Energy Consumption Diagnosis System and Method for Gas-fired Steam Boilers

Abstract

This invention relates to the field of boiler energy consumption diagnosis technology, specifically to a gas-fired steam boiler energy consumption diagnosis system and method, comprising: collecting furnace flame radiation spectrum data and flue gas component concentration data from the exhaust pipe during the boiler operation cycle; calculating the flame flicker frequency by extracting frequency domain features from the flame radiation spectrum data using a combustion stability analyzer; calculating the exhaust heat loss value from the flue gas component concentration data using a heat loss calculator based on the excess air coefficient and flue gas specific heat capacity; synchronously transmitting the flame flicker frequency to the heat loss calculator to correct the exhaust temperature calculation deviation; the heat loss calculator outputs a boiler operating efficiency benchmark value, which is compared with a preset theoretical energy efficiency threshold to generate an energy efficiency deviation vector; using this vector to drive a fault tracing engine to traverse a multi-dimensional fault feature database, identifying suspected faulty components and their corresponding confidence scores. This invention enables accurate diagnosis and fault tracing of boiler energy consumption.

Description

Technical Field

[0001] This invention relates to the field of boiler energy consumption diagnosis technology, and in particular to a gas-fired steam boiler energy consumption diagnosis system and method. Background Technology

[0002] Gas-fired steam boilers are widely used in industrial production and heating, and their energy consumption directly affects operating costs and environmental benefits. Therefore, energy consumption diagnosis is a crucial aspect of boiler operation and maintenance. Existing energy consumption diagnosis technologies for gas-fired steam boilers mostly rely on collecting basic parameters such as flue gas temperature and flue gas component concentration, combined with empirical formulas to calculate flue gas heat loss, and then assess boiler operating efficiency. Troubleshooting mainly depends on manual judgment of single fault characteristics or on pre-set simple fault correspondences.

[0003] Existing technologies, in calculating flue gas heat loss, rely solely on flue gas component concentration and excess air coefficient, neglecting the impact of combustion stability on flue gas temperature calculations. This leads to inaccuracies in the calculated flue gas heat loss values, resulting in inaccurate calculations of boiler operating efficiency benchmarks and an inability to accurately reflect the actual energy consumption of the boiler. In fault tracing, existing technologies lack a systematic multi-dimensional fault feature correlation and fail to establish a clear mapping relationship between various abnormal operating conditions and energy efficiency indicators. They rely solely on single fault characteristics for judgment, making it difficult to accurately pinpoint suspected faulty components and quantify the probability of faults, resulting in low diagnostic efficiency and a high misjudgment rate.

[0004] Currently, it is necessary to address the problem of inaccurate energy efficiency assessment caused by deviations in flue gas temperature calculation, as well as the lack of clear mapping relationships in fault tracing, the inability to accurately locate suspected faulty components, and the inability to quantify the probability of faults, in order to improve the accuracy and reliability of energy consumption diagnosis for gas-fired steam boilers. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing an energy consumption diagnosis system and method for gas-fired steam boilers.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for diagnosing the energy consumption of a gas-fired steam boiler, comprising: During the operation cycle of the gas-fired steam boiler, flame radiation spectrum data inside the furnace and flue gas component concentration data in the flue gas duct are collected. The flame radiation spectrum data is sent to a combustion stability analyzer for processing. The combustion stability analyzer calculates the flame flashing frequency by extracting the frequency domain features of the spectral data. The flue gas component concentration data is sent to a heat loss calculator for processing. The heat loss calculator calculates the exhaust heat loss value based on the excess air coefficient and the specific heat capacity of the flue gas. The flame flicker frequency data output by the combustion stability analyzer will be synchronously transmitted to the heat loss calculator to correct the calculation deviation of the exhaust gas temperature. The heat loss calculator outputs the current boiler operating efficiency baseline value; The boiler operating efficiency benchmark value is compared with the preset theoretical energy efficiency threshold to generate an energy efficiency deviation vector; The energy efficiency deviation vector is used as input to drive the fault tracing engine to traverse the multi-dimensional fault feature library, which records the mapping relationship between various abnormal operating conditions and energy efficiency indicators. The fault tracing engine identifies several suspected faulty components and their corresponding confidence scores based on the traversal results.

