Boiler combustion state identification method and system based on flame spectrum and sound wave fusion

Abstract

This application discloses a method and system for boiler combustion state identification based on flame spectrum and acoustic wave fusion. The method includes: acquiring flame emission spectral signals from the target combustion zone of the boiler, acoustic wave signals generated during the boiler combustion process, and boiler operating parameters; constructing a spectral fluctuation sequence based on the flame emission spectral signals and an acoustic envelope sequence based on the acoustic wave signals; determining the cross-modal time delay compensation amount based on the spectral fluctuation sequence and the acoustic envelope sequence, and performing time alignment; determining the reliability of the flame spectral mode and the acoustic wave mode; extracting flame spectral features, acoustic features, and cross-modal coupling features, and performing fusion identification in conjunction with boiler operating parameters to obtain the boiler combustion state identification result. The method provided in this application can improve the comprehensiveness and anti-interference ability of boiler combustion state characterization, and enhance the accuracy, stability, and real-time performance of boiler combustion state identification.

Description

Technical Field

[0001] This application relates to the field of boiler combustion detection technology, and in particular to a method and system for identifying boiler combustion status based on the fusion of flame spectrum and sound waves. Background Technology

[0002] Boilers are widely used in industrial settings such as thermal power generation, district heating, metallurgy, and chemical processing. The stability and completeness of their combustion process directly affect the boiler's thermal efficiency, pollutant emission levels, and operational safety. Due to the high temperature, high dust levels, strong disturbances, and continuous changes inherent in the combustion process within the boiler furnace, monitoring and assessing the combustion status are crucial for timely detection of combustion anomalies, optimization of operating conditions, and maintenance of stable boiler operation. Therefore, effectively identifying the boiler's combustion status has always been a critical issue in the field of boiler control and operation monitoring.

[0003] In related technologies, monitoring the combustion status of boilers typically involves indirect judgment using operating parameters such as furnace temperature, flue gas oxygen content, furnace negative pressure, and flue gas composition, or analysis of the combustion process using information such as flame images, flame optical signals, and acoustic signals. While parameter-based monitoring is convenient for engineering implementation, it often only reflects the overall boiler operating condition and cannot promptly characterize rapid changes in the local combustion status within the furnace. Monitoring based on a single optical signal is easily affected by factors such as smoke, window contamination, and light fluctuations. Monitoring based on a single acoustic signal is easily interfered with by factors such as structural vibration, mechanical noise, and airflow disturbances. Especially under conditions of boiler load fluctuations, changes in fuel characteristics, and adjustments to air distribution, existing technologies often suffer from incomplete combustion status characterization, insufficient anti-interference capabilities, and poor stability of identification results, thus affecting the accuracy and timeliness of subsequent operational adjustments.

[0004] Therefore, in boiler combustion status identification, insufficient combustion status characterization, limited anti-interference capability of single monitoring methods, and poor accuracy and real-time performance of combustion status identification have become urgent problems to be solved. Summary of the Invention

[0005] This application provides a boiler combustion state identification method and system based on the fusion of flame spectrum and sound waves, aiming to solve the problems of insufficient combustion state characterization, limited anti-interference ability of single monitoring methods, and poor accuracy and real-time performance of combustion state identification in existing technologies for boiler combustion state identification.

[0006] Firstly, a method for identifying boiler combustion status based on the fusion of flame spectrum and acoustic waves, the method comprising: Collect flame emission spectrum signals from the target combustion zone of the boiler, acoustic signals generated during the boiler combustion process, and boiler operating parameters; A spectral wave sequence is constructed based on the flame emission spectral signal, and an acoustic envelope sequence is constructed based on the acoustic wave signal; The cross-modal time delay compensation amount is determined based on the correlation or coherence between the spectral fluctuation sequence and the acoustic envelope sequence within a preset time window; Based on the cross-modal time delay compensation, the flame emission spectrum signal and the acoustic signal are time-aligned to obtain the aligned flame emission spectrum signal and the aligned acoustic signal. Based on the aligned flame emission spectrum signal and the aligned acoustic signal, the reliability of the flame spectral mode and the reliability of the acoustic mode are determined. Based on the aligned flame emission spectrum signal and the aligned acoustic signal, flame spectral features, acoustic features and cross-modal coupling features are extracted; The boiler combustion state identification result is obtained by fusing and identifying the reliability of the flame spectral mode, the reliability of the acoustic mode, the flame spectral features, the acoustic features, the cross-modal coupling features, and the boiler operating parameters.

[0007] Optionally, in the above scheme, constructing a spectral fluctuation sequence based on the flame emission spectral signal includes: Based on the flame emission spectrum signal, spectral change parameters for characterizing the change of flame emission spectrum over time are extracted; The values ​​of the spectral variation parameters are arranged in chronological order to form the spectral fluctuation sequence; The construction of the acoustic envelope sequence based on the acoustic signal includes: Based on the sound wave signal, acoustic envelope parameters are extracted to characterize the change of sound wave amplitude over time. The values ​​of the acoustic envelope parameters are arranged in chronological order to form the acoustic envelope sequence; The step of determining the cross-modal time delay compensation amount based on the correlation or coherence between the spectral fluctuation sequence and the acoustic envelope sequence within a preset time window includes: Within the data segment corresponding to the preset time window, a relative time shift search is performed on the spectral fluctuation sequence and the acoustic envelope sequence within a preset time shift range, and the time shift amount corresponding to the maximum correlation coefficient or coherence coefficient between the spectral fluctuation sequence and the acoustic envelope sequence is determined as the cross-modal time delay compensation amount.

[0008] In the above scheme, optionally, the spectral change parameters include at least one of the following: the change in spectral intensity of the preset characteristic band between adjacent sampling times, the peak shift, and the change in the centroid of spectral energy. The step of extracting acoustic envelope parameters based on the acoustic signal to characterize the change of acoustic amplitude over time includes: The acoustic signal is subjected to bandpass filtering and frame windowing to obtain the processed acoustic signal; Envelope detection is performed on the processed acoustic signal to obtain the acoustic envelope value corresponding to each sampling time. The acoustic envelope value is determined as the acoustic envelope parameter, and / or the acoustic envelope change is determined based on the acoustic envelope value corresponding to adjacent sampling times, and the acoustic envelope change is determined as the acoustic envelope parameter.

[0009] Optionally, in the above scheme, determining the reliability of the flame spectral mode and the reliability of the acoustic mode based on the aligned flame emission spectral signal and the aligned acoustic signal includes: Based on the aligned flame emission spectrum signal, determine the flame spectral mode quality index; Based on the aligned acoustic wave signal, determine the acoustic wave modal quality index; Based on the flame spectral mode quality index, the reliability of the flame spectral mode is determined; The reliability of the acoustic mode is determined based on the acoustic mode quality index.

[0010] In the above scheme, optionally, the flame spectral modal quality index includes at least one of the following: reference spectral band attenuation rate, spectral line saturation ratio, continuous spectrum baseline drift, and spectral signal-to-noise ratio; The acoustic modal quality indicators include at least one of the following: standing wave resonance interference coefficient, mechanical noise frequency band ratio, sudden impact noise ratio, and acoustic signal-to-noise ratio.

