetric least squares smoothing algorithm. The global search operation finds the optimal convergence point by simulating the foraging behavior of particles in multidimensional space and calculates the optimal smoothing weights. The differential stripping module applies the asymmetric least squares smoothing algorithm with the optimal smoothing weights to the vertical polarization branch. The differential stripping module performs a fitting operation on the vertical polarization branch. The fitting operation gradually approximates the true background contour according to the set penalty constraints, extracting pure absorption baseline data. The differential stripping module performs a spatial dimension subtraction operation on the parallel polarization branch and the absorption baseline data. The spatial dimension subtraction operation forcibly removes the mixed chemical absorption background in the parallel polarization branch, and the differential stripping module derives preliminary decoupled scattering data.

[0029] To address the potential physical distortion risk of the initially decoupled scattering data, the photon verification module constructs a low-level error correction and verification mechanism. The module extracts a first set of flight timestamps and a second set of flight timestamps from the absolute arrival timestamp sequence. The first set of flight timestamps represents surface-scattered photons that return after experiencing very few collisions, while the second set represents long-range diffusely reflected photons that experience dense collisions within the pores of coal. The module maps the first set of flight timestamps to a Gaussian probability density function and the second set to an asymmetric gamma probability density function. These Gaussian and asymmetric gamma probability density functions are then input into a spatiotemporal joint Monte Carlo photon tracing physics model. This model sets boundary constraints for the random walk of photons based on the input probability density functions, simulating the multiple scattering paths of real photons in complex media. The module performs integral calculations through the spatiotemporal joint Monte Carlo photon tracing physics model, outputting unbiased theoretical scattering characteristic data.

[0030] Using theoretical scattering characteristic data as a standard reference, the photon verification module performs matrix subtraction on the theoretical scattering characteristic data and the preliminary decoupled scattering data. This matrix subtraction quantifies the deviation between the measured data and the theoretical model, calculating the residual evaluation matrix. The photon verification module extracts the element with the largest absolute value within the residual evaluation matrix. When the element with the largest absolute value is determined to be greater than a preset tolerance threshold, it indicates that the current stripping parameters have caused a deviation from the physical rules, and the photon verification module sends the residual evaluation matrix back to the differential stripping module. The differential stripping module uses a built-in adaptive Kalman filter algorithm to update the state covariance matrix based on the residual evaluation matrix. This update eliminates system observation noise during particle optimization and calculates the Kalman gain. The differential stripping module uses the Kalman gain to perform scaling correction on the particle flight velocity of the particle swarm optimization algorithm. The differential stripping module then recalculates the optimal smoothing weights and re-extracts the absorption baseline data using the corrected particle swarm optimization algorithm. When the maximum absolute value of an element is less than or equal to a preset tolerance threshold, it indicates that the signal processing result conforms to the evolution law of the underlying physical scattering, and the photon verification module outputs purified scattering data.

[0031] Faced with the measurement barriers posed by the irregular shapes of raw coal from thermal power plants, the morphology inversion module undertakes the task of geometric size reconstruction. The module extracts the scattering phase function and polarization degree physical quantities from the purified scattering data. It then imports these quantities into a light scattering morphology numerical analysis model. This model pre-establishes morphological representations using Chebyshev polynomials and includes geometric topological vectors. The module performs an inner product matching operation between these vectors and the scattering phase function. This operation searches a vast topology library for non-spherical structures that highly match the measured phase function. The module retrieves a second time-of-flight set and restricts the aspect ratio extrema of the geometric topological vectors based on this set. This restriction utilizes deep photon hysteresis to inversely eliminate interference from flat or elongated extreme shapes. Finally, the module performs a cross-sectional extinction inverse operation on the purified scattering data using the restricted geometric topological vectors. The cross-sectional extinction inverse operation eliminates the light intensity distortion caused by irregular shape, and calculates the equivalent spherical volume distribution probability function. The morphology inversion module performs matrix inverse operation on the equivalent spherical volume distribution probability function to calculate the true particle size characteristic data.

[0032] The system's endpoint transforms virtual data into macroscopic heavy machinery movements. The scheduling module receives actual particle size characteristic data. It then performs a Boolean logic operation on this data against a pre-set particle size threshold. This Boolean logic operation filters and determines the compliance of the current coal block size. When the actual particle size characteristic data exceeds the pre-set threshold, it means the current coal block size does not meet the combustion standards, and the scheduling module generates a separation command. This command is then sent to the sorting equipment. Based on the separation command, the sorting equipment opens a high-pressure gas valve or deflects a drive motor to perform mechanical separation.

[0033] The scheduling module synchronously receives the absorption baseline data. It extracts the absorption peak height and width characteristics from the baseline data. The peak height reflects the absolute concentration of moisture and mineral impurities, while the peak width characterizes the coupling strength of complex chemical bonds within the coal matrix. Based on these characteristics, the scheduling module constructs a composition correlation matrix. It then retrieves features from the composition correlation matrix and performs calculations to generate a duty cycle adjustment signal. This signal is converted into an electrical control variable for the equipment drive layer. The scheduling module then sends the duty cycle adjustment signal to the pulverizing equipment. Upon receiving the signal, the pulverizing equipment adjusts its rotational speed.

[0034] A single constant threshold cannot cope with combustion fluctuations in the furnace caused by blending multiple coal types. The scheduling module extracts nonlinear weight parameters from the component correlation matrix. These nonlinear weight parameters quantify the potential destructive effects of high humidity and high ash content on the overall combustion flame stability. The scheduling module inputs the nonlinear weight parameters and actual particle size characteristic data into the built-in combustion efficiency evaluation model. The scheduling module performs analytical operations on the nonlinear weight parameters and actual particle size characteristic data through the combustion efficiency evaluation model. The analytical operations comprehensively consider the synergistic combustion kinetic effects of water-ash concentration and physical particle size in the furnace, calculating a dynamic compensation coefficient. The scheduling module uses the dynamic compensation coefficient to perform a multiplicative correction operation on the preset particle size threshold. The multiplicative correction operation forcibly converts the static threshold into a variable that adapts to combustion fluctuations, calculating an updated particle size threshold. The scheduling module replaces the preset particle size threshold with the updated particle size threshold.

