A coal-fired power plant boiler combustion stability online monitoring method and system based on multi-modal signal fusion

Abstract

The present application relates to the technical field of boiler combustion monitoring, in particular to a coal-fired power plant boiler combustion stability online monitoring method and system based on multi-modal signal fusion, the method synchronously collects radiation images, radiation intensity signals, radiation heat flow signals and boiler load signals Load, and inverts temperature signals; after filtering and smoothing the multi-modal signals, Z-score standardization is realized; based on a sliding window, the variation coefficients of each standardized signal are calculated, mapped into single-modal stability indexes and fused into a comprehensive stability index; combined with four-level state determination, when or amplitude does not match the current load, and are corrected to avoid misjudgment caused by high load weak fire or low load abnormally strong fire. The present application realizes high-precision, robust combustion state monitoring within the full load range.

Description

Technical Field

[0001] This invention relates to the field of boiler combustion monitoring technology, and in particular to an online monitoring method and system for combustion stability of coal-fired power plant boilers based on multimodal signal fusion. Background Technology

[0002] In industrial production, flame stability is crucial for the safe, efficient, and clean operation of boilers. Unstable combustion not only reduces energy efficiency and increases pollutant emissions but can also lead to equipment damage and even safety accidents. Therefore, real-time, accurate, and reliable monitoring and stability assessment of flame conditions has always been a key research direction in the fields of industrial automation and safe production.

[0003] Traditional monitoring methods mainly rely on flame television images or single thermal signals (such as furnace negative pressure and CO concentration). These methods suffer from several drawbacks: image-based methods are highly subjective and difficult to quantify stability; single signals are susceptible to interference (such as thermocouple failure due to ash accumulation); fixed thresholds cannot adapt to variable load operation, resulting in high false alarm rates at high loads and serious missed alarms at low loads; and there is a lack of verification on the physical rationality of whether the flame intensity matches the load, which can easily lead to misjudging "high load and weak flame" as stable conditions.

[0004] In recent years, although some studies have introduced multi-sensor fusion, most of them have not considered the dynamic correlation between load and signal amplitude, nor have they established an adaptive reference system with load as the anchor point, resulting in insufficient reliability under deep peak shaving conditions, so improvements are needed. Summary of the Invention

[0005] In view of this, the purpose of this invention is to propose an online monitoring method and system for combustion stability of coal-fired power plant boilers based on multimodal signal fusion, so as to solve the problems of false alarms and missed alarms caused by fixed thresholds, single signal dependence and lack of load matching verification in the prior art.

[0006] To achieve the above objectives, this invention provides an online monitoring method for combustion stability of coal-fired power plant boilers based on multimodal signal fusion, comprising the following steps:

[0007] S1. Synchronously acquire radiation images and radiation intensity signals of the combustion zone. and radiative heat flow signal And the boiler load signal, based on the radiation image, is used to invert the temperature signal through a radiation thermometry model. ;

[0008] S2, Temperature signal Radiation intensity signal and radiative heat flow signal Savitzky-Golay filtering was performed separately, and then smoothed using an exponentially weighted moving average to obtain the preprocessed signal;

[0009] S3. Determine the load reference range corresponding to the current operating condition based on the boiler load signal Load, and acquire the temperature signal within the load reference range. Radiation intensity signal and radiative heat flow signal Their respective historical baseline mean Standard deviation from the benchmark ,and The preprocessed signals are Z-score normalized to obtain the normalized signals. , and ;

[0010] S4. Within the sliding time window, calculate respectively , and coefficient of variation , and Each coefficient of variation is mapped to its corresponding single-mode stability index, and the single-mode stability indices are weighted and summed to obtain the comprehensive stability index. ;

[0011] S5, Based on absolute value of radiation intensity Absolute value of the rate of change of radiation intensity and absolute value of radiative heat flux The probability of flame presence is calculated using a pre-trained sigmoid function model. ;

[0012] S6, if If the value is less than the first threshold, it is determined to be a flameless state; otherwise, it is determined according to the comprehensive stability index. The combustion state is classified into severely unstable, unstable, or stable based on the comparison results with multiple preset stability thresholds, and the corresponding warning information is output.

[0013] In step S6, based on the load reference range determined in step S3, the current temperature signal is... Radiation intensity signal The amplitude is checked for load consistency. If it exceeds the reasonable amplitude range under the load, then... or Make corrections.