[0011] Optionally, in the above scheme, the step of extracting flame spectral features, acoustic features, and cross-modal coupling features based on the aligned flame emission spectrum signal and the aligned acoustic wave signal includes: Based on the aligned flame emission spectrum signal, flame spectral features are extracted; Based on the aligned acoustic signal, acoustic features are extracted; Based on the correspondence between the flame spectral features and the acoustic features, cross-modal coupling features are extracted to characterize the correlation between the flame spectral features and the acoustic features.

[0012] In the above scheme, optionally, the flame spectral characteristics include at least one of the following: preset band spectral energy, band energy ratio, characteristic spectral peak intensity, characteristic spectral peak area, spectral peak drift, spectral entropy, spectral slope, and scintillation frequency characteristics. The acoustic features include at least one of the following: sound pressure level, dominant frequency, frequency band energy distribution, harmonic energy ratio, spectral centroid, spectral entropy, wavelet packet energy features, pulsation amplitude, and abnormal impact count. The cross-modal coupling characteristics include at least one of the following: the coherence between the spectral fluctuation sequence and the acoustic envelope sequence; the phase difference between the spectral fluctuation sequence and the acoustic envelope sequence after time alignment based on the cross-modal time delay compensation; the phase difference drift between the spectral fluctuation sequence and the acoustic envelope sequence after time alignment based on the cross-modal time delay compensation; the energy correlation between the flame spectral characteristics and the acoustic characteristics in a common frequency band; and the coupling strength between the changes in the flame spectral characteristics and the changes in the acoustic characteristics within a corresponding time window.

[0013] Optionally, in the above scheme, the fusion identification based on the confidence level of the flame spectral mode, the confidence level of the acoustic mode, the flame spectral features, the acoustic features, the cross-modal coupling features, and the boiler operating parameters to obtain the boiler combustion state identification result includes: Based on the credibility of the flame spectral modes, the first weight of the flame spectral features in the fusion identification is determined; Based on the credibility of the acoustic mode, the second weight of the acoustic feature in the fusion identification is determined; The flame spectral features are weighted based on the first weight to obtain the weighted flame spectral features. The acoustic features are weighted based on the second weight to obtain the weighted acoustic features. The weighted flame spectral features, the weighted acoustic features, the cross-modal coupling features, and the boiler operating parameters are fused to obtain fused identification input information; Based on the fused identification input information, the boiler combustion state identification result is determined; The boiler combustion state identification results include combustion state category, state intensity level and / or state confidence level. The combustion state category is stable combustion state, oxygen-deficient combustion state, excess air combustion state, combustion oscillation state, incomplete combustion trend state, flameout risk state or coking risk state.

[0014] Optionally, in the above scheme, the method further includes: Based on the combustion state category in the boiler combustion state identification results, and in combination with the state intensity level and / or state confidence, determine the parameter optimization strategy corresponding to the combustion state category; Based on the parameter optimization strategy, the optimization results of the boiler combustion parameters are determined; Based on the optimization results of the boiler combustion parameters, the parameters of the boiler combustion process are adjusted. The optimization results of the boiler combustion parameters include the adjustment amount of at least one of the following: fuel feed rate, primary air volume, secondary air volume, induced draft volume, and the ratio of primary air to secondary air.

[0015] Secondly, a boiler combustion state identification system based on the fusion of flame spectrum and sound waves, the system comprising: The data acquisition module is used to acquire flame emission spectrum signals from the target combustion zone of the boiler, acoustic signals generated during the boiler combustion process, and boiler operating parameters. A sequence construction module is used to construct a spectral wave sequence based on the flame emission spectral signal and an acoustic envelope sequence based on the acoustic wave signal; The time delay compensation module is used to determine the cross-modal time delay compensation amount based on the correlation or coherence between the spectral fluctuation sequence and the acoustic envelope sequence within a preset time window. The time alignment module is used to perform time alignment between the flame emission spectrum signal and the acoustic signal based on the cross-modal time delay compensation amount, so as to obtain the aligned flame emission spectrum signal and the aligned acoustic signal. A credibility determination module is used to determine the credibility of the flame spectral mode and the credibility of the acoustic mode based on the aligned flame emission spectral signal and the aligned acoustic signal. The feature extraction module is used to extract flame spectral features, acoustic features, and cross-modal coupling features based on the aligned flame emission spectrum signal and the aligned acoustic wave signal. The fusion identification module is used to perform fusion identification based on the credibility of the flame spectral mode, the credibility of the acoustic mode, the flame spectral features, the acoustic features, the cross-modal coupling features, and the boiler operating parameters to obtain the boiler combustion state identification result.

[0016] Compared with the prior art, this application has at least the following beneficial effects: Based on further analysis and research of existing technical problems, this application recognizes that existing technologies for identifying boiler combustion states suffer from insufficient combustion state characterization, limited anti-interference capabilities of single monitoring methods, and poor accuracy and real-time performance. By acquiring flame emission spectral signals from the target combustion zone of the boiler, acoustic signals generated during the boiler combustion process, and boiler operating parameters, this application expands the original method of judging boiler combustion state based solely on a single monitoring method to a multi-dimensional joint characterization based on optical, acoustic, and operating condition information. Since flame emission spectral signals reflect flame radiation and combustion reaction characteristics, acoustic signals reflect combustion pulsation and changes in combustion stability, and boiler operating parameters reflect the current operating conditions, the combination of these three elements improves the completeness of the combustion state characterization. Furthermore, this application constructs a spectral fluctuation sequence based on the flame emission spectral signal and an acoustic envelope sequence based on the acoustic signal. It then determines the cross-modal time delay compensation amount based on the correlation or coherence of the two within a preset time window, and uses the cross-modal time delay compensation amount to time-align the flame emission spectral signal and the acoustic signal, ensuring that data from different modes corresponding to the same combustion process are synchronized. Maintaining consistency in time avoids the impact of inconsistent response timing between signals from different sources on subsequent identification results. Simultaneously, after time alignment, this application separately determines the reliability of the flame spectral mode and the acoustic mode, ensuring that different modes can participate in subsequent identification based on their reliability when affected by smoke obstruction, noise interference, or operating condition fluctuations. This mitigates the adverse effects of the limited anti-interference capability of a single monitoring method on the identification results. Furthermore, this application extracts flame spectral features, acoustic features, and cross-modal coupling features, and performs fusion identification in conjunction with boiler operating parameters. Because the flame spectrum... The features and acoustic features respectively characterize the combustion state information under different modes, and the cross-modal coupling features further characterize the correlation between the two modes. Therefore, compared with the existing technology that judges based on a single signal or a single parameter, this application can more comprehensively reflect the boiler combustion state and improve the ability to identify changes in combustion state under complex operating conditions. Ultimately, it achieves the technical effect of improving the accuracy, stability and real-time performance of boiler combustion state identification, thereby solving the problems of insufficient combustion state characterization, limited anti-interference ability of single monitoring methods and poor accuracy and real-time performance of combustion state identification in the background technology. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a boiler combustion state identification method based on the fusion of flame spectrum and acoustic waves, provided as an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] In one embodiment, such as Figure 1 As shown, a method for identifying boiler combustion status based on the fusion of flame spectrum and acoustic waves is provided, including the following steps: Collect flame emission spectrum signals from the target combustion zone of the boiler, acoustic signals generated during the boiler combustion process, and boiler operating parameters; A spectral wave sequence is constructed based on the flame emission spectral signal, and an acoustic envelope sequence is constructed based on the acoustic wave signal; The cross-modal time delay compensation amount is determined based on the correlation or coherence between the spectral fluctuation sequence and the acoustic envelope sequence within a preset time window; Based on the cross-modal time delay compensation, the flame emission spectrum signal and the acoustic signal are time-aligned to obtain the aligned flame emission spectrum signal and the aligned acoustic signal. Based on the aligned flame emission spectrum signal and the aligned acoustic signal, the reliability of the flame spectral mode and the reliability of the acoustic mode are determined. Based on the aligned flame emission spectrum signal and the aligned acoustic signal, flame spectral features, acoustic features and cross-modal coupling features are extracted; The boiler combustion state identification result is obtained by fusing and identifying the reliability of the flame spectral mode, the reliability of the acoustic mode, the flame spectral features, the acoustic features, the cross-modal coupling features, and the boiler operating parameters.