[0035] In the complex operating environment of raw coal transportation at thermal power plants, the blending of multiple coal types often causes nonlinear and drastic fluctuations in the surface moisture and mineral ash concentration of coal particles. To fundamentally sever the spectral coupling chain between moisture absorption and particle scattering, the physical device employs a transmitting device at the front end consisting of a high-frequency thermal excitation array and a collimated infrared light source, and a receiving device at the receiving end integrating an avalanche diode array, a parallel polarization filter, and a vertical polarization filter. The data acquisition module, acting as the underlying control center, sends a high-frequency synchronous trigger level to the high-frequency thermal excitation array. Upon receiving the high-frequency synchronous trigger level, the high-frequency thermal excitation array releases transient thermal pulses onto the coal flow on the conveyor belt. The high-density thermal energy carried by these transient thermal pulses directly acts on the surface of the coal, causing the liquid water molecules in the coal flow to absorb a large amount of heat energy and rapidly cross the latent heat of vaporization threshold, transforming into gaseous water molecules. These gaseous water molecules form a transient vapor layer on the surface of the coal particles, causing a sudden change in the absorption characteristics of a specific infrared band. The data acquisition module synchronously drives the collimated infrared light source to emit a single linearly polarized beam to irradiate the coal flow. After a single linearly polarized light beam penetrates a transient steam layer and irradiates a coal flow, the photons undergo complex physical collisions with the coal particles, generating reflected photons. An avalanche diode array distributed at the periphery of the space captures these reflected photons through parallel and vertical polarization filters. These filters perform polarization state filtering on the reflected photons, separating optical signals with different polarization properties. The data acquisition module appends an absolute arrival timestamp sequence to the reflected photons. This sequence records the photon's flight hysteresis within the microscopic pores. Based on this signal with the absolute arrival timestamp sequence, the data acquisition module encapsulates and generates a mixed spectral stream.

[0036] Eliminating interference caused by the vaporization phase transition of water molecules is a prerequisite for extracting pure scattering signals. The differential stripping module reads the absolute arrival timestamp sequence embedded within the mixed spectral stream. According to the time nodes set by the absolute arrival timestamp sequence, the differential stripping module divides the mixed spectral stream into a first subset, a second subset, and a third subset in the time dimension. The first subset corresponds to the background detection state before the arrival of the thermal pulse, the second subset corresponds to the disturbance state during the instantaneous vaporization of water, and the third subset corresponds to the steady-state physical detection state after the thermal disturbance has completely dissipated. The differential stripping module performs an operation by subtracting the first subset from the second subset. The subtraction operation strips out the independent chemical absorption features caused by the water vaporization phase transition, and calculates and derives the first-order time-domain difference matrix. The differential stripping module then performs an operation by subtracting the first-order time-domain difference matrix from the third subset. The process of subtracting the first-order time-domain difference matrix removes potential residual phase transition thermal disturbance components from the steady-state signal. After completing the subtraction operation, the differential stripping module generates a steady-state orthogonal polarization dataset.

[0037] The steady-state orthogonal polarization dataset physically comprises both vertical and parallel polarization branches. The vertical polarization branch primarily carries deep depolarization absorption information, while the parallel polarization branch superimposes shallow polarization-preserving scattering information and some absorption attenuation information. The differential stripping module initiates a particle swarm optimization algorithm to perform a global search operation on the hyperparameter solution space of the asymmetric least squares smoothing algorithm. The global search operation finds the globally optimal convergence point by simulating the movement of particles in multidimensional space and calculates the optimal smoothing weights. The differential stripping module applies the asymmetric least squares smoothing algorithm with the optimal smoothing weights to the vertical polarization branch. The differential stripping module performs a fitting operation on the vertical polarization branch. The fitting operation gradually approximates the true background contour according to the set penalty constraints, extracting the absorption baseline data reflecting the concentration of moisture and impurities. The differential stripping module performs a spatial dimension subtraction operation on the parallel polarization branch and the absorption baseline data. The spatial dimension subtraction operation forcibly removes the mixed chemical absorption background in the parallel polarization branch, and the differential stripping module derives preliminary decoupled scattering data through the spatial dimension subtraction operation.

[0038] To address the potential physical distortion risk of the initially decoupled scattering data, the photon verification module constructs a low-level physical error correction and verification mechanism. The module extracts a first set of flight timestamps and a second set of flight timestamps from the absolute arrival timestamp sequence. The first set of flight timestamps represents shallow surface-scattered photons that return after minimal collisions, while the second set represents deep diffuse-reflected photons that undergo dense collisions within the pores of coal. The module maps the first set of flight timestamps to a Gaussian probability density function and the second set to an asymmetric gamma probability density function. These Gaussian and asymmetric gamma probability density functions are then input into a spatiotemporal joint Monte Carlo photon tracing physics calculation model. This model sets boundary constraints for the random walk of photons based on the input probability density functions, simulating the multiple scattering paths of real photons in complex media. The module performs integral calculations through the spatiotemporal joint Monte Carlo photon tracing physics calculation model, outputting unbiased theoretical scattering characteristic data.