[0007] As a further aspect of the present invention, the flame radiation spectrum data is fed into a combustion stability analyzer for processing. The combustion stability analyzer calculates the flame flicker frequency by extracting the frequency domain features of the spectral data, including: The flame radiation spectrum data is subjected to bandpass filtering to remove infrared interference noise from the environment; The filtered spectral data is converted into a grayscale image sequence, which reflects the changes in the transient morphology of the flame; The Fourier transform algorithm is applied to the grayscale image sequence to extract the spectral curve of the flame brightness changing over time; Find the characteristic peak with the largest amplitude on the spectrum curve, and determine the frequency corresponding to the characteristic peak as the flame flashing frequency; A combustion fluctuation coefficient is obtained by comparing the flame flicker frequency with the standard frequency range under the boiler's rated load.

[0008] As a further aspect of the present invention, the flue gas component concentration data is sent to a heat loss calculator for processing. The heat loss calculator calculates the exhaust heat loss value based on the excess air coefficient and the specific heat capacity of the flue gas, including: The volume fractions of oxygen and carbon monoxide were extracted from the flue gas component concentration data. Based on the volume fraction of oxygen, the excess air coefficient is derived by using the conversion relationship between oxygen content and air coefficient. Collect real-time exhaust temperature data in the exhaust duct and calculate the exhaust temperature difference by combining it with ambient temperature data; The actual exhaust heat loss value is calculated based on the excess air coefficient, exhaust temperature difference, and preset average specific heat capacity of the flue gas. The combustion fluctuation coefficient is incorporated into the calculation process of the exhaust heat loss value. When the combustion fluctuation coefficient exceeds the warning value, the exhaust temperature data is smoothed and noise-reduced.

[0009] As a further aspect of the present invention, the heat loss calculator outputs the current boiler operating efficiency benchmark value, including: Simultaneously collect boiler feedwater flow rate data and steam outlet pressure data to calculate the effective heat absorption per unit time; Dividing the effective heat absorption by the sum of the fuel input heat and the exhaust heat loss yields a preliminary efficiency calculation result. The preliminary efficiency calculation results are weighted and fused with the combustion fluctuation coefficient to generate the boiler operating efficiency benchmark value; The boiler operating efficiency benchmark value, along with the current excess air coefficient, is packaged and sent to the fault tracing engine.

[0010] As a further aspect of the present invention, the boiler operating efficiency benchmark value is compared with a preset theoretical energy efficiency threshold to generate an energy efficiency deviation vector, including: Retrieve the design thermal parameter table corresponding to the current boiler model and read the preset theoretical energy efficiency threshold from it; The difference between the boiler operating efficiency benchmark value and the preset theoretical energy efficiency threshold is calculated and defined as the instantaneous energy efficiency gap; Collect the current operating load rate data of the boiler, and normalize the instantaneous energy efficiency gap with the operating load rate data; The normalized data is combined into a multidimensional array, which is the energy efficiency deviation vector. The energy efficiency deviation vector is sent to the fault tracing engine as a retrieval key.

[0011] As a further aspect of the present invention, the energy efficiency deviation vector is used as input to drive the fault tracing engine to traverse a multi-dimensional fault feature library. This multi-dimensional fault feature library records the mapping relationship between various abnormal operating conditions and energy efficiency indicators, including: The energy efficiency deviation vector is mapped onto the coordinate system of the multidimensional fault feature library to determine its position in the multidimensional space; Retrieve several historical fault case nodes that are adjacent to the location in the multidimensional fault feature database; Calculate the Euclidean distance between the energy efficiency deviation vector and each historical fault case node, and select the nodes with the closest distance. Read the fault labels and feature weights corresponding to the selected nodes, and summarize them to generate the confidence score; The fault label with the highest confidence score is marked as the name of the suspected faulty component.