[0020] In some embodiments, the flame emission spectrum signal of the target combustion zone of the boiler is acquired by a spectral acquisition device located at the boiler observation port. This spectral acquisition device can be a fiber optic spectrometer, an array-type spectrophotometer, or other device capable of acquiring the flame radiation spectrum. The acoustic signal generated during boiler combustion is acquired by an acoustic sensor located on the outside of the furnace wall, near the observation port, or in the burner area. This acoustic sensor can be a microphone, a sound pressure sensor, or a piezoelectric acoustic sensor. Boiler operating parameters are provided by the existing boiler control system, distributed control system, or field sensing devices. These parameters may include one or more of the following: boiler load, fuel feed rate, primary air volume, secondary air volume, induced draft volume, primary and secondary air ratio, furnace negative pressure, flue gas oxygen content, flue gas temperature, steam flow rate, and steam pressure.

[0021] In some embodiments, the flame emission spectrum signal of the target combustion zone of the boiler, the acoustic signal generated during the boiler combustion process, and the boiler operating parameters are first acquired. Then, a spectral fluctuation sequence is constructed based on the flame emission spectrum signal, and an acoustic envelope sequence is constructed based on the acoustic signal. Next, the cross-modal time delay compensation amount is determined based on the correlation or coherence between the spectral fluctuation sequence and the acoustic envelope sequence within a preset time window. Then, the flame emission spectrum signal and the acoustic signal are time-aligned based on the cross-modal time delay compensation amount to obtain aligned flame emission spectrum signals and aligned acoustic signals. Next, the reliability of the flame spectral mode and the acoustic mode are determined based on the aligned flame emission spectrum signal and the aligned acoustic signal. Then, flame spectral features, acoustic features, and cross-modal coupling features are extracted based on the aligned flame emission spectrum signal and the aligned acoustic signal. Finally, the reliability of the flame spectral mode, the reliability of the acoustic mode, the flame spectral features, the acoustic features, the cross-modal coupling features, and the boiler operating parameters are fused and identified to obtain the boiler combustion state identification result.

[0022] In some embodiments, flame emission spectral signals, acoustic signals, and boiler operating parameters can all be appended with a uniform timestamp for subsequent time correlation. Boiler combustion state identification results can include one or more of the following: combustion state category, state intensity level, and state confidence level.

[0023] Through the above implementation methods, it is possible to simultaneously utilize flame emission spectrum signals, acoustic signals, and boiler operating parameters during the boiler combustion state identification process. Furthermore, by combining cross-modal time delay compensation, the reliability of flame spectrum modes, the reliability of acoustic modes, and cross-modal coupling characteristics, a fusion identification can be performed. This improves the comprehensiveness and anti-interference capability of boiler combustion state characterization, and enhances the accuracy, stability, and real-time performance of boiler combustion state identification.

[0024] In this embodiment, constructing a spectral fluctuation sequence based on the flame emission spectral signal includes: Based on the flame emission spectrum signal, spectral change parameters for characterizing the change of flame emission spectrum over time are extracted; The values ​​of the spectral variation parameters are arranged in chronological order to form the spectral fluctuation sequence; The construction of the acoustic envelope sequence based on the acoustic signal includes: Based on the sound wave signal, acoustic envelope parameters are extracted to characterize the change of sound wave amplitude over time. The values ​​of the acoustic envelope parameters are arranged in chronological order to form the acoustic envelope sequence; The step of determining the cross-modal time delay compensation amount based on the correlation or coherence between the spectral fluctuation sequence and the acoustic envelope sequence within a preset time window includes: Within the data segment corresponding to the preset time window, a relative time shift search is performed on the spectral fluctuation sequence and the acoustic envelope sequence within a preset time shift range, and the time shift amount corresponding to the maximum correlation coefficient or coherence coefficient between the spectral fluctuation sequence and the acoustic envelope sequence is determined as the cross-modal time delay compensation amount.

[0025] In some embodiments, when constructing a spectral fluctuation sequence based on flame emission spectral signals, flame emission spectral data corresponding to multiple consecutive sampling times can be acquired first according to a preset sampling period. To construct the spectral fluctuation sequence, spectral variation parameters characterizing the change of flame emission spectra over time can be extracted first, and then the values ​​of the spectral variation parameters can be arranged in chronological order to form the spectral fluctuation sequence.

[0026] In some embodiments, when constructing an acoustic envelope sequence based on acoustic signals, acoustic envelope parameters characterizing the change of acoustic amplitude over time can be extracted from continuously acquired acoustic signals. Then, the values ​​of these acoustic envelope parameters are arranged in chronological order to form an acoustic envelope sequence. Here, the spectral fluctuation sequence is used to characterize the dynamic fluctuation of the flame emission spectrum over time, and the acoustic envelope sequence is used to characterize the dynamic envelope of the acoustic amplitude over time.

[0027] In some embodiments, when determining the cross-modal time delay compensation amount based on the correlation or coherence between the spectral fluctuation sequence and the acoustic envelope sequence within a preset time window, a sliding time window approach can be used. Let the current analysis window be of length [missing information]. Within the preset time window, spectral fluctuation sequences are taken respectively within the data segment corresponding to that time window. Harmony and acoustic envelope sequence and within the preset time shift range Within this context, a relative time-shift search is performed on the two.

[0028] Normalized cross-correlation coefficient It can be represented as: ;in, Indicates the first The spectral fluctuation sequence values ​​corresponding to each sampling point Indicates the time shift as The corresponding acoustic envelope sequence value at that time, This indicates the number of sampling points within the preset time window. Indicates relative time shift. Indicates the time shift as The normalized correlation between the temporal spectral wave sequence and the acoustic envelope sequence. The largest As a cross-modal time delay compensation quantity.

[0029] In other embodiments, coherence analysis can be used to determine the cross-modal time delay compensation amount. This involves calculating the coherence coefficient between the spectral wave sequence and the acoustic envelope sequence in the frequency domain, and selecting the time shift corresponding to the maximum coherence coefficient as the cross-modal time delay compensation amount. After determining the cross-modal time delay compensation amount, the flame emission spectral signal and the acoustic signal can be time-aligned accordingly. For example, one sequence can be shifted forward or backward by a corresponding time interval, or corrected based on the original timestamp, thereby obtaining the aligned flame emission spectral signal and the aligned acoustic signal.