[0039] Theoretical scattering characteristic data serves as a standard reference to measure the physical compliance of the front-end data stripping algorithm. The photon verification module performs matrix subtraction on the theoretical scattering characteristic data and the initially decoupled scattering data. This matrix subtraction quantifies the deviation between the measured data and the theoretical model, calculating the residual evaluation matrix. The photon verification module extracts the element with the largest absolute value within the residual evaluation matrix. When the element with the largest absolute value is greater than a preset tolerance threshold, it indicates that the current stripping parameters have caused a deviation from the physical rules, and the photon verification module sends the residual evaluation matrix back to the differential stripping module. The differential stripping module uses a built-in adaptive Kalman filter algorithm to update the state covariance matrix based on the residual evaluation matrix. The update of the state covariance matrix eliminates system observation noise during particle optimization and calculates the Kalman gain. The differential stripping module uses the Kalman gain to perform scaling correction on the particle flight velocity of the particle swarm optimization algorithm. The differential stripping module uses the corrected particle swarm optimization algorithm to recalculate the optimal smoothing weights and re-extract the absorption baseline data. When the maximum absolute value of an element is less than or equal to a preset tolerance threshold, it indicates that the signal processing result conforms to the evolution law of the underlying physical scattering, and the photon verification module outputs purified scattering data that completely eliminates absorption interference.

[0040] Faced with measurement barriers posed by the irregular shapes of raw coal from thermal power plants, the morphology inversion module undertakes the task of geometric size reconstruction. The module extracts the scattering phase function and degree of polarization from the purified scattering data. It then imports these values ​​into a light scattering morphology numerical analysis model, which pre-establishes morphological representations using Chebyshev polynomials and includes geometric topological vectors. The module performs an inner product matching operation between these vectors and the scattering phase function. This inner product matching operation searches a vast topology library for non-spherical structures that highly match the measured phase function. The module retrieves a second time-of-flight set and restricts the aspect ratio extrema of the geometric topological vectors based on this set. This restriction utilizes deep photon hysteresis to inversely eliminate interference from flat or elongated extreme shapes. Finally, the module uses the restricted geometric topological vectors to perform cross-sectional extinction inverse operations on the purified scattering data. The cross-sectional extinction inverse operation eliminates the light intensity distortion caused by irregular shape, and calculates the equivalent spherical volume distribution probability function. The morphology inversion module performs matrix inverse operation on the equivalent spherical volume distribution probability function to calculate the true particle size characteristic data.

[0041] The system control layer transforms the virtual data from the underlying inversion into macroscopic heavy machinery drive commands. The scheduling module receives the actual particle size characteristic data. The scheduling module performs Boolean logic operations on the actual particle size characteristic data and the system's built-in preset particle size threshold. The Boolean logic operations perform condition filtering to determine whether the current coal block size meets the combustion standards for the furnace. When the actual particle size characteristic data is greater than the preset particle size threshold, it means the current coal block size exceeds the standard, and the scheduling module generates a separation command. The scheduling module sends the separation command to the sorting equipment. The sorting equipment, based on the separation command, opens the high-pressure air valve or deflects the drive motor to perform mechanical separation. The scheduling module simultaneously receives the absorption baseline data. The scheduling module extracts the absorption peak height and absorption peak width characteristics from the absorption baseline data. The absorption peak height reflects the absolute concentration of moisture and mineral impurities, while the absorption peak width characterizes the coupling strength of complex chemical bonds within the coal matrix. The scheduling module constructs a component correlation matrix based on the absorption peak height and absorption peak width characteristics. The scheduling module retrieves the internal features of the component correlation matrix and performs operations to generate a duty cycle adjustment electrical signal. The duty cycle adjustment signal is converted into an electrical control variable in the equipment drive layer, and the scheduling module sends the duty cycle adjustment signal to the pulverizing equipment. The pulverizing equipment receives the duty cycle adjustment signal and executes the speed adjustment action.

[0042] The co-firing of multiple coal types in thermal power plants leads to frequent fluctuations in furnace combustion kinetics, making it difficult to optimize efficiency under all operating conditions with a constant screening setpoint. The scheduling module extracts nonlinear weight parameters from the component correlation matrix. These nonlinear weight parameters quantify the potential destructive effects of high humidity and high ash content on overall combustion flame stability. The scheduling module inputs the nonlinear weight parameters and actual particle size characteristic data into a built-in combustion efficiency evaluation model. The scheduling module then performs analytical operations on the nonlinear weight parameters and actual particle size characteristic data through the combustion efficiency evaluation model. These analytical operations comprehensively consider the synergistic combustion response relationship between water-ash concentration and physical particle size within the furnace, calculating a dynamic compensation coefficient. The scheduling module uses this dynamic compensation coefficient to perform a multiplicative correction operation on the preset particle size threshold. This multiplicative correction operation forcibly converts the static setpoint into a dynamic variable adapted to current combustion fluctuations, calculating an updated particle size threshold. The scheduling module then replaces the preset particle size threshold with the updated particle size threshold.

[0043] A transient thermal pulse refers to a high-density collection of thermal energy released directionally onto the surface of a coal flow within a microsecond-level time window by a high-frequency thermal excitation array. The energy peak of the transient thermal pulse is strictly controlled within a range sufficient to break through the physical structure of the liquid water film on the surface of the coal particles, causing the liquid water molecules to rapidly absorb thermal energy and cross the latent heat of vaporization threshold to transform into gaseous water molecules. At the same time, the pulse duration is on the order of microseconds to avoid altering the internal skeletal structure of the coal and the primary mineral phase.

[0044] A single linearly polarized beam refers to directional light radiation emitted by a collimated infrared source, where the electric field vector is confined to vibrating within a single fixed plane. When a single linearly polarized beam penetrates a phase-change vapor layer and irradiates a solid coal surface, the shallow surface scattering caused by the physical morphology of the coal surface retains the initial optical polarization state. However, the deep diffuse reflection within the microscopic pores of the coal, which undergoes multiple physical collisions, completely loses its polarization characteristics. This single fixed polarization plane lays the underlying optical isolation foundation for subsequent use of polarization filter components to separate physical scattering from chemical absorption.