[0014] Preferably, in step S1, the radiation thermometry model is constructed based on Planck's blackbody radiation law or the two-color thermometry method, and the temperature signal This is a time series of the average or highest temperature in the combustion zone.

[0015] Preferably, in step S3, the load reference interval is defined by dividing the boiler load range into multiple discrete intervals, each of which has a pre-stored historical reference average value. Standard deviation from the benchmark .

[0016] Preferably, in step S3, the pre-processed , , Perform Z-score standardization separately:

[0017] , .

[0018] Preferably, in step S4, the coefficient of variation is calculated for each standardized signal:

[0019] , ;

[0020] in, This represents the average value. Indicates standard deviation;

[0021] ;

[0022] ;

[0023] in, Represents the first normalized signal within the sliding time window. Each sample value, For sampling sequence number, This represents the number of sampling points within the sliding time window;

[0024] Map each coefficient of variation to a single-mode stability index:

[0025] , ;

[0026] in, This is an empirical attenuation coefficient used to adjust the stability's sensitivity to fluctuations;

[0027] The overall stability index is obtained by weighted summation of the individual mode stability indices. ;

[0028] ;

[0029] in, , and These correspond to the combined weights of the temperature stability index, radiation intensity stability index, and radiation heat flux stability index, respectively. ,and .

[0030] Preferably, in step S5, the probability of the flame existing is... Calculated using a pre-trained Sigmoid function model:

[0031] ;

[0032] in, , , , For the Sigmoid function, , and The model parameters are obtained by training based on historical combustion sample data, and For feature fusion weights, The characteristic sensitivity coefficient, These are bias parameters;

[0033] in, , ,and .

[0034] Preferably, in step S6, the first threshold is 0.2, and the multiple stability thresholds include 0.3 and 0.6. The specific steps for judging the flame state are as follows:

[0035] like If the value is less than 0.2, it is determined that there is no flame and a no-flame alarm is triggered;

[0036] like If the value is ≥0.2, then according to the comprehensive stability index... Make a judgment;

[0037] like ≥0.2, and If the value is less than 0.3, it is considered severely unstable and a severe instability alarm is triggered.

[0038] If 0.3≤ If the value is less than 0.6, it is considered unstable and an instability alarm is triggered.

[0039] like If the value is ≥0.6, it is considered to be in a stable state.

[0040] Preferably, in step S6, the reasonable amplitude range is based on the historical benchmark average value corresponding to the load benchmark interval to which the current boiler load belongs. Standard deviation from the benchmark Confirmed, among which The lower limit of the reasonable amplitude is: The upper limit of the reasonable amplitude is ,in ;

[0041] The load consistency verification includes: determining the current temperature signal. or radiation intensity signal Whether it falls within their respective reasonable amplitude range;

[0042] If the verification results are inconsistent, the following correction operations will be performed:

[0043] If the current boiler load is greater than or equal to the preset load threshold, and or When the value is below its corresponding reasonable lower limit, the probability of flame existence will be... Force it to be less than 0.2 to trigger the no-flame state determination;

[0044] If the current boiler load is less than the preset load threshold, and or When the value exceeds its corresponding reasonable upper limit, the overall stability index will be adjusted. Multiply by a consistency weighting factor less than 1 ,in ,and This is a fixed constant calibrated based on historical operating data, and the corrected... Used for determining combustion status;

[0045] The preset load threshold is the percentage of boiler load that characterizes the switching point between high and low load conditions.

[0046] An online monitoring system for combustion stability of a coal-fired power plant boiler based on multimodal signal fusion includes:

[0047] Intelligent flame spectral image detector, used to acquire radiation images and with built-in temperature inversion module;

[0048] The radiation intensity sensor and the radiation heat flux sensor output signals respectively. and ;

[0049] The load interface module is used to receive the boiler load signal (Load) from the DCS system.

[0050] An edge computing unit is configured with a processor and a memory, wherein the memory stores a computer program, and the computer program is executed by the processor to implement the above-mentioned monitoring method.

[0051] The alarm and visualization module is used to output the combustion status determination result and the probability of flame presence, and generate early warning information. It also supports data interaction with the DCS system.

[0052] The beneficial effects of this invention are as follows: This invention simultaneously collects multiple physical quantities such as temperature, radiation intensity, radiant heat flux, and boiler load, and dynamically calls the historical benchmark average value of the corresponding load benchmark interval according to the current boiler load. Standard deviation from the benchmark Z-score normalization is applied to temperature, radiation intensity, and radiative heat flux signals to effectively eliminate the systematic shift in the absolute amplitude of multimodal signals under different operating loads.