[0030] The above implementation methods provide a basic implementation path for spectral fluctuation sequences, acoustic envelope sequences, and cross-modal time delay compensation quantities, providing a unified data and temporal foundation for subsequent reliability determination, feature extraction, and fusion identification, thereby improving the feasibility of the entire boiler combustion state identification process.

[0031] In this embodiment, the spectral variation parameters include at least one of the following: the change in spectral intensity of a preset characteristic band between adjacent sampling times, the peak shift, and the change in the centroid of spectral energy. The step of extracting acoustic envelope parameters based on the acoustic signal to characterize the change of acoustic amplitude over time includes: The acoustic signal is subjected to bandpass filtering and frame windowing to obtain the processed acoustic signal; Envelope detection is performed on the processed acoustic signal to obtain the acoustic envelope value corresponding to each sampling time. The acoustic envelope value is determined as the acoustic envelope parameter, and / or the acoustic envelope change is determined based on the acoustic envelope value corresponding to adjacent sampling times, and the acoustic envelope change is determined as the acoustic envelope parameter.

[0032] In some embodiments, the spectral variation parameters include at least one of the following: the change in spectral intensity of a preset characteristic band between adjacent sampling times, the peak shift between adjacent sampling times, and the change in the centroid of spectral energy between adjacent sampling times.

[0033] Integrated spectral intensity of preset characteristic bands It can be represented as: ;in, Indicates the first The integrated spectral intensity of the preset characteristic band corresponding to each sampling time. Indicates the first Each sampling time, wavelength is The intensity of the flame emission spectrum at that location, Indicates the starting wavelength of the preset characteristic band. Indicates the termination wavelength of the preset characteristic band. Indicates the first Each sampling time is used. By accumulating the spectral intensities within a preset characteristic band, the integrated spectral intensity of that sampling time can be obtained.

[0034] In some embodiments, the change in spectral intensity between adjacent sampling times can be determined by the difference in integrated spectral intensity corresponding to adjacent sampling times; the peak shift can be determined by comparing the target peak positions corresponding to adjacent sampling times; and the change in the spectral energy centroid can be determined by comparing the spectral energy centroids corresponding to adjacent sampling times. For the spectral energy centroid, the energy distribution of the flame emission spectrum can be weighted and statistically analyzed within a preset characteristic band to determine the energy center position corresponding to the current sampling time. Then, the difference in energy center positions corresponding to adjacent sampling times is compared to obtain the change in the spectral energy centroid. Through the above methods, the change process of the flame emission spectrum over time can be characterized from different perspectives.

[0035] In some embodiments, extracting acoustic envelope parameters characterizing the change of acoustic amplitude over time based on the acoustic signal includes: performing bandpass filtering and frame-by-frame windowing on the acoustic signal to obtain a processed acoustic signal; performing envelope detection on the processed acoustic signal to obtain acoustic envelope values ​​corresponding to each sampling time; determining the acoustic envelope values ​​as the acoustic envelope parameters; and / or determining the acoustic envelope change based on the acoustic envelope values ​​corresponding to adjacent sampling times, and determining the acoustic envelope change as the acoustic envelope parameters. Envelope detection can be implemented using Hilbert transform, short-time energy analysis, low-pass filtering after absolute value rectification, etc., as long as the acoustic envelope values ​​characterizing the change of acoustic amplitude over time can be obtained.

[0036] The above implementation methods can further define the specific implementation methods of spectral variation parameters and acoustic envelope parameters, making the formation process of spectral fluctuation sequences and acoustic envelope sequences clearer, thereby enhancing the engineering feasibility of cross-modal time delay compensation and subsequent multimodal identification.

[0037] In this embodiment, determining the reliability of the flame spectral mode and the reliability of the acoustic mode based on the aligned flame emission spectrum signal and the aligned acoustic signal includes: Based on the aligned flame emission spectrum signal, determine the flame spectral mode quality index; Based on the aligned acoustic wave signal, determine the acoustic wave modal quality index; Based on the flame spectral mode quality index, the reliability of the flame spectral mode is determined; The reliability of the acoustic mode is determined based on the acoustic mode quality index.

[0038] In some embodiments, determining the reliability of the flame spectral mode and the reliability of the acoustic mode based on the aligned flame emission spectral signal and the aligned acoustic signal includes: determining a flame spectral mode quality index based on the aligned flame emission spectral signal; determining an acoustic mode quality index based on the aligned acoustic signal; determining the reliability of the flame spectral mode based on the flame spectral mode quality index; and determining the reliability of the acoustic mode based on the acoustic mode quality index.

[0039] In some embodiments, the flame spectral modal quality index is used to reflect the status of the current flame emission spectral signal in terms of acquisition quality, integrity of effective information, and degree of interference; the acoustic modal quality index is used to reflect the status of the current acoustic signal in terms of signal-to-noise level, echo interference, mechanical noise influence, and sudden abnormal noise influence. When determining the credibility based on the quality index, empirical threshold mapping, piecewise function mapping, linear weighted mapping, or model learning mapping can be used to convert the quality index into the corresponding credibility value.

[0040] In a default embodiment, the quality indices for the flame spectral modes and acoustic modes can be normalized first. Then, preset weights are assigned to each quality index based on its influence on the validity of the corresponding mode. The normalized quality indices are then weighted and summed to obtain the reliability of the flame spectral modes and the acoustic modes. The obtained reliability can be further limited to a preset reliability range, such as 0 to 1, so that it can be directly used for feature weighting or weight allocation later. Higher reliability indicates that the corresponding mode is more reliable within the current time window; lower reliability indicates that the corresponding mode is more susceptible to interference.

[0041] Through the above implementation methods, the credibility of the flame spectral mode and the acoustic mode can be determined separately before fusion identification, and a default credibility mapping implementation method is given, thereby avoiding the mode that is severely interfered with from occupying too high a weight in subsequent identification and improving the robustness of multimodal fusion identification.

[0042] In this embodiment, the flame spectral modal quality index includes at least one of the following: reference spectral band attenuation rate, spectral line saturation ratio, continuous spectrum baseline drift, and spectral signal-to-noise ratio. The acoustic modal quality indicators include at least one of the following: standing wave resonance interference coefficient, mechanical noise frequency band ratio, sudden impact noise ratio, and acoustic signal-to-noise ratio.

[0043] In some embodiments, the flame spectral modal quality indicators include at least one of the following: reference spectral band attenuation rate, spectral line saturation ratio, continuous spectrum baseline drift, and spectral signal-to-noise ratio. Specifically, the reference spectral band attenuation rate characterizes the degree of intensity reduction of the flame emission spectrum within a preset reference spectral band relative to baseline conditions, reflecting the impact of smoke obstruction, observation window contamination, or optical path attenuation on the flame emission spectral signal; the spectral line saturation ratio characterizes whether there are problems such as local overexposure, sampling saturation, or insufficient dynamic range in the flame emission spectrum during acquisition; the continuous spectrum baseline drift characterizes the fluctuation of the overall background radiation level of the flame emission spectrum, reflecting baseline instability, changes in background thermal radiation, or changes in the acquisition environment; and the spectral signal-to-noise ratio characterizes the relative strength of the effective signal components in the flame emission spectrum compared to the noise components.