[0045] An absolute arrival timestamp sequence refers to a set of high-precision time stamps with picosecond-level time resolution attached to each returned photon captured by the data acquisition module from the avalanche diode array. The absolute arrival timestamp sequence accurately records the complete flight delay process of a photon from its departure from the transmitter to its return to the receiver. This flight delay process reflects the difference in physical path length between a single bounce on the surface of a coal particle and the dense collisions it undergoes within internal pores. The time stamp set provides a physical basis for the temporal separation between shallow polarization-maintaining surface scattering and deep depolarization diffuse reflection.

[0046] The first-order temporal difference matrix is ​​a purified data structure derived by performing algebraic subtraction on a subset of data from a mixed spectral stream at a specific time node. The first-order temporal difference matrix extracts the difference between the spectral features of the instantaneously vaporized water state and the spectral features of the background state before pulse arrival. The elements within the first-order temporal difference matrix purely characterize the phase-specific infrared chemical absorption transient features induced by the instantaneous vaporization of surface liquid water. The algebraic subtraction operation completely filters out static background interference from coal entities that have not undergone phase change.

[0047] The steady-state orthogonal polarization dataset is a physical collection of spectral information formed after eliminating transient thermal disturbances and moisture vaporization phase transition interference. Internally, the steady-state orthogonal polarization dataset is rigorously divided into vertical polarization branches and parallel polarization branches according to polarization attributes. The vertical polarization branch specifically filters out surface-scattered photons with polarization-preserving properties, thus retaining only deep depolarization absorption information. The parallel polarization branch retains surface scattering information with physical morphological characteristics. The steady-state orthogonal polarization dataset provides a pure reference source for subsequent signal decoupling and stripping in the spatial dimension.

[0048] The absorption baseline data is the underlying substrate profile variable extracted by the differential stripping module using an asymmetric least squares smoothing algorithm to perform fitting approximation operations on the vertical polarization branch. The absorption baseline data reflects the abrupt changes in absorption depth in the infrared characteristic bands caused by the drastic and nonlinear fluctuations in fuel surface moisture and mineral impurity concentration under multi-coal blending conditions. The fluctuation amplitude of the absorption baseline data quantifies the degree of irregular dynamic drift of the spectral diffuse reflectance baseline caused by chemical absorption effects.

[0049] The preliminary decoupled scattering data is the purified physical characteristic parameter output after subtracting the spatial dimension of the parallel polarization branch from the absorption baseline data. The subtraction of the parallel polarization branch from the absorption baseline data forcibly removes the background attenuation caused by deep chemical absorption. The preliminary decoupled scattering data is free from the physical absorption masking effect caused by high humidity and high ash content. The preliminary decoupled scattering data shows a Mie scattering intensity distribution law controlled by physical grain size.

[0050] The residual evaluation matrix is ​​a dynamic quantitative mathematical indicator used to measure the physical compliance of the underlying data stripping algorithm. It is generated by subtracting the matrix derivative of theoretical scattering characteristic data from the initial decoupled scattering data. The absolute values ​​of the elements within the residual evaluation matrix directly map the degree of deviation between the measured decoupled data and the actual physical propagation laws deduced by the Monte Carlo photon tracing physical calculation model. The residual evaluation matrix constitutes the core error feedback source driving the closed-loop feedback correction mechanism and the Kalman gain calculation.

[0051] The purified scattering data is the optical characteristic parameter output unidirectionally after the maximum absolute value element in the residual evaluation matrix has completely converged to within the preset tolerance threshold range. The purified scattering data has undergone multiple data purification processes, including transient thermal perturbation temporal differential stripping, polarization spatial dimension decoupling, and spatiotemporal joint Monte Carlo photon tracing reverse verification. The purified scattering data completely eliminates the deep coupling interference between chemical absorption effects and physical Mie scattering effects, presenting pure physical optical morphological information free from environmental noise.

[0052] Geometric topological vectors are mathematical arrays built upon Chebyshev polynomials within the numerical analysis model of light scattering morphology, used to characterize the three-dimensional morphology of non-spherical particles. Geometric topological vectors overcome the spatial dimensional limitations of traditional standard spherical analytical models. They constrain aspect ratio extrema based on hysteresis parameters included in the second flight time set of real photons penetrating irregular pores in coal. Geometric topological vectors possess a three-dimensional spatial mapping capability that highly matches the actual physical morphology of complex mixed coal flows.

[0053] The true particle size characteristic data is the final size measurement result output by the morphology inversion module after performing cross-sectional extinction inverse operation on the purified scattering data using constrained geometric topological vectors and completing matrix inverse solution. The true particle size characteristic data objectively recreates the probability function of the equivalent spherical volume distribution of the mixed coal flow in three-dimensional physical space. It quantifies the true size distribution of fuel particles unaffected by moisture and ash content, providing a hard real-time basis for the scheduling module to generate separation commands and drive the sorting equipment's actions.

[0054] The composition correlation matrix is ​​a two-dimensional digital structure representing the chemical properties of coal flow, constructed by the scheduling module based on the absorption peak height and width characteristics of the absorption baseline data. The absorption peak height characteristics within the composition correlation matrix directly correlate with the absolute concentrations of moisture and mineral impurities, while the absorption peak width characteristics map the physical coupling strength of complex chemical bonds within the coal matrix. The composition correlation matrix enables dynamic quantitative modeling of the intrinsic chemical quality of fuel under multi-coal blending conditions.

[0055] The dynamic compensation coefficient is an adaptive adjustment multiplier derived from the combustion efficiency assessment model by performing analytical calculations on the nonlinear weight parameters and the actual particle size characteristic data. The dynamic compensation coefficient comprehensively considers the synergistic combustion response relationship between water and ash concentration and physical particle size in the furnace of a thermal power plant. The dynamic compensation coefficient changes its value in real time with fluctuations in furnace heat load and coal blending ratio, and directly participates in breaking the fixed mechanical screening size setting limit.