[0053] Based on this, the coefficient of variation of each standardized signal is calculated using a sliding time window to accurately quantify the instantaneous fluctuation intensity of the combustion process and integrate them into a comprehensive stability index. Furthermore, load consistency verification is carried out by combining reasonable amplitude ranges within the same load reference interval. Physically unreasonable flame signals are corrected before state determination, thereby achieving highly sensitive and robust online assessment of combustion stability across the entire operating range, significantly improving the safety monitoring capabilities of coal-fired power plant boilers in complex operating scenarios such as deep peak shaving. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a flowchart of the steps of the present invention;

[0056] Figure 2 This is a schematic diagram of the overall process of the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0058] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this invention should have the ordinary meaning understood by those skilled in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0059] like Figure 1 , Figure 2 As shown, an online monitoring method for combustion stability of coal-fired power plant boilers based on multimodal signal fusion includes the following steps:

[0060] S1. Synchronously acquire radiation images and radiation intensity signals of the combustion zone. and radiative heat flow signal And the boiler load signal, based on the radiation image, is used to invert the temperature signal through a radiation thermometry model. The boiler load signal is read in real time from the power plant's distributed control system (DCS) and represents the percentage of the current boiler output relative to the maximum continuous evaporation capacity (BMCR).

[0061] In step S1, the radiation thermometry model is constructed based on Planck's blackbody radiation law or the two-color thermometry method, and the temperature signal This is a time series of the average or highest temperature in the combustion zone.

[0062] S2, Temperature signal Radiation intensity signal and radiative heat flow signal Savitzky-Golay filtering was performed separately, and then smoothed using an exponentially weighted moving average to obtain the preprocessed signal;

[0063] Perform Savitzky-Golay filtering on the original signal:

[0064]

[0065] in, This is the filtered output signal. For the first in the window The original signal values ​​of each sampling point These are the filter coefficients obtained by fitting using the least squares method. The width of the window is half its width, and the length of the filter window is [value missing]. ;

[0066] In this embodiment, the window length is 5 and the polynomial order is 3.

[0067] After Savitzky-Golay filtering, an exponentially weighted moving average (EWMA) is used for smoothing.

[0068]

[0069] in, This is the smoothed output value at the current moment. Enter the value for the current moment. This is the smoothed output value from the previous time step. As a smoothing factor, satisfying , The sampling time.

[0070] S3. Determine the load reference range corresponding to the current operating condition based on the boiler load signal Load, and acquire the temperature signal within the load reference range. Radiation intensity signal and radiative heat flow signal Their respective historical baseline mean Standard deviation from the benchmark ,and The preprocessed signals are Z-score normalized to obtain the normalized signals. , and ;

[0071] In step S3, the load reference interval is defined by dividing the boiler load range into multiple discrete intervals, each of which has a pre-stored historical reference average value. Standard deviation from the benchmark .

[0072] In step S3, the preprocessed , , Perform Z-score standardization separately:

[0073] , .

[0074] Divide the boiler's load range (e.g., 30%~100% of rated output) into several segments, for example:

[0075] Range 1, 30%~50%

[0076] Range 2, 50%~70%

[0077] Range 3, 70%~90%

[0078] Range 4, 90%~100%

[0079] Each segment is called a load reference interval, representing a typical operating condition.

[0080] For each load reference interval, the system pre-stores the statistical parameters corresponding to each modal signal:

[0081] , ,

[0082] in, Represents a temperature signal. Indicates radiation intensity signal, Indicates the radiative heat flux signal; This indicates the signal under the load reference range. Based on the historical average value during normal stable combustion, This represents the corresponding historical standard deviation.

[0083] These parameters are typically obtained offline through statistical analysis of historical operating data (e.g., collecting data from all periods of 50%–70% load and stable combustion over the past 6 months, and calculating...). , , (mean and standard deviation).

[0084] During real-time operation, the system reads the current boiler load (e.g., Load=63%), determines which discrete interval (e.g., 50%~70%) 63% belongs to, and calls the corresponding function for that interval. and Z-score normalization is used to normalize the current signal, so that signals under different loads are mapped to a uniform dimensionless scale, ensuring that the subsequent coefficient of variation only reflects the intensity of combustion disturbance, rather than the difference in operating conditions.