[0012] As a further aspect of the present invention, it also includes: For the suspected faulty component with the highest confidence score, a maintenance procedure template associated with the suspected faulty component is retrieved. The maintenance procedure template defines the disassembly steps and measurement points. According to the maintenance procedure template, disassembly and measurement operations are performed on the suspected faulty component on site, and the measured physical parameters are recorded. The measured physical parameters will be sent back to update the multidimensional fault feature database. For the suspected faulty component with the highest confidence score, a maintenance procedure template associated with the suspected faulty component is retrieved. This maintenance procedure template defines the disassembly steps and measurement points, including: Based on the name of the suspected faulty component, access an equipment asset management database to obtain the factory serial number and assembly drawing version number of the suspected faulty component; Based on the factory serial number and assembly drawing version number, download the corresponding maintenance procedure template from the cloud server; The structured data in the maintenance procedure template is analyzed to extract the standard torque parameters and sealing surface inspection standards for suspected faulty components. The standard torque parameters and sealing surface inspection standards are compiled into a work instruction card that can be executed on a mobile terminal, and the work instruction card is distributed to the handheld terminal of the on-site maintenance personnel.

[0013] As a further aspect of the present invention, according to the maintenance procedure template, disassembly and measurement operations are performed on the suspected faulty component on-site, and the measured physical parameters are recorded. These measured physical parameters will be fed back to update the multi-dimensional fault feature database, including: According to the work instruction card, the on-site maintenance personnel disassembled the suspected faulty component and used measuring tools to measure the geometric dimensions of its key parts; Record the measured values ​​of the geometric dimensions and photos of the wear marks on the components, and package the information into a measured data package; The measured data packets are uploaded to the data center via a wireless network and associated with the previous energy efficiency deviation vector; In the database of the data center, the feature threshold range of the corresponding fault label in the multidimensional fault feature library is adjusted according to the numerical range of the measured data packet. After the adjustment is completed, the suspected faulty component is reassembled back into the boiler system.

[0014] As a further aspect of the present invention, before retrieving several historical fault case nodes adjacent to the location in the multi-dimensional fault feature database, a database self-cleaning step is also included: Periodically scan all historical fault case nodes in the multidimensional fault feature library to identify obsolete nodes that have not been called for a long time; Retrieve the original operation logs associated with the old nodes and check for any misjudgment records caused by sensor drift; If there are misjudged records, the old nodes are removed, and the gaps are filled with recent correct case nodes. The purified multidimensional fault feature library index is then reloaded into the memory of the fault tracing engine.

[0015] As a further aspect of the present invention, the present invention includes a gas-fired steam boiler energy consumption diagnosis system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the gas-fired steam boiler energy consumption diagnosis method described above.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: Frequency domain feature extraction is performed on the flame radiation spectrum data inside the furnace to calculate the flame flicker frequency. This flame flicker frequency data is then synchronously transmitted to the heat loss calculator to correct the calculation deviation of the flue gas temperature. The flame flicker frequency obtained through frequency domain feature extraction can accurately reflect the combustion stability state. Using it to correct the calculation deviation of the flue gas temperature can compensate for the deficiency of conventional technology in ignoring the influence of combustion stability on the flue gas temperature, reduce the calculation deviation of the flue gas heat loss value, make the output boiler operating efficiency benchmark value more consistent with the actual operating conditions of the boiler, reduce the error of energy efficiency assessment, and make the energy consumption diagnosis results more realistic and referential.