[0056] The updated particle size threshold is a new generation of critical discrimination boundary conditions generated by the scheduling module after performing a multiplicative correction operation on the system's built-in preset particle size threshold using dynamic compensation coefficients. The updated particle size threshold forcibly transforms the original static mechanical interception standard into a dynamic variable that adapts to the current combustion flame fluctuations. The updated particle size threshold guides the sorting equipment to perform mechanical separation actions that conform to the efficiency optimization logic under different coal quality and chemical conditions.

[0057] The differential stripping module receives the mixed spectral stream and reads the embedded absolute arrival timestamp sequence. Based on this sequence, it divides the stream into a first subset, a second subset, and a third subset along the time dimension. The module then subtracts the first subset from the second subset to perform an algebraic subtraction operation, extracting the independent absorption features caused by the water vaporization phase transition and deriving the first-order time-domain difference matrix. The specific calculation formula is as follows: In the formula, Represents the first-order time-domain difference matrix. Represents the second subset. It represents the first subset.

[0058] After calculating and deriving the first-order time-domain difference matrix, the difference stripping module subtracts the first-order time-domain difference matrix from the third subset to remove potential residual phase transition thermal disturbance components from the steady-state signal, generating a steady-state orthogonal polarization dataset. The specific calculation formula is as follows: In the formula, Represents a steady-state orthogonal polarization dataset. Represents the third subset. This represents a first-order time-domain difference matrix.

[0059] The steady-state orthogonal polarization dataset includes both vertical and parallel polarization branches. The differential stripping module initiates a particle swarm optimization algorithm to perform a global search operation on the hyperparameter solution space of the asymmetric least squares smoothing algorithm, finding the globally optimal convergence point and calculating the optimal smoothing weights. The differential stripping module then applies the asymmetric least squares smoothing algorithm with the optimal smoothing weights to the vertical polarization branch to perform a fitting operation, approximating the true background contour and extracting the absorption baseline data. Subsequently, the differential stripping module performs a spatial dimension subtraction operation on the parallel polarization branch and the absorption baseline data, forcibly stripping away the mixed chemical absorption background in the parallel polarization branch, deriving preliminary decoupled scattering data. The specific calculation formula is as follows: In the formula, This represents preliminary decoupled scattering data. Represents parallel polarization branches. This represents the absorbed baseline data.

[0060] The photon verification module extracts the first set of flight timestamps and the second set of flight timestamps covered by the absolute arrival timestamp sequence. To construct the boundary constraints for the photon random walk, the photon verification module maps the first set of flight timestamps to a Gaussian probability density function, calculated using the following formula: In the formula, This represents the probability density function of a Gaussian distribution. Represents standard deviation, Represents pi (π) This represents an exponential function with the natural constant as its base. Represents the first set of flight times. This represents the mean.

[0061] The photon verification module synchronously maps the second time-of-flight set to an asymmetric gamma distribution probability density function, and the specific calculation formula is as follows: In the formula, This represents the probability density function of an asymmetric gamma distribution. Represents the scale parameter. Represents shape parameters, Represents the gamma function. Represents the second set of flight times. This represents an exponential function with the natural constant as its base.

[0062] The photon verification module inputs the Gaussian probability density function and the asymmetric gamma probability density function into the spatiotemporal joint Monte Carlo photon tracing physics calculation model. The photon verification module performs integral operations on the simulated path through the spatiotemporal joint Monte Carlo photon tracing physics calculation model, outputting unbiased theoretical scattering characteristic data. The specific calculation formula is as follows: In the formula, Represents theoretical scattering characteristic data, Represents the integral operator, This represents the probability density function of a Gaussian distribution. This represents the probability density function of an asymmetric gamma distribution. Represents the phase function of scattering through micropores. This represents the time integral variable.

[0063] The photon verification module performs matrix differentiation and subtraction on the theoretical scattering characteristic data and the preliminary decoupled scattering data to quantify the degree of deviation between the measured data and the theoretical model, and calculates the residual evaluation matrix. The specific calculation formula is as follows: In the formula, Represents the residual evaluation matrix. Represents the partial derivative operator, Represents theoretical scattering characteristic data, This represents preliminary decoupled scattering data.

[0064] When the maximum absolute value element in the residual evaluation matrix is ​​determined to be greater than a preset tolerance threshold, the photon verification module sends the residual evaluation matrix back to the differential stripping module. The differential stripping module uses a built-in adaptive Kalman filter algorithm to update the state covariance matrix based on the residual evaluation matrix, eliminating system observation noise and calculating the Kalman gain. The specific calculation formula is as follows: In the formula, Represents Kalman gain, This represents the state covariance matrix of the previous period. The transpose of the observation mapping matrix. Represents the observation mapping matrix, This represents the observation noise covariance matrix of the system.

[0065] The differential stripping module uses Kalman gain to perform scaling correction on the particle flight velocity of the particle swarm optimization algorithm and updates the state covariance matrix synchronously. The specific calculation formula is as follows: In the formula, This represents the updated state covariance matrix. Represents the identity matrix. Represents Kalman gain, Represents the observation mapping matrix, This represents the state covariance matrix of the previous period.

[0066] The topography inversion module receives purified scattering data, extracts the scattering phase function and degree of polarization physical quantity from the purified scattering data, and imports the scattering phase function and degree of polarization physical quantity into the light scattering topography numerical analysis model. The topography inversion module retrieves the second time-of-flight set and, based on the second time-of-flight set, restricts the aspect ratio extrema of the geometric topology vectors included in the light scattering topography numerical analysis model. Subsequently, the topography inversion module performs an inner product matching operation on the restricted geometric topology vectors and the scattering phase function; the specific calculation formula is as follows: In the formula, This represents the result of the inner product matching operation. Represents a geometric topological vector. This represents the scattering phase function.