[0085] S4. Within the sliding time window, calculate respectively , and coefficient of variation , and Each coefficient of variation is mapped to its corresponding single-mode stability index, and the single-mode stability indices are weighted and summed to obtain the comprehensive stability index. ;

[0086] In step S4, the coefficient of variation is calculated for each standardized signal:

[0087] , ;

[0088] in, This represents the average value. Indicates standard deviation; It reflects the relative degree of fluctuation; the smaller the value, the more "stable" the signal.

[0089] ;

[0090] ;

[0091] in, Represents the first normalized signal within the sliding time window. Each sample value, For sampling sequence number, This represents the number of sampling points within the sliding time window;

[0092] Each coefficient of variation is mapped to a single-mode stability index (the larger the value, the more stable the mode):

[0093] , ;

[0094] in, This is an empirical attenuation coefficient used to adjust the stability's sensitivity to fluctuations;

[0095] The overall stability index is obtained by weighted summation of the individual mode stability indices. ;

[0096] ;

[0097] in, , and These correspond to the combined weights of the temperature stability index, radiation intensity stability index, and radiation heat flux stability index, respectively. ,and .

[0098] S5, Based on absolute value of radiation intensity Absolute value of the rate of change of radiation intensity and absolute value of radiative heat flux The probability of flame presence is calculated using a pre-trained sigmoid function model. ;

[0099] In step S5, the probability of the flame existing is... Calculated using a pre-trained Sigmoid function model:

[0100] ;

[0101] in, , , , For the Sigmoid function, , and The model parameters are obtained by training based on historical combustion sample data, and For feature fusion weights, The characteristic sensitivity coefficient, These are bias parameters;

[0102] in, , ,and .

[0103] S6, if If the value is less than the first threshold, it is determined to be a flameless state; otherwise, it is determined according to the comprehensive stability index. The combustion state is classified into severely unstable, unstable, or stable based on the comparison results with multiple preset stability thresholds, and the corresponding warning information is output.

[0104] In step S6, based on the load reference range determined in step S3, the current temperature signal is... Radiation intensity signal The amplitude is checked for load consistency. If it exceeds the reasonable amplitude range under the load, then... or Make corrections.

[0105] In step S6, the first threshold is 0.2, and the multiple stability thresholds include 0.3 and 0.6. The specific steps for determining the flame state are as follows:

[0106] like If the value is less than 0.2, it is determined that there is no flame and a no-flame alarm is triggered;

[0107] like If the value is ≥0.2, then according to the comprehensive stability index... Make a judgment;

[0108] like ≥0.2, and If the value is less than 0.3, it is considered severely unstable and a severe instability alarm (Level 1 alarm) is triggered.

[0109] If 0.3≤ If the value is less than 0.6, it is considered unstable and an instability alarm (level 2 alarm) is triggered.

[0110] like If the value is ≥0.6, it is considered to be in a stable state.

[0111] All judgment results are output to the DCS system through the human-machine interface or communication interface. In addition, the system also verifies the duration of the status. If the preset time is exceeded, a higher level alarm will be issued.

[0112] In step S6, the reasonable amplitude range is based on the historical benchmark average value corresponding to the load benchmark interval to which the current boiler load belongs. Standard deviation from the benchmark Confirmed, among which The lower limit of the reasonable amplitude is: The upper limit of the reasonable amplitude is ,in ;

[0113] The load consistency verification includes: determining the current temperature signal. or radiation intensity signal Whether it falls within their respective reasonable amplitude range;

[0114] If the verification results are inconsistent, the following correction operations will be performed:

[0115] If the current boiler load is greater than or equal to the preset load threshold, and or When the value is below its corresponding reasonable lower limit, the probability of flame existence will be... Force it to be less than 0.2 to trigger the no-flame state determination;

[0116] If the current boiler load is less than the preset load threshold, and or When the value exceeds its corresponding reasonable upper limit, the overall stability index will be adjusted. Multiply by a consistency weighting factor less than 1 ,in ,and This is a fixed constant calibrated based on historical operating data, and the corrected... Used for determining combustion status;

[0117] The preset load threshold is the percentage of boiler load that characterizes the switching point between high and low load conditions.

[0118] The preset load threshold can be set to 70% of BMCR. Choose 2, consistency weight factor The value is calibrated to 0.8. This value is determined by analyzing the impact of 20 low-load, abnormally high-radiation events over the past year on the stability index, resulting in a corrected value. The consistency with expert judgments has been significantly improved.