[0044] In some embodiments, the acoustic modal quality indicators include at least one of the following: standing wave resonance interference coefficient, mechanical noise bandwidth proportion, sudden impact noise proportion, and acoustic signal-to-noise ratio. Specifically, the standing wave resonance interference coefficient can be used to characterize the degree of influence of furnace structural echoes, cavity resonances, or local standing wave phenomena on the acoustic signal; the mechanical noise bandwidth proportion can be used to characterize the proportion of noise generated by mechanical devices such as fans, coal feeders, and induced draft equipment in the acoustic signal; the sudden impact noise proportion can be used to characterize the proportion of pulsed abnormal noise caused by slag falling, structural collisions, or other non-combustion events; and the acoustic signal-to-noise ratio can be used to characterize the effectiveness of combustion-related acoustic components relative to background noise components.

[0045] In some embodiments, the aforementioned flame spectral modal quality indices and acoustic modal quality indices can be calculated separately first, and then the reliability of the flame spectral modes and acoustic modes can be obtained through normalization, weighted summarization, threshold mapping, or model mapping. As a default implementation, each quality index can be normalized first, and then summed according to preset weights to form the reliability of the corresponding mode. Higher reliability indicates that the corresponding mode is more reliable within the current time window; lower reliability indicates that the corresponding mode is more likely to be disturbed. When the reliability of the flame spectral mode is lower than a preset first reliability threshold, the contribution of flame spectral features to fusion identification can be reduced; when the reliability of the acoustic mode is lower than a preset second reliability threshold, the contribution of acoustic features to fusion identification can be reduced.

[0046] Through the above implementation methods, the quality indicators of flame spectral modes and acoustic modes can be specifically defined, and these can be used as the basis for determining modal reliability. This provides clear data support and a default implementation path for the reliability of flame spectral modes and acoustic modes, thereby improving the robustness and reliability of the multimodal fusion identification process.

[0047] In this embodiment, the extraction of flame spectral features, acoustic features, and cross-modal coupling features based on the aligned flame emission spectrum signal and the aligned acoustic wave signal includes: Based on the aligned flame emission spectrum signal, flame spectral features are extracted; Based on the aligned acoustic signal, acoustic features are extracted; Based on the correspondence between the flame spectral features and the acoustic features, cross-modal coupling features are extracted to characterize the correlation between the flame spectral features and the acoustic features.

[0048] In some embodiments, the step of extracting flame spectral features, acoustic features, and cross-modal coupling features based on the aligned flame emission spectral signal and the aligned acoustic signal includes: extracting flame spectral features based on the aligned flame emission spectral signal; extracting acoustic features based on the aligned acoustic signal; and extracting cross-modal coupling features to characterize the correlation between the flame spectral features and the acoustic features based on the correspondence between the flame spectral features and the acoustic features.

[0049] In some embodiments, flame spectral features are used to reflect the intensity of flame radiation, peak distribution, spectral energy changes, and flame fluctuations over time; acoustic features are used to reflect the intensity of combustion pulsations, frequency band distribution characteristics, periodic fluctuations, and abnormal impact events; and cross-modal coupling features are used to reflect the correspondence between flame spectral features and acoustic features in the time domain, frequency domain, or trend of change.

[0050] In some embodiments, when extracting cross-modal coupling features, the synchronous change relationship between flame spectral features and acoustic features within the same preset time window can be utilized, or the alignment result after cross-modal time delay compensation can be used to analyze the correlation, synchronization, phase relationship, energy distribution relationship, or consistency of change between the two types of features. As a first representative calculation example, the coherence between the time-aligned spectral fluctuation sequence and acoustic envelope sequence can be calculated within a preset frequency band to characterize the consistency of the two sequences in the frequency domain. As a second representative calculation example, the coupling strength between the changes in flame spectral features and acoustic features can be calculated within the corresponding time window to characterize the consistency of the changing trends of the two types of features. In this way, fusion identification can not only utilize the independent characterization capabilities of the flame emission spectral signal and the acoustic signal, but also further enhance the ability to judge the boiler combustion state by utilizing the coupling relationship between the two types of signals.

[0051] Through the above implementation method, flame spectral features, acoustic features, and cross-modal coupling features can be extracted as three types of features that are related but at different levels. Representative cross-modal coupling feature calculation examples are added to enhance the utilization of cross-modal correlation information while retaining effective information of single modes, thereby improving the comprehensiveness and distinguishability of boiler combustion state identification.

[0052] In this embodiment, the flame spectral characteristics include at least one of the following: preset band spectral energy, band energy ratio, characteristic peak intensity, characteristic peak area, peak shift, spectral entropy, spectral slope, and scintillation frequency. The acoustic features include at least one of the following: sound pressure level, dominant frequency, frequency band energy distribution, harmonic energy ratio, spectral centroid, spectral entropy, wavelet packet energy features, pulsation amplitude, and abnormal impact count. The cross-modal coupling characteristics include at least one of the following: the coherence between the spectral fluctuation sequence and the acoustic envelope sequence; the phase difference between the spectral fluctuation sequence and the acoustic envelope sequence after time alignment based on the cross-modal time delay compensation; the phase difference drift between the spectral fluctuation sequence and the acoustic envelope sequence after time alignment based on the cross-modal time delay compensation; the energy correlation between the flame spectral characteristics and the acoustic characteristics in a common frequency band; and the coupling strength between the changes in the flame spectral characteristics and the changes in the acoustic characteristics within a corresponding time window.

[0053] In some embodiments, the flame spectral characteristics include at least one of the following: preset band spectral energy, band energy ratio, characteristic peak intensity, characteristic peak area, peak shift, spectral entropy, spectral slope, and scintillation frequency. Specifically, the preset band spectral energy and band energy ratio can be used to characterize the distribution of radiative energy in different spectral intervals; the characteristic peak intensity and characteristic peak area can be used to characterize the strength of the radiative characteristics corresponding to the target peak; the peak shift can be used to characterize the shift of the position of the characteristic peak in the flame emission spectrum over time; the spectral entropy can be used to characterize the complexity of the spectral distribution; the spectral slope can be used to characterize the overall morphological changes in the flame emission spectrum; and the scintillation frequency characteristic can be used to characterize the periodic fluctuations of the flame over time.

[0054] In some embodiments, the acoustic features include at least one of sound pressure level, dominant frequency, frequency band energy distribution, harmonic energy ratio, spectral centroid, spectral entropy, wavelet packet energy characteristics, pulsation amplitude, and abnormal impact count. Specifically, sound pressure level can be used to characterize the overall intensity of combustion noise; dominant frequency and frequency band energy distribution can be used to characterize the distribution of combustion pulsations across different frequency ranges; harmonic energy ratio can be used to characterize periodic fluctuations or resonance phenomena; spectral centroid and spectral entropy can be used to characterize the overall shape of the sound wave spectrum; wavelet packet energy characteristics can be used to characterize multi-scale frequency band energy distribution; pulsation amplitude can be used to characterize the intensity of combustion fluctuations; and abnormal impact count can be used to characterize the frequency of unstable combustion or abnormal events.