[0067] The topography inversion module uses the constrained geometric topology vectors to perform cross-sectional extinction inverse operation on the purified scattering data, eliminating the light intensity distortion caused by irregular shape, and calculating the equivalent spherical volume distribution probability function. The specific calculation formula is as follows: In the formula, The probability function representing the equivalent spherical volume distribution. Represents the integral operator, Represents the lower limit of particle size integration. Represents the upper limit of particle size integration. The cross-sectional extinction integral kernel matrix, Represents true particle size characteristic data. This represents the integral variable of the actual particle size.

[0068] The morphology inversion module performs matrix inverse operation on the probability function of the equivalent spherical volume distribution to calculate the true particle size characteristic data that objectively restores the geometric size distribution law of the coal flow. The specific calculation formula is as follows: In the formula, Represents true particle size characteristic data. The transpose of the cross-sectional extinction integral kernel matrix. The cross-sectional extinction integral kernel matrix, Represents the smoothing regularization parameter. Represents the Laplacian operator matrix. This represents the probability function of the equivalent spherical volume distribution.

[0069] The scheduling module extracts the nonlinear weight parameters within the component correlation matrix and inputs these parameters, along with actual particle size characteristic data, into the built-in combustion efficiency evaluation model. The scheduling module then performs analytical calculations on the nonlinear weight parameters and actual particle size characteristic data using the combustion efficiency evaluation model, comprehensively considering the synergistic response relationship between water-ash concentration and physical particle size, and calculates the dynamic compensation coefficient. The specific calculation formula is as follows: In the formula, Represents the dynamic compensation coefficient. This represents an exponential function with the natural constant as its base. Represents the combustion kinetics decay constant. Represents a non-linear weighting parameter. This represents a logarithmic function with the natural constant as its base. This represents the true particle size characteristic data.

[0070] The scheduling module uses a dynamic compensation coefficient to perform a multiplicative correction operation on the system's built-in preset particle size threshold, forcibly converting the static setpoint into a dynamic variable that adapts to the current combustion fluctuations, and calculating the updated particle size threshold. The specific calculation formula is as follows: In the formula, This represents the updated particle size threshold. Represents the dynamic compensation coefficient. This represents the preset particle size threshold.

[0071] Embodiment 1 of the present invention: Under the adverse conditions of a thermal power plant experiencing continuous rainy seasons, resulting in extremely high levels of free moisture adhering to the surface of raw coal entering the furnace, the liquid water film on the surface of the coal particles will induce strong infrared photochemical absorption attenuation. The data acquisition module in the control device sends a high-frequency synchronous trigger level to a physical device arranged above the coal conveyor belt. The high-frequency thermal excitation array integrated within the transmitting device of the physical device receives the high-frequency synchronous trigger level and instantaneously releases a transient thermal pulse to the coal flow below. The high-energy pulse energy breaks down the physical structure of the water film, causing the liquid water molecules in the coal flow to cross the latent heat of vaporization limit and transform into gaseous water molecules within microseconds. The data acquisition module synchronously drives the collimated infrared light source to emit a single linearly polarized beam of light to irradiate the coal flow. The single linearly polarized beam of light penetrates the gaseous water molecule layer and strikes the coal's solid skeleton, generating a large number of reflected photons. The avalanche diode array inside the receiving device captures the reflected photons through parallel polarization filter components and vertical polarization filter components. The data acquisition module appends an absolute arrival timestamp sequence to the returned photon hardware, encapsulates it along the time dimension to generate a mixed spectral stream, and transmits it to the differential stripping module. The differential stripping module reads the absolute arrival timestamp sequence embedded in the mixed spectral stream and rigorously divides it into a first subset, a second subset, and a third subset according to the physical boundary of the thermal pulse. The differential stripping module performs an operation by subtracting the first subset, representing the background state, from the second subset, which reflects the absorption peak at the moment of vaporization, to derive a first-order time-domain difference matrix. The differential stripping module then performs an operation by subtracting the first-order time-domain difference matrix from the third subset, representing the state after pulse dissipation, to remove residual thermal perturbations through algebraic subtraction, generating a steady-state orthogonal polarization dataset including both vertical and parallel polarization branches. The differential stripping module initiates a particle swarm optimization algorithm to perform a global search operation on the hyperparameter solution space of the asymmetric least squares smoothing algorithm. The operation outputs the optimal smoothing weights, and the differential stripping module applies the asymmetric least squares smoothing algorithm with the optimal smoothing weights to the vertical polarization branch dominated by pure chemical absorption. The differential stripping module performs a fitting operation on the vertical polarization branch to extract the absorption baseline data characterizing the moisture concentration. The differential stripping module then performs a spatial dimension subtraction operation between the parallel polarization branch and the absorption baseline data to eliminate the absorption background masking effect caused by high humidity, deriving preliminary decoupled scattering data.