[0017] The boiler operating efficiency benchmark value output by the heat loss calculator is compared with the preset theoretical energy efficiency threshold to generate an energy efficiency deviation vector. This deviation vector is then used as input to drive the fault tracing engine to traverse a multi-dimensional fault feature library. This library records the mapping relationship between various abnormal operating conditions and energy efficiency indicators, ultimately identifying several suspected faulty components and their corresponding confidence scores. By constructing a bridge between energy efficiency status and fault characteristics through the energy efficiency deviation vector, combined with the traversal analysis of the multi-dimensional fault feature library, the limitations of single fault feature judgment in conventional technologies are overcome. This achieves a precise correlation between abnormal operating conditions and energy efficiency indicators, enabling a more comprehensive investigation of potential faults. Simultaneously, the confidence scores can distinguish the probability of failure of different suspected faulty components, reducing ineffective investigation steps and improving the accuracy and efficiency of fault tracing. Attached Figure Description

[0018] Figure 1 This is a flowchart of the energy consumption diagnosis method for gas-fired steam boilers described in this invention; Figure 2 A flowchart for calculating and correcting flue gas heat loss values; Figure 3 The graph shows the effect of combustion fluctuation coefficient on heat loss and efficiency. Figure 4 A bar chart showing the confidence score for the fault tracing stage of a gas-fired steam boiler; Figure 5 This is a time-series graph showing the change in the number of nodes in the multidimensional fault feature library. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1The energy consumption diagnosis method for gas-fired steam boilers is achieved through a sensor network and data analysis system deployed in the boiler system. The sensor network includes a flame spectral sensor installed on the furnace wall and a flue gas analyzer installed in the exhaust duct. The data analysis system consists of a combustion stability analyzer, a heat loss calculator, and a fault tracing engine. During a complete boiler operation cycle, the flame spectral sensor continuously collects flame radiation spectrum data, and the flue gas analyzer simultaneously collects flue gas component concentration data. This data is transmitted to the data analysis system in real time via an industrial bus. The combustion stability analyzer processes the flame radiation spectrum data, extracting frequency domain features to calculate the flame flicker frequency, which characterizes the combustion state. The heat loss calculator processes the flue gas component concentration data, calculating the exhaust heat loss value based on the excess air coefficient and the specific heat capacity of the flue gas. The flame flicker frequency data output by the combustion stability analyzer is synchronously transmitted to the heat loss calculator to correct for exhaust temperature measurement deviations caused by combustion fluctuations. The heat loss calculator integrates various inputs and outputs a corrected benchmark value for boiler operating efficiency. This benchmark value is then compared with a preset theoretical energy efficiency threshold, and the difference is normalized to generate a multi-dimensional energy efficiency deviation vector. This energy efficiency deviation vector serves as input to drive the fault tracing engine. This engine traverses its built-in multi-dimensional fault feature library, which pre-stores the mapping relationship between various abnormal operating conditions and energy efficiency indicators. Based on the traversal results, the fault tracing engine ultimately identifies one or more suspected faulty components and calculates a confidence score for each component, thus completing the diagnostic process from data acquisition to fault location.

[0022] In one embodiment of the invention, a flame spectral sensor acquires raw flame spectral data at a rate of 100 frames per second. The raw flame spectral data is a continuous spectral signal containing wavelengths ranging from 400 nanometers to 900 nanometers. In a specific implementation, a combustion stability analyzer performs bandpass filtering on the raw flame spectral data. The purpose of bandpass filtering is to filter out infrared thermal radiation interference noise with wavelengths greater than 850 nanometers generated by the radiation of high-temperature refractory materials in the furnace, while retaining the characteristic spectral signals of the flame itself in the visible and near-infrared bands. The filtered spectral data is then sent to a subsequent image conversion module.

[0023] In some embodiments, the step of converting the filtered spectral data into a grayscale image sequence specifically involves converting the complete spectral data acquired at each sampling time into a single pixel grayscale value according to a preset wavelength-intensity weighted mapping relationship. This grayscale value represents the overall relative brightness of the flame at that moment. Arranging the grayscale values ​​from multiple consecutive sampling times in chronological order generates a one-dimensional grayscale image sequence, which reflects the transient shape of the flame brightness changing over time. In some embodiments, applying a Fourier transform algorithm to the grayscale image sequence to extract the spectral curve of the flame brightness changing over time involves inputting a discrete grayscale value sequence as a time-domain signal and calculating its frequency component distribution using a fast Fourier transform algorithm. The fast Fourier transform algorithm outputs a complex array. By performing a modulo operation on the complex array, the corresponding amplitude-frequency response curve can be obtained. This amplitude-frequency response curve is the spectral curve of the flame brightness changing over time.