[0119] Load consistency verification logic:

[0120] High load scenario (Load ≥ preset load threshold, such as 70%): If or If the value is below the lower limit of the reasonable amplitude, it indicates a "high load, weak flame" state, and the flame is very likely to be shut down soon. Therefore, Forced to be less than 0.2, triggering the no-flame detection;

[0121] Low load scenario (Load < preset load threshold): If or A value exceeding the reasonable upper limit indicates a "low-load abnormally strong fire" phenomenon (such as rich fuel combustion) in the current combustion state. In this case, the comprehensive stability index will be... Multiply by consistency weight factor Make corrections, where 0 (If taken as 0.8, it can be dynamically calculated based on the degree of deviation) to reduce the overall stability index. This allows the modified overall stability index Sv to be used for subsequent combustion state determination, and prompts abnormal operating conditions to be determined as unstable or severely unstable states.

[0122] For example, the boiler load drops rapidly from 100% to 50%, but the pulverized coal is not reduced in time, resulting in a brief period of "rich fuel combustion" and an abnormally high flame temperature.

[0123] Temperature signal without load consistency verification Small fluctuations lead to a lower coefficient of variation. Low overall stability index The value is relatively high, which may lead to a misjudgment as a stable state;

[0124] During load consistency verification, if Load=50% is detected, but the current temperature signal... If the temperature of 1500℃ is higher than the corresponding reasonable upper limit of 1200℃, it is determined that the load and flame intensity are mismatched, and the overall stability index is reduced. This allows for the correct identification of an unstable state and triggers an early warning message.

[0125] An online monitoring system for combustion stability of a coal-fired power plant boiler based on multimodal signal fusion includes:

[0126] Intelligent flame spectral image detector, used to acquire radiation images and with built-in temperature inversion module;

[0127] The radiation intensity sensor and the radiation heat flux sensor output signals respectively. and ;

[0128] The load interface module is used to receive the boiler load signal (Load) from the DCS system.

[0129] An edge computing unit is configured with a processor and a memory, the memory storing a computer program that, when executed by the processor, implements the aforementioned monitoring method.

[0130] In this embodiment, the processor can be an ARM Cortex-A72 processor, and the operating system can be Linux. The edge computing unit synchronously acquires and processes signals from each channel at a sampling frequency of 20Hz, with a multi-channel synchronization error of less than 50ms.

[0131] The edge computing unit supports the Modbus TCP communication protocol and can interact with the power plant's DCS system and SIS system to upload monitoring results, output alarm information, and store historical data.

[0132] The alarm and visualization module is used to output the combustion status determination result and the probability of flame presence, and generate early warning information. It also supports data interaction with the DCS system.

[0133] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention is limited to these examples; within the framework of the invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of the different aspects of the invention as described above, which are not provided in detail for the sake of brevity.

[0134] The embodiments of this invention are intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for online monitoring of combustion stability of coal-fired power plant boilers based on multimodal signal fusion, characterized in that, Includes the following steps: S1. Synchronously acquire radiation images and radiation intensity signals of the combustion zone. and radiative heat flow signal And the boiler load signal, based on the radiation image, is used to invert the temperature signal through a radiation thermometry model. ; S2, Temperature signal Radiation intensity signal and radiative heat flow signal Savitzky-Golay filtering was performed separately, and then smoothed using an exponentially weighted moving average to obtain the preprocessed signal; S3. Determine the load reference range corresponding to the current operating condition based on the boiler load signal Load, and acquire the temperature signal within the load reference range. Radiation intensity signal and radiative heat flow signal Their respective historical baseline mean Standard deviation from the benchmark ,and The preprocessed signals are Z-score standardized to obtain standardized signals. , and ; S4. Within the sliding time window, calculate respectively , and coefficient of variation , and Each coefficient of variation is mapped to its corresponding single-mode stability index, and the single-mode stability indices are weighted and summed to obtain the comprehensive stability index. ; In step S4, the coefficient of variation is calculated for each standardized signal: ， ； in, This represents the average value. Indicates standard deviation; ； ； in, Represents the first normalized signal within the sliding time window. Each sample value, For sampling sequence number, This represents the number of sampling points within the sliding time window; Map each coefficient of variation to a single-mode stability index: ， ； in, This is an empirical attenuation coefficient used to adjust the stability's sensitivity to fluctuations; The overall stability index is obtained by weighted summation of the individual mode stability indices. ; ； in, , and These correspond to the combined weights of the temperature stability index, radiation intensity stability index, and radiation heat flux stability index, respectively. ,and ; S5, Based on absolute value of radiation intensity Absolute value of the rate of change of radiation intensity and absolute value of radiative heat flux The probability of flame presence is calculated using a pre-trained sigmoid function model. ; In step S5, the probability of the flame existing is... Calculated using a pre-trained Sigmoid function model: ； in, , , , For the Sigmoid function, , and The model parameters are obtained by training based on historical combustion sample data, and For feature fusion weights, The characteristic sensitivity coefficient, These are bias parameters; in, , ,and ; S6, if If the value is less than the first threshold, it is determined to be a flameless state; otherwise, it is determined according to the comprehensive stability index. The combustion state is classified into severely unstable, unstable, or stable based on the comparison results with multiple preset stability thresholds, and the corresponding warning information is output. In step S6, based on the load reference range determined in step S3, the current temperature signal is... Radiation intensity signal The amplitude is checked for load consistency. If it exceeds the reasonable amplitude range under the load, then... or Make corrections.