[0072] Embodiment 2 of this invention: The blending of high-ash, low-quality coal often results in a nonlinear increase in the porosity of coal particles. Irregular internal micropores can cause severe photon tracing phenomena. The photon verification module extracts the first and second time-of-flight sets encompassed by the absolute arrival timestamp sequence. The module maps the first time-of-flight set, representing shallow rapid return, to a Gaussian probability density function, and the second time-of-flight set, representing diffuse reflection from deep, complex pores, to an asymmetric gamma probability density function. The module inputs the Gaussian and asymmetric gamma probability density functions into a spatiotemporal joint Monte Carlo photon tracing physics model. The model injects one million virtual photons into a three-dimensional virtual mesh for physical path tracing. The module performs integration operations through the spatiotemporal joint Monte Carlo photon tracing physics model, outputting theoretical scattering characteristic data without physical bias. The module performs matrix differentiation and subtraction on the theoretical scattering characteristic data and the initially decoupled scattering data to calculate the residual evaluation matrix. The module extracts the element with the largest absolute value within the residual evaluation matrix. When the maximum absolute value element is greater than the preset tolerance threshold, it indicates that the initial stripping parameters incorrectly filtered out useful high-frequency scattering components. The photon verification module sends the residual evaluation matrix back to the differential stripping module. The differential stripping module uses a built-in adaptive Kalman filter algorithm to update the state covariance matrix based on the residual evaluation matrix, suppresses environmental optical noise interference, and calculates the Kalman gain. The differential stripping module uses the Kalman gain to perform scaling correction on the particle flight velocity of the particle swarm optimization algorithm. The differential stripping module uses the corrected particle swarm optimization algorithm to recalculate the optimal smoothing weights and re-extract the absorption baseline data. After multiple rounds of high-frequency iterations, when the maximum absolute value element is less than or equal to the preset tolerance threshold, the photon verification module outputs cleaned scattering data. The topography inversion module extracts the scattering phase function and polarization degree physical quantity from the cleaned scattering data and imports the scattering phase function and polarization degree physical quantity into the light scattering topography numerical analysis model. The light scattering topography numerical analysis model includes a geometric topology vector characterizing non-spherical distortion. The topography inversion module performs an inner product matching operation between the geometric topology vector and the scattering phase function. To prevent overfitting in matrix operations, the topography inversion module retrieves a second time-of-flight set and, based on the long-range temporal characteristics included in this set, restricts the aspect ratio extrema of the geometric topology vector. Using the constrained geometric topology vector, the module performs a cross-sectional extinction inverse operation on the purified scattering data, converting the diffuse reflection intensity distortion caused by porosity into an equivalent spherical volume distribution probability function through integration. The module then performs matrix inverse solving on the equivalent spherical volume distribution probability function to calculate the true particle size characteristic data.

[0073] Embodiment 3 of this invention: The operating energy consumption of a thermal power plant's pulverizing system is strongly coupled with the physical size and chemical properties of the coal fed into the furnace. The scheduling module receives the actual particle size characteristic data transmitted by the morphology inversion module. The scheduling module performs Boolean logic judgment operations on the actual particle size characteristic data and the system's built-in preset particle size threshold. When the actual particle size characteristic data is greater than the preset particle size threshold, it means that a large piece of gangue exceeding the grinding standard is passing through the conveyor belt. The scheduling module generates a separation command and sends the separation command to the sorting equipment. The sorting equipment opens the high-pressure air valve or drives the motor to deflect according to the separation command to perform mechanical separation actions, removing the oversized gangue from the main coal stream. For the fine coal stream that has not been removed, the scheduling module receives the absorption baseline data output by the differential stripping module across stages. The scheduling module extracts the absorption peak height and absorption peak width characteristics from the absorption baseline data. The scheduling module constructs a component correlation matrix based on the absorption peak height and absorption peak width characteristics. The scheduling module retrieves the internal features of the component correlation matrix, performs operations to generate a duty cycle adjustment electrical signal, and sends the duty cycle adjustment electrical signal to the pulverizing equipment. The pulverizing equipment receives a duty cycle adjustment signal, which directly drives the frequency converter to adjust the rotational speed. Faced with the complex operating conditions of fluctuating furnace heat load due to the blending of multiple coal types, a fixed mechanical screening size cannot meet the requirements for optimizing combustion efficiency. The scheduling module extracts the nonlinear weight parameters within the component correlation matrix. The scheduling module inputs the nonlinear weight parameters and actual particle size characteristic data into the built-in combustion efficiency evaluation model. The scheduling module performs analytical operations on the nonlinear weight parameters and actual particle size characteristic data through the combustion efficiency evaluation model to calculate the dynamic compensation coefficient. Under high moisture conditions, the dynamic compensation coefficient will adaptively decrease. The scheduling module uses the dynamic compensation coefficient to perform a multiplicative correction operation on the preset particle size threshold to calculate the updated particle size threshold. The scheduling module replaces the preset particle size threshold with the updated particle size threshold.

Claims

1. An online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy, characterized in that, It includes a control device and a physical device, wherein the control device establishes a communication connection with the physical device; the physical device includes a transmitting device, a receiving device, a sorting device, and a powder-making device; The control device includes: a data acquisition module, which drives the transmitting device to emit transient thermal pulses and single linearly polarized beams toward the coal flow, controls the receiving device to acquire the mixed spectral stream, and transmits the mixed spectral stream to the differential stripping module; The differential stripping module receives the mixed spectral stream, performs spatiotemporal differential operations on the mixed spectral stream, extracts absorption baseline data and separates preliminary decoupled scattering data, transmits the absorption baseline data to the scheduling module, and transmits the preliminary decoupled scattering data to the photon verification module. The photon verification module receives the preliminary decoupled scattering data, calculates the residual evaluation matrix using the preliminary decoupled scattering data, and sends the residual evaluation matrix back to the differential stripping module to trigger closed-loop iterative calculation until the residual evaluation matrix converges. It then outputs the purified scattering data and transmits the purified scattering data unidirectionally to the topography inversion module. The morphology inversion module receives the purified scattering data, performs inversion calculations on the purified scattering data to obtain the true particle size characteristic data, and transmits the true particle size characteristic data to the scheduling module. The scheduling module receives the actual particle size characteristic data and the absorption baseline data respectively, drives the sorting equipment to perform mechanical separation action according to the actual particle size characteristic data, and drives the powder making equipment to perform speed adjustment action according to the absorption baseline data.

2. The online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy according to claim 1, characterized in that, The transmitting device includes a high-frequency thermally excited array and a collimated infrared light source; the receiving device includes an avalanche diode array, a parallel polarization filter component, and a vertical polarization filter component. The data acquisition module sends a high-frequency synchronous trigger level to the high-frequency thermal excitation array; The high-frequency thermal excitation array receives the high-frequency synchronous trigger level and releases the transient thermal pulse to the coal flow, causing the liquid water molecules in the coal flow to transform into gaseous water molecules; The data acquisition module drives the collimated infrared light source to emit a single linearly polarized beam of light to irradiate the coal flow, and the single linearly polarized beam of light generates reflected photons after irradiating the coal flow; The avalanche diode array captures the reflected photons through the parallel polarization filter and the vertical polarization filter; The data acquisition module adds an absolute arrival timestamp sequence to the returned photons and encapsulates them to generate the mixed spectral stream.