[0024] In practical implementation, the process of finding the characteristic peak with the largest amplitude on the spectrum curve and determining the frequency corresponding to the characteristic peak as the flame flicker frequency is automatically completed by the peak detection algorithm within the combustion stability analyzer. The peak detection algorithm traverses all data points on the spectrum curve, identifies local maxima points whose amplitudes are greater than several adjacent points, and selects the point with the largest amplitude from all local maxima points. The abscissa value of this point, i.e., the frequency value, is directly output as the flame flicker frequency. The process of comparing the flame flicker frequency with the standard frequency range under the boiler's rated load to calculate the combustion fluctuation coefficient can be expressed by the following formula:

[0025] in: This represents the real-time flame flicker frequency obtained from spectrum analysis. This represents the reference frequency value obtained from a standard database under the current boiler operating load. The half-width represents the standard frequency range. This is the calculated combustion fluctuation coefficient. Optionally, during the comparison process, if the real-time flame flicker frequency falls within the standard frequency range, the combustion fluctuation coefficient will take a small value close to zero, indicating a stable combustion state. If the real-time flame flicker frequency deviates from the standard range, the value of the combustion fluctuation coefficient will increase accordingly, and its magnitude directly quantifies the degree of combustion instability.

[0026] In one embodiment of the present invention, see [reference] Figure 2The flue gas analyzer measures and transmits flue gas component concentration data in real time, and the heat loss calculator parses the volume fractions of oxygen and carbon monoxide from the received data. In some embodiments, the process of deducing the excess air coefficient based on the oxygen volume fraction is based on the chemical reaction equation for complete fuel combustion, combined with the known fuel elemental composition, and using the theoretical relationship between oxygen and air volume for calculation. The calculation process involves querying fuel characteristic parameters, and the specific value of the excess air coefficient reflects the ratio between the actual amount of air entering the furnace and the theoretically required amount of air.

[0027] In practical implementation, the heat loss calculator simultaneously collects real-time flue gas temperature data and ambient temperature data from the flue gas duct. The real-time flue gas temperature data is measured by thermocouples installed in the flue, while the ambient temperature data is provided by temperature sensors in the boiler room. The difference between the two constitutes the flue gas temperature difference. The heat loss calculator calculates the flue gas heat loss value based on the excess air coefficient, the flue gas temperature difference, and a preset average specific heat capacity of the flue gas. The preset average specific heat capacity of the flue gas is an empirical constant related to the fuel type and the excess air coefficient, stored in the parameter database of the heat loss calculator. In practical implementation, the combustion fluctuation coefficient from the combustion stability analyzer is incorporated into the calculation process of the flue gas heat loss value. When the combustion fluctuation coefficient exceeds a preset warning value, for example, if the combustion fluctuation coefficient is greater than 0.1, the heat loss calculator performs smoothing and noise reduction processing on the real-time collected flue gas temperature data. The smoothing and noise reduction processing uses a moving average filtering algorithm to eliminate the impact of instantaneous peaks or troughs in flue gas temperature caused by severe combustion fluctuations on the accuracy of the heat loss calculation.

[0028] In some embodiments, the calculation of the boiler operating efficiency benchmark value requires the simultaneous acquisition of boiler feedwater flow rate data and steam outlet pressure data. The boiler feedwater flow rate data is provided by an electromagnetic flowmeter installed on the feedwater pipeline, and the steam outlet pressure data is provided by a pressure transmitter installed on the main steam pipeline. It is understood that calculating the effective heat absorption per unit time requires combining the feedwater flow rate and the enthalpy difference between steam and feedwater. The enthalpy of steam and feedwater is obtained by consulting a steam property table, while the steam outlet pressure is one of the key parameters for determining the steam enthalpy. It is understood that the heat loss calculator divides the calculated effective heat absorption by the sum of the fuel input heat and the corrected flue gas heat loss value to obtain a preliminary efficiency calculation result. The fuel input heat is obtained by multiplying the fuel flow rate measured by the fuel flowmeter by the lower heating value of the fuel. In a specific implementation, the preliminary efficiency calculation result is weighted and fused with the combustion fluctuation coefficient to generate the boiler operating efficiency benchmark value. The weighted fusion uses a linear interpolation method, where the combustion fluctuation coefficient is used as a weighting factor in the calculation. The formula for weighted fusion is expressed as:

[0029] in: This represents the final generated baseline value for boiler operating efficiency. This represents the boiler efficiency result obtained from preliminary calculations. This represents a baseline efficiency empirical value set based on combustion conditions. This represents the weighting factor derived from the combustion fluctuation coefficient mapping. The value ranges from zero to one. Optionally, when the combustion fluctuation coefficient is very small, the weighting factor... Approaching zero, the boiler operating efficiency benchmark value It mainly depends on the preliminary efficiency calculation results. Optionally, when the combustion fluctuation coefficient is large, the weighting factor... Increase the boiler operating efficiency benchmark value It will move towards the baseline efficiency empirical value The direction was adjusted to smooth out the drastic jumps in efficiency calculation results caused by intense combustion fluctuations.

[0030] In one embodiment of the present invention, the fault tracing engine retrieves the design thermodynamic parameter table corresponding to the current boiler model. This table is stored in a database and records the theoretical efficiency design values ​​of the boiler under different load rates. The fault tracing engine matches and reads a preset theoretical energy efficiency threshold based on the current operating conditions. In some embodiments, the fault tracing engine calculates the difference between the boiler operating efficiency benchmark value and the preset theoretical energy efficiency threshold, defining this difference as the instantaneous energy efficiency gap. For example, if the boiler operating efficiency benchmark value is 88% and the preset theoretical energy efficiency threshold is 91.5%, then the instantaneous energy efficiency gap is -3.5%. In a specific implementation, the fault tracing engine simultaneously collects the current boiler operating load rate data, provided by the boiler main controller. The load rate data represents the percentage of the boiler's actual evaporation to its rated evaporation. The fault tracing engine normalizes the instantaneous energy efficiency gap and the operating load rate data using a minimum-maximum scaling method, converting two physical quantities with different dimensions into a numerical range of 0 to 1. It can be understood that the normalized data is combined into a multidimensional array containing two elements, which is the energy efficiency deviation vector. For example, if the normalized instantaneous energy efficiency deficit is 0.65 and the normalized operating load rate is 0.8, then the energy efficiency deviation vector is represented as [0.65, 0.8]. In specific implementation, the fault tracing engine uses the energy efficiency deviation vector as input to drive it to traverse the built-in multidimensional fault feature library. The multidimensional fault feature library stores historical fault cases in the form of a high-dimensional vector space. Each historical fault case node has a coordinate position in the vector space composed of energy efficiency features. It can be understood that the fault tracing engine maps the energy efficiency deviation vector to the coordinate system of the multidimensional fault feature library, determining the specific position of the energy efficiency deviation vector in the multidimensional space composed of the efficiency deviation dimension and the load rate dimension.

[0031] In some embodiments, the fault tracing engine retrieves several historical fault case nodes adjacent to the energy efficiency deviation vector in a multi-dimensional fault feature database. The retrieval process employs the K-nearest neighbor algorithm, calculating the geometric distance between the energy efficiency deviation vector and all nodes in the feature database, centered on the coordinates of the energy efficiency deviation vector. The fault tracing engine then calculates the Euclidean distance between the energy efficiency deviation vector and each historical fault case node. The formula for calculating the Euclidean distance is:

[0032] in: and These represent the first and second elements of the energy efficiency deviation vector, namely the normalized efficiency deviation and load factor, respectively. and These represent the corresponding coordinate values ​​of a historical fault case node in the multidimensional fault feature database. This refers to the calculated Euclidean distance. Optionally, the fault tracing engine selects the 5 or 10 nodes with the smallest distances from all the calculated Euclidean distances as the nearest neighbor set. Optionally, the fault tracing engine reads the fault labels and feature weights corresponding to the selected nodes. Fault labels are, for example, "air preheater blockage" or "burner air distribution imbalance," and feature weights represent the significance of the fault feature in historical diagnoses. The fault tracing engine summarizes and generates a confidence score. The confidence score is inversely proportional to the Euclidean distance and directly proportional to the feature weights, specifically implemented through a weighted average algorithm.