2. The method for online monitoring of combustion stability of coal-fired power plant boilers based on multimodal signal fusion according to claim 1, characterized in that, In step S1, the radiation thermometry model is constructed based on Planck's blackbody radiation law or the two-color thermometry method, and the temperature signal This is a time series of the average or highest temperature in the combustion zone.

3. The method for online monitoring of combustion stability of coal-fired power plant boilers based on multimodal signal fusion according to claim 1, characterized in that, In step S3, the load reference interval is defined by dividing the boiler load range into multiple discrete intervals, each of which has a pre-stored historical reference average value. Standard deviation from the benchmark .

4. The method for online monitoring of combustion stability of coal-fired power plant boilers based on multimodal signal fusion according to claim 1, characterized in that, In step S3, the preprocessed , , Perform Z-score standardization separately: ， 。 5. The method for online monitoring of combustion stability of coal-fired power plant boilers based on multimodal signal fusion according to claim 1, characterized in that, In step S6, the first threshold is 0.2, and the multiple stability thresholds include 0.3 and 0.

6. The specific steps for determining the flame state are as follows: like If the value is less than 0.2, it is determined that there is no flame and a no-flame alarm is triggered; like If the value is ≥0.2, then according to the comprehensive stability index... Make a judgment; like ≥0.2, and If the value is less than 0.3, it is considered severely unstable and a severe instability alarm is triggered. If 0.3≤ If the value is less than 0.6, it is considered unstable and an instability alarm is triggered. like If the value is ≥0.6, it is considered to be in a stable state.

6. The method for online monitoring of combustion stability of coal-fired power plant boilers based on multimodal signal fusion according to claim 5, characterized in that, In step S6, the reasonable amplitude range is based on the historical benchmark average value corresponding to the load benchmark interval to which the current boiler load belongs. Standard deviation from the benchmark Confirmed, among which The lower limit of the reasonable amplitude is: The upper limit of the reasonable amplitude is ,in ; The load consistency verification includes: determining the current temperature signal. or radiation intensity signal Whether it falls within their respective reasonable amplitude range; If the verification results are inconsistent, the following correction operations will be performed: If the current boiler load is greater than or equal to the preset load threshold, and or When the value is below its corresponding reasonable lower limit, the probability of flame existence will be... Force it to be less than 0.2 to trigger the no-flame state determination; If the current boiler load is less than the preset load threshold, and or When the value exceeds its corresponding reasonable upper limit, the overall stability index will be adjusted. Multiply by a consistency weighting factor less than 1 ,in ,and This is a fixed constant calibrated based on historical operating data, and the corrected... Used for determining combustion status; The preset load threshold is the percentage of boiler load that characterizes the switching point between high and low load conditions.

7. An online monitoring system for combustion stability of a coal-fired power plant boiler based on multimodal signal fusion, implementing the method as described in any one of claims 1-6, characterized in that, include: Intelligent flame spectral image detector, used to acquire radiation images and with built-in temperature inversion module; The radiation intensity sensor and the radiation heat flux sensor output signals respectively. and ; The load interface module is used to receive the boiler load signal (Load) from the DCS system. An edge computing unit is configured with a processor and a memory, the memory storing a computer program that, when executed by the processor, implements the method as described in any one of claims 1-6; The alarm and visualization module is used to output the combustion status determination result and the probability of flame presence, and generate early warning information. It also supports data interaction with the DCS system.

Images