[0055] In some embodiments, the cross-modal coupling features include at least one of the following: coherence between the spectral wave sequence and the acoustic envelope sequence; phase difference between the spectral wave sequence and the acoustic envelope sequence after time alignment based on cross-modal time delay compensation; phase difference drift between the spectral wave sequence and the acoustic envelope sequence after time alignment based on cross-modal time delay compensation; energy correlation between flame spectral features and acoustic features in a common frequency band; and coupling strength between changes in flame spectral features and changes in acoustic features within a corresponding time window. As one representative calculation method, the common frequency components of the spectral wave sequence and the acoustic envelope sequence can be compared within a preset frequency band to determine the coherence between them. As another representative calculation method, the degree of unidirectional change between changes in flame spectral features and changes in acoustic features can be compared within multiple consecutive time windows to determine the coupling strength between them. The above-mentioned cross-modal coupling features characterize the cooperative relationship between the flame emission spectral signal and the acoustic signal from multiple perspectives, including synchronization, phase consistency, frequency band consistency, and change consistency.

[0056] Through the above implementation methods, flame spectral features, acoustic features, and cross-modal coupling features can be specifically listed, and representative cross-modal coupling feature implementation methods can be added, making the feature system on which boiler combustion state identification is based more complete, which is conducive to improving the distinguishability between different combustion states and enhancing the stability of identification results.

[0057] In this embodiment, the fusion identification based on the confidence level of the flame spectral mode, the confidence level of the acoustic mode, the flame spectral features, the acoustic features, the cross-modal coupling features, and the boiler operating parameters to obtain the boiler combustion state identification result includes: Based on the credibility of the flame spectral modes, the first weight of the flame spectral features in the fusion identification is determined; Based on the credibility of the acoustic mode, the second weight of the acoustic feature in the fusion identification is determined; The flame spectral features are weighted based on the first weight to obtain the weighted flame spectral features. The acoustic features are weighted based on the second weight to obtain the weighted acoustic features. The weighted flame spectral features, the weighted acoustic features, the cross-modal coupling features, and the boiler operating parameters are fused to obtain fused identification input information; Based on the fused identification input information, the boiler combustion state identification result is determined; The boiler combustion state identification results include combustion state category, state intensity level and / or state confidence level. The combustion state category is stable combustion state, oxygen-deficient combustion state, excess air combustion state, combustion oscillation state, incomplete combustion trend state, flameout risk state or coking risk state.

[0058] In some embodiments, the step of fusing and identifying the boiler combustion state based on the confidence level of the flame spectral mode, the confidence level of the acoustic mode, the flame spectral features, the acoustic features, the cross-modal coupling features, and the boiler operating parameters to obtain the boiler combustion state identification result includes: determining a first weight of the flame spectral features in the fusion identification based on the confidence level of the flame spectral mode; determining a second weight of the acoustic features in the fusion identification based on the confidence level of the acoustic mode; weighting the flame spectral features based on the first weight to obtain weighted flame spectral features; weighting the acoustic features based on the second weight to obtain weighted acoustic features; fusing the weighted flame spectral features, the weighted acoustic features, the cross-modal coupling features, and the boiler operating parameters to obtain fusion identification input information; and determining the boiler combustion state identification result based on the fusion identification input information.

[0059] In some embodiments, the higher the confidence level of the flame spectral mode, the greater the first weight of the flame spectral features in the fusion identification; the higher the confidence level of the acoustic mode, the greater the second weight of the acoustic features in the fusion identification. In this way, the contribution of each feature in the fusion identification can be adaptively adjusted according to the reliability of the two modes within the current time window.

[0060] In some embodiments, the fusion identification input information can be obtained through feature-level concatenation, weighted aggregation, feature mapping, or other multi-source data fusion methods. Boiler operating parameters can be used as auxiliary operating condition information to participate in the fusion identification, reflecting the current boiler operating status and its impact on the combustion state.

[0061] In some embodiments, the boiler combustion state identification result includes a combustion state category, a state intensity level, and / or a state confidence level. The combustion state category can be selected from one of the following: stable combustion state, oxygen-deficient combustion state, excess air combustion state, combustion oscillation state, incomplete combustion trend state, flameout risk state, and coking risk state. The state intensity level can be used to characterize the degree to which the current combustion state deviates from the normal operating state, and the state confidence level can be used to characterize the reliability of the current identification result.

[0062] In some embodiments, the boiler combustion state identification result can be obtained through rule-based determination or model-based identification. As a preferred embodiment of this application, the boiler combustion state identification result is determined using a rule-based determination method. Specifically, a set of determination rules between different feature combinations and different combustion state categories can be pre-established. During fusion identification, the weights of flame spectral features and acoustic features are first determined based on the confidence levels of the flame spectral mode and acoustic mode, respectively. Then, the flame spectral features and acoustic features are weighted and combined with cross-modal coupling features and boiler operating parameters to form fusion identification input information. Subsequently, the fusion identification input information is matched with the preset determination rule set to output the boiler combustion state identification result. The state intensity level can be determined based on the degree to which the relevant features deviate from a preset threshold, and the state confidence level can be determined based on the number of rules satisfied, the confidence level of the modes involved in the determination, and the consistency of identification results within multiple consecutive time windows.

[0063] In other embodiments, a model identification method can be used as an alternative, in which the fused identification input information is input into a pre-trained classification or regression model to output the combustion state category, state intensity level, and / or state confidence. To improve the stability of the identification results, temporal consistency correction can also be performed on the identification results obtained from multiple consecutive time windows.

[0064] Through the above implementation methods, the credibility of flame spectral modes and acoustic modes can be introduced to adaptively weight the features, and cross-modal coupling features and boiler operating parameters can be combined for fusion identification. At the same time, the fixed rule judgment method is used as the preferred embodiment, thereby making fuller use of multi-source information and improving the accuracy, stability and engineering adaptability of boiler combustion state identification.

[0065] In this embodiment, the method further includes: Based on the combustion state category in the boiler combustion state identification results, and in combination with the state intensity level and / or state confidence, determine the parameter optimization strategy corresponding to the combustion state category; Based on the parameter optimization strategy, the optimization results of the boiler combustion parameters are determined; Based on the optimization results of the boiler combustion parameters, the parameters of the boiler combustion process are adjusted. The optimization results of the boiler combustion parameters include the adjustment amount of at least one of the following: fuel feed rate, primary air volume, secondary air volume, induced draft volume, and the ratio of primary air to secondary air.

[0066] In some embodiments, the method further includes: determining a parameter optimization strategy corresponding to the combustion state category based on the combustion state category in the boiler combustion state identification result, and in combination with the state intensity level and / or state confidence level; determining the boiler combustion parameter optimization result based on the parameter optimization strategy; and adjusting the parameters of the boiler combustion process based on the boiler combustion parameter optimization result.

[0067] In some embodiments, a correspondence table between combustion state categories and parameter optimization strategies can be pre-established. Specifically, when the combustion state category is an oxygen-deficient combustion state, one or more strategies can be adopted, such as increasing the secondary air volume, appropriately reducing the fuel feed rate, and adjusting the primary air to secondary air ratio; when the combustion state category is an excess air combustion state, one or more strategies can be adopted, such as reducing the total air volume, reducing the induced draft volume, and optimizing the primary air to secondary air ratio; when the combustion state category is a combustion oscillation state, strategies can be adopted, such as reducing the adjustment amplitude, prioritizing the stabilization of airflow distribution, and gradually correcting the primary air to secondary air ratio or induced draft volume; when the combustion state category is an incomplete combustion trend state, strategies can be adopted, such as appropriately increasing the secondary air volume and optimizing the matching relationship between the fuel feed rate and the air distribution; when the combustion state category is a flameout risk state, strategies can be adopted, such as appropriately increasing the combustion stabilization-related air volume and limiting actions that are detrimental to combustion stabilization; when the combustion state category is a coking risk state, strategies can be adopted, such as adjusting the flame center position, optimizing local air distribution, or controlling the fuel feed rate.