[0033] See Figure 3 This graph illustrates the impact of combustion fluctuation coefficient on heat loss and efficiency in a gas-fired steam boiler, showing the relationship between combustion fluctuation coefficient and flue gas heat loss and boiler operating efficiency. Flue gas heat loss generally increases with increasing combustion fluctuation coefficient, gradually rising from approximately 2.7% to approximately 5.8%, indicating that the more unstable the combustion, the more heat is carried away by the flue gas. Boiler operating efficiency generally decreases with increasing combustion fluctuation coefficient, gradually decreasing from approximately 88.6% to approximately 85.9%, indicating that combustion fluctuation directly reduces boiler energy efficiency. The two curves show a clear inverse relationship. When the combustion fluctuation coefficient is <0.3, the boiler efficiency remains above 87.7%, and the flue gas heat loss is controlled below 3.8%, belonging to a stable and efficient operating range. When the combustion fluctuation coefficient is >0.7, the flue gas heat loss exceeds 5.0%, and the boiler efficiency drops below 86.5%, showing a significant energy efficiency decline, belonging to a high-loss and low-efficiency range.

[0034] In one embodiment of the invention, the diagnostic system accesses the equipment asset management database based on the name of the suspected faulty component, "air preheater," using the component name "air preheater" as the query condition. The query returns the specific factory serial number and assembly drawing version number of the air preheater in the database. For example, the factory serial number is APH-2023-0876, and the assembly drawing version number is Rev.2.1. In some embodiments, based on the factory serial number APH-2023-0876 and the assembly drawing version number Rev.2.1, the diagnostic system generates a unique retrieval identifier and downloads a maintenance procedure template that perfectly matches the retrieval identifier from the maintenance procedure library on a cloud server. The maintenance procedure template is a structured data file containing disassembly steps, measurement points, and safety precautions. In a specific implementation, the diagnostic system parses the structured data in the maintenance procedure template and extracts key assembly parameters for the air preheater. These key assembly parameters include standard torque parameters and sealing surface inspection standards, which may include flatness requirements and cleanliness requirements. Understandably, the diagnostic system compiles the extracted standard torque parameters and sealing surface inspection standards into a work instruction card that can be executed on a mobile terminal. The work instruction card uses a mixed text and graphics interface; for example, it highlights the bolt torque value "120 N·m" and lists the sealing surface inspection standard "flatness ≤ 0.1 mm" as an inspection item. In practice, the diagnostic system distributes the generated work instruction card to the handheld terminal of the on-site maintenance personnel via a wireless network. The handheld terminal runs a dedicated maintenance client program to receive and display the contents of the work instruction card.

[0035] Following the illustrated instructions on the work instruction card on their handheld terminals, on-site maintenance personnel disassembled the air preheater, which was marked as a suspected faulty component. The disassembly process followed the steps defined in the work instruction card, such as first closing the flue gas valve, then disconnecting the connecting pipes, and finally removing the fixing bolts. After on-site disassembly, maintenance personnel used measuring tools such as calipers, feeler gauges, and dial indicators to measure the geometric dimensions of key components specified in the work instruction card. These key geometric dimensions included the outer diameter of the heat exchanger tube bundle, the spacing of the tube sheets, and the thickness of the sealing flanges. The maintenance personnel recorded the measured values ​​of the geometric dimensions and simultaneously used the built-in camera of the handheld terminal to take photos of dust accumulation, tube bundle corrosion, or wear marks on the sealing surfaces inside the air preheater. The measured values ​​and wear mark photos were automatically packaged into a measurement data packet. In some embodiments, the measurement data packet is uploaded to the data center via the on-site wireless network. The data center's receiving service program associates the measurement data packet with the original energy efficiency deviation vector that triggered this maintenance task. This association is achieved by establishing the same transaction serial number in the database. In practice, the data center's backend system adjusts the feature threshold range of the corresponding fault labels in the multidimensional fault feature library based on the numerical range of the measured data packets. The adjustment method involves Bayesian updating of historical statistical intervals using measured data. The update logic of the multidimensional fault feature library can be described by the following formula:

[0036] in: This represents the upper limit of the updated feature threshold range. This represents the upper limit of the feature threshold range before the update. This represents the maximum value corresponding to the actual measured data in this field. The confidence weight coefficient represents historical data, with a value between 0 and 1. Optionally, the update process also applies to the lower limit of the feature threshold interval. Optionally, after completing the online adjustment of the multi-dimensional fault feature library, the diagnostic system sends instructions to the handheld terminal, directing on-site maintenance personnel to reassemble the air preheater back into the boiler system according to the reverse sequence steps defined in the work instruction card. Refer to Table 1 for examples of key dimensions recorded during disassembly and measurement.

[0037] Table 1: Example Table of Measurement Data for Key Components of Air Preheater Measurement location Standard value (mm) Measured value (mm) Permissible deviation (mm) outer diameter of heat exchange tube bundle 40.0 39.8 ±0.5 Tube sheet spacing 1200.0 1201.5 ±2.0 Sealing flange thickness 25.0 24.7 ±0.5 See Figure 4This is a bar chart showing the confidence score during the fault tracing phase of a gas-fired steam boiler. It quantifies the degree of matching between each suspected faulty component and the current energy efficiency deviation vector. The air preheater has the highest confidence score, significantly higher than other components, making it the most likely cause of the current low boiler energy efficiency. The burner and economizer have the next highest scores and should be considered as secondary targets for investigation to verify whether there are compound faults. The flue damper and feedwater preheater have lower scores and can be temporarily excluded from priority investigation. Prioritizing the disassembly, unblocking, and sealing surface inspection of the air preheater is the most likely way to resolve the current energy efficiency problem. The confidence score, based on the K-nearest neighbor algorithm and Euclidean distance calculation, reflects the degree of matching between fault characteristics and current operating conditions, providing a quantitative basis for diagnosis. This score result can be linked with data such as flame flicker frequency and flue gas heat loss to further pinpoint the root cause of the fault.

[0038] In one embodiment of the present invention, the database self-cleaning step is initiated by the background management module of the fault tracing engine, which periodically performs a full scan of the multi-dimensional fault feature database. Specifically, it periodically scans all historical fault case nodes in the multi-dimensional fault feature database to identify stale nodes that have not been accessed for a long time. The identification process relies on the built-in timestamp attribute of each historical fault case node, which records the time when the node was last retrieved by the fault tracing engine during the diagnostic process. In some embodiments, the criterion for identifying stale nodes is based on a preset time threshold; for example, if a historical fault case node has not been accessed within the last 365 days, the node is marked as a stale node. It is understood that the system retrieves the original operating logs associated with the stale nodes. These original operating logs are stored in a separate log database and associated with the stale nodes through the unique number of the fault case. Specifically, it checks the original operating logs associated with the stale nodes to see if there are any misjudgment records caused by sensor drift. The check includes reviewing the consistency of the calibration status markers, signal quality indices, and readings of other associated sensors recorded in the logs during the fault occurrence period. For example, examining the logs associated with an old node marked as "fuel flow meter failure" might reveal a persistent zero drift in the fuel flow meter signal at the time, while other parameters such as oxygen content and flue gas temperature remained normal. This suggests that the node may have originated from a misjudgment caused by sensor drift. In some embodiments, if a misjudgment record exists, the old node is removed. The removal operation involves removing the node's coordinates and feature vector from the active index of the multidimensional