3. The online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy according to claim 2, characterized in that, The differential stripping module is configured as follows: The differential stripping module reads the absolute arrival timestamp sequence embedded in the mixed spectral stream; The differential stripping module divides the mixed spectral stream into a first subset, a second subset, and a third subset according to the absolute arrival timestamp sequence. The differential stripping module performs an operation by subtracting the first subset from the second subset to calculate and derive a first-order time-domain difference matrix. The differential stripping module performs an operation by subtracting the first-order time-domain difference matrix from the third subset to generate a steady-state orthogonal polarization dataset.

4. The online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy according to claim 3, characterized in that, The differential stripping module is also configured to: The steady-state orthogonal polarization dataset includes both vertical polarization branches and parallel polarization branches; The differential stripping module initiates a particle swarm optimization algorithm to perform a global search operation on the hyperparameter solution space of the asymmetric least squares smoothing algorithm, and calculates and outputs the optimal smoothing weights. The differential stripping module applies an asymmetric least squares smoothing algorithm with the optimal smoothing weights to the vertical polarization branch. The differential stripping module performs a fitting operation on the vertical polarization branch to extract the absorption baseline data; The differential stripping module performs a spatial dimension subtraction operation between the parallel polarization branch and the absorption baseline data to derive the preliminary decoupled scattering data.

5. The online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy according to claim 4, characterized in that, The photon verification module is configured as follows: The photon verification module extracts the first flight time set and the second flight time set covered by the absolute arrival timestamp sequence. The photon verification module maps the first time-of-flight set to a Gaussian probability density function. The photon verification module maps the second time-of-flight set to an asymmetric gamma distribution probability density function. The photon verification module inputs the Gaussian probability density function and the asymmetric gamma probability density function into the spatiotemporal joint Monte Carlo photon tracing physics calculation model. The photon verification module performs integral calculations using the spatiotemporal joint Monte Carlo photon tracing physics calculation model and outputs theoretical scattering characteristic data.

6. The online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy according to claim 5, characterized in that, The photon verification module is further configured to: perform matrix derivative subtraction on the theoretical scattering characteristic data and the preliminary decoupled scattering data to calculate the residual evaluation matrix; The photon verification module extracts the element with the largest absolute value within the residual evaluation matrix; When the maximum absolute value element is determined to be greater than the preset tolerance threshold, the photon verification module will send the residual evaluation matrix back to the differential stripping module. The differential stripping module uses a built-in adaptive Kalman filter algorithm to update the state covariance matrix based on the residual evaluation matrix and calculates the Kalman gain. The differential stripping module uses the Kalman gain to perform scaling correction on the particle flight velocity of the particle swarm optimization algorithm; The differential stripping module recalculates the optimal smoothing weights and re-extracts the absorption baseline data using the modified particle swarm optimization algorithm. When the maximum absolute value element is determined to be less than or equal to the preset tolerance threshold, the photon verification module outputs the purified scattering data.

7. The online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy according to claim 6, characterized in that, The topography inversion module is configured as follows: The topography inversion module extracts the scattering phase function and the degree of polarization physical quantity from the purified scattering data. The topography inversion module imports the scattering phase function and the polarization degree physical quantity into the light scattering topography numerical analysis model, which includes geometric topology vectors. The topography inversion module performs an inner product matching operation between the geometric topology vector and the scattering phase function; The topography inversion module retrieves the second time-of-flight set and limits the aspect ratio extrema of the geometric topology vector based on the second time-of-flight set; The topography inversion module uses the constrained geometric topology vector to perform cross-sectional extinction inverse operation on the purified scattering data to calculate the equivalent spherical volume distribution probability function. The morphology inversion module performs matrix inverse solving on the equivalent spherical volume distribution probability function to calculate the true particle size characteristic data.

8. The online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy according to claim 7, characterized in that, The scheduling module is configured as follows: The scheduling module receives the actual particle size characteristic data; The scheduling module performs a Boolean logic judgment operation on the actual particle size feature data and the preset particle size threshold built into the system. When the actual particle size feature data is greater than the preset particle size threshold, the scheduling module generates a separation instruction; The scheduling module sends the separation command to the sorting device; The sorting equipment performs the mechanical separation action by opening the high-pressure air valve or deflecting the drive motor according to the separation command.

9. The online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy according to claim 8, characterized in that, The scheduling module is also configured to: The scheduling module receives the absorption baseline data; The scheduling module extracts the absorption peak height and absorption peak width features from the absorption baseline data; The scheduling module constructs a component correlation matrix based on the absorption peak height feature and the absorption peak width feature; The scheduling module retrieves the internal features of the component correlation matrix, performs calculations, and generates a duty cycle adjustment electrical signal. The scheduling module sends the duty cycle adjustment electrical signal to the pulverizing equipment; The pulverizing equipment receives the duty cycle adjustment electrical signal and executes the speed adjustment action.

10. The online detection and automatic sorting system for power plant fuel particle size based on infrared spectroscopy according to claim 9, characterized in that, The scheduling module is also configured to: The scheduling module extracts the nonlinear weight parameters within the component correlation matrix; The scheduling module inputs the nonlinear weight parameter and the real particle size characteristic data into the built-in combustion efficiency evaluation model; The scheduling module performs analytical operations on the nonlinear weight parameters and the real particle size characteristic data through the combustion efficiency evaluation model to calculate the dynamic compensation coefficient. The scheduling module uses the dynamic compensation coefficient to perform a multiplication correction operation on the preset particle size threshold to calculate the updated particle size threshold. The scheduling module replaces the preset particle size threshold with the updated particle size threshold.

Images