[0068] In some embodiments, the state intensity level can be used to determine the parameter adjustment range. For example, for the same combustion state category, a smaller adjustment amount is used if the state intensity level is low, and a larger adjustment amount is used if the state intensity level is high. State confidence can be used to adjust the conservatism of strategy execution. For example, when the state confidence is high, the target adjustment amount can be directly executed; when the state confidence is low, a trial-and-error small-step adjustment can be performed first, and then corrected based on subsequent identification results.

[0069] In some embodiments, the boiler combustion parameter optimization results include adjustments to at least one of the following: fuel feed rate, primary air volume, secondary air volume, induced draft volume, and primary air to secondary air ratio. Based on the boiler combustion parameter optimization results, the boiler combustion process parameters can be adjusted via a coal feeder, forced draft fan, induced draft fan, damper actuator, or other control devices.

[0070] In some embodiments, the boiler combustion parameter optimization results should meet adjustment constraints within the boiler's allowable operating range during execution. These adjustment constraints may include at least one of the following: furnace negative pressure constraint, upper limit constraint for airflow adjustment, step size constraint for fuel feed rate variation, amplitude constraint for induced draft air volume adjustment, and range constraint for the primary and secondary air ratio variation. When the target adjustment amount obtained according to the parameter optimization strategy exceeds the corresponding adjustment constraint, the target adjustment amount can be limited, executed in stages, or conservatively corrected to ensure that the final boiler combustion parameter optimization results remain within the boiler's allowable operating range. After parameter adjustment, flame emission spectrum signals, acoustic signals, and boiler operating parameters can be continuously collected to re-identify the boiler combustion state, thus forming a closed-loop adjustment process.

[0071] Through the above implementation methods, the boiler combustion state identification results can be further used for the selection of parameter optimization strategies and the adjustment of boiler combustion parameters, and adjustment constraints within the allowable operating range of the boiler can be introduced, so that the boiler combustion state identification can directly serve the boiler combustion operation control, thereby improving the timeliness, optimization capability and engineering rationality of boiler combustion process adjustment.

[0072] In one embodiment, a boiler combustion state identification system based on the fusion of flame spectrum and acoustic waves is provided, comprising: The data acquisition module is used to acquire flame emission spectrum signals from the target combustion zone of the boiler, acoustic signals generated during the boiler combustion process, and boiler operating parameters. A sequence construction module is used to construct a spectral wave sequence based on the flame emission spectral signal and an acoustic envelope sequence based on the acoustic wave signal; The time delay compensation module is used to determine the cross-modal time delay compensation amount based on the correlation or coherence between the spectral fluctuation sequence and the acoustic envelope sequence within a preset time window. The time alignment module is used to perform time alignment between the flame emission spectrum signal and the acoustic signal based on the cross-modal time delay compensation amount, so as to obtain the aligned flame emission spectrum signal and the aligned acoustic signal. A credibility determination module is used to determine the credibility of the flame spectral mode and the credibility of the acoustic mode based on the aligned flame emission spectral signal and the aligned acoustic signal. The feature extraction module is used to extract flame spectral features, acoustic features, and cross-modal coupling features based on the aligned flame emission spectrum signal and the aligned acoustic wave signal. The fusion identification module is used to perform fusion identification based on the credibility of the flame spectral mode, the credibility of the acoustic mode, the flame spectral features, the acoustic features, the cross-modal coupling features, and the boiler operating parameters to obtain the boiler combustion state identification result.

[0073] The specific implementation details of each module can be found in the above description of the limitations of the boiler combustion state identification method based on the fusion of flame spectrum and sound waves, and will not be repeated here.

[0074] In some embodiments, a boiler combustion state identification system based on the fusion of flame spectrum and acoustic waves includes: a data acquisition module, a sequence construction module, a time delay compensation module, a time alignment module, a reliability determination module, a feature extraction module, and a fusion identification module. The above modules can be implemented using an industrial controller, industrial computer, embedded processor, edge computing device, or a combination thereof; they can also be implemented using a combination of hardware circuits and software programs.

[0075] The system comprises the following modules: a data acquisition module for acquiring flame emission spectral signals from the target combustion zone of the boiler, acoustic signals generated during the boiler combustion process, and boiler operating parameters; a sequence construction module for constructing a spectral fluctuation sequence based on the flame emission spectral signals and an acoustic envelope sequence based on the acoustic signals; a time delay compensation module for determining the cross-modal time delay compensation amount based on the correlation or coherence between the spectral fluctuation sequence and the acoustic envelope sequence within a preset time window; and a time alignment module for performing time alignment of the flame emission spectral signals and acoustic signals based on the cross-modal time delay compensation amount to obtain the aligned flame emission spectrum. The system consists of a signal and an aligned acoustic signal; a credibility determination module is used to determine the credibility of the flame spectral mode and the acoustic mode based on the aligned flame emission spectrum signal and the aligned acoustic signal; a feature extraction module is used to extract flame spectral features, acoustic features and cross-modal coupling features based on the aligned flame emission spectrum signal and the aligned acoustic signal; and a fusion identification module is used to perform fusion identification based on the credibility of the flame spectral mode, the credibility of the acoustic mode, the flame spectral features, the acoustic features, the cross-modal coupling features and the boiler operating parameters to obtain the boiler combustion state identification result.

[0076] In some embodiments, the modules can communicate with each other via bus, industrial Ethernet, fieldbus, or internal data interface. The modules can be centrally deployed on the same control platform or distributed across the acquisition and analysis ends. Specifically, the data acquisition module and some preprocessing functions can be deployed at the field acquisition end, while time delay compensation, reliability determination, feature extraction, and fusion identification functions can be deployed at the edge computing end or the host computer.

[0077] Through the above implementation methods, the boiler combustion state identification method can be mapped to a modular system structure, which is conducive to the engineering deployment and online operation of the boiler combustion state identification function, thereby improving the feasibility of the solution in industrial sites.

[0078] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for identifying boiler combustion state based on the fusion of flame spectrum and acoustic waves, characterized in that, The method includes: Collect flame emission spectrum signals from the target combustion zone of the boiler, acoustic signals generated during the boiler combustion process, and boiler operating parameters; A spectral wave sequence is constructed based on the flame emission spectral signal, and an acoustic envelope sequence is constructed based on the acoustic wave signal; The cross-modal time delay compensation amount is determined based on the correlation or coherence between the spectral fluctuation sequence and the acoustic envelope sequence within a preset time window. Based on the cross-modal time delay compensation, the flame emission spectrum signal and the acoustic signal are time-aligned to obtain the aligned flame emission spectrum signal and the aligned acoustic signal. Based on the aligned flame emission spectrum signal and the aligned acoustic signal, the reliability of the flame spectral mode and the reliability of the acoustic mode are determined. Based on the aligned flame emission spectrum signal and the aligned acoustic signal, flame spectral features, acoustic features and cross-modal coupling features are extracted; The boiler combustion state identification result is obtained by fusing and identifying the reliability of the flame spectral mode, the reliability of the acoustic mode, the flame spectral features, the acoustic features, the cross-modal coupling features, and the boiler operating parameters.

2. The method according to claim 1, characterized in that, The construction of the spectral fluctuation sequence based on the flame emission spectral signal includes: Based on the flame emission spectrum signal, spectral change parameters for characterizing the change of flame emission spectrum over time are extracted; The values ​​of the spectral variation parameters are arranged in chronological order to form the spectral fluctuation sequence; The construction of the acoustic envelope sequence based on the acoustic signal includes: Based on the sound wave signal, acoustic envelope parameters are extracted to characterize the change of sound wave amplitude over time. The values ​​of the acoustic envelope parameters are arranged in chronological order to form the acoustic envelope sequence; The step of determining the cross-modal time delay compensation amount based on the correlation or coherence between the spectral fluctuation sequence and the acoustic envelope sequence within a preset time window includes: Within the data segment corresponding to the preset time window, a relative time shift search is performed on the spectral fluctuation sequence and the acoustic envelope sequence within a preset time shift range, and the time shift amount corresponding to the maximum correlation coefficient or coherence coefficient between the spectral fluctuation sequence and the acoustic envelope sequence is determined as the cross-modal time delay compensation amount.

3. The method according to claim 2, characterized in that, The spectral variation parameters include at least one of the following: the change in spectral intensity, the peak shift, and the change in the centroid of spectral energy between adjacent sampling times for a preset characteristic band. The step of extracting acoustic envelope parameters based on the acoustic signal to characterize the change of acoustic amplitude over time includes: The acoustic signal is subjected to bandpass filtering and frame windowing to obtain the processed acoustic signal; Envelope detection is performed on the processed acoustic signal to obtain the acoustic envelope value corresponding to each sampling time. The acoustic envelope value is determined as the acoustic envelope parameter, and / or the acoustic envelope change is determined based on the acoustic envelope value corresponding to adjacent sampling times, and the acoustic envelope change is determined as the acoustic envelope parameter.

4. The method according to claim 1, characterized in that, The step of determining the reliability of the flame spectral mode and the reliability of the acoustic mode based on the aligned flame emission spectrum signal and the aligned acoustic wave signal includes: Based on the aligned flame emission spectrum signal, determine the flame spectral mode quality index; Based on the aligned acoustic wave signal, determine the acoustic wave modal quality index; Based on the flame spectral mode quality index, the reliability of the flame spectral mode is determined; The reliability of the acoustic mode is determined based on the acoustic mode quality index.

5. The method according to claim 4, characterized in that, The flame spectral modal quality index includes at least one of the following: reference band attenuation rate, spectral line saturation ratio, continuous spectrum baseline drift, and spectral signal-to-noise ratio. The acoustic modal quality indicators include at least one of the following: standing wave resonance interference coefficient, mechanical noise frequency band ratio, sudden impact noise ratio, and acoustic signal-to-noise ratio.

6. The method according to claim 1, characterized in that, The extraction of flame spectral features, acoustic features, and cross-modal coupling features based on the aligned flame emission spectrum signal and the aligned acoustic wave signal includes: Based on the aligned flame emission spectrum signal, flame spectral features are extracted; Based on the aligned acoustic signal, acoustic features are extracted; Based on the correspondence between the flame spectral features and the acoustic features, cross-modal coupling features are extracted to characterize the correlation between the flame spectral features and the acoustic features.

7. The method according to claim 6, characterized in that, The flame spectral characteristics include at least one of the following: preset band spectral energy, band energy ratio, characteristic peak intensity, characteristic peak area, peak shift, spectral entropy, spectral slope, and scintillation frequency. The acoustic features include at least one of the following: sound pressure level, dominant frequency, frequency band energy distribution, harmonic energy ratio, spectral centroid, spectral entropy, wavelet packet energy features, pulsation amplitude, and abnormal impact count. The cross-modal coupling feature includes at least one of the following: the coherence between the spectral fluctuation sequence and the acoustic envelope sequence; The phase difference between the spectral fluctuation sequence and the acoustic envelope sequence after time alignment is completed based on the cross-modal time delay compensation; The phase difference drift between the spectral fluctuation sequence and the acoustic envelope sequence after time alignment based on the cross-modal time delay compensation; the energy correlation between the flame spectral features and the acoustic features in the common frequency band; The coupling strength between the changes in flame spectral characteristics and the changes in acoustic characteristics within the corresponding time window.

8. The method according to claim 1, characterized in that, The boiler combustion state identification result is obtained by fusing and identifying the reliability of the flame spectral mode, the reliability of the acoustic mode, the flame spectral features, the acoustic features, the cross-modal coupling features, and the boiler operating parameters, including: Based on the credibility of the flame spectral modes, the first weight of the flame spectral features in the fusion identification is determined; Based on the credibility of the acoustic mode, the second weight of the acoustic feature in the fusion identification is determined; The flame spectral features are weighted based on the first weight to obtain the weighted flame spectral features. The acoustic features are weighted based on the second weight to obtain the weighted acoustic features. The weighted flame spectral features, the weighted acoustic features, the cross-modal coupling features, and the boiler operating parameters are fused to obtain fused identification input information; Based on the fused identification input information, the boiler combustion state identification result is determined; The boiler combustion state identification results include combustion state category, state intensity level and / or state confidence level. The combustion state category is stable combustion state, oxygen-deficient combustion state, excess air combustion state, combustion oscillation state, incomplete combustion trend state, flameout risk state or coking risk state.

9. The method according to claim 8, characterized in that, The method further includes: Based on the combustion state category in the boiler combustion state identification results, and in combination with the state intensity level and / or state confidence, determine the parameter optimization strategy corresponding to the combustion state category; Based on the parameter optimization strategy, the optimization results of the boiler combustion parameters are determined; Based on the optimization results of the boiler combustion parameters, the parameters of the boiler combustion process are adjusted. The optimization results of the boiler combustion parameters include the adjustment amount of at least one of the following: fuel feed rate, primary air volume, secondary air volume, induced draft volume, and the ratio of primary air to secondary air.

10. A boiler combustion state identification system based on the fusion of flame spectrum and acoustic waves, characterized in that, The system includes: The data acquisition module is used to acquire flame emission spectrum signals from the target combustion zone of the boiler, acoustic signals generated during the boiler combustion process, and boiler operating parameters. A sequence construction module is used to construct a spectral wave sequence based on the flame emission spectral signal and an acoustic envelope sequence based on the acoustic wave signal; The time delay compensation module is used to determine the cross-modal time delay compensation amount based on the correlation or coherence between the spectral fluctuation sequence and the acoustic envelope sequence within a preset time window. The time alignment module is used to perform time alignment between the flame emission spectrum signal and the acoustic signal based on the cross-modal time delay compensation amount, so as to obtain the aligned flame emission spectrum signal and the aligned acoustic signal. A credibility determination module is used to determine the credibility of the flame spectral mode and the credibility of the acoustic mode based on the aligned flame emission spectral signal and the aligned acoustic signal. The feature extraction module is used to extract flame spectral features, acoustic features, and cross-modal coupling features based on the aligned flame emission spectrum signal and the aligned acoustic wave signal. The fusion identification module is used to perform fusion identification based on the credibility of the flame spectral mode, the credibility of the acoustic mode, the flame spectral features, the acoustic features, the cross-modal coupling features, and the boiler operating parameters to obtain the boiler combustion state identification result.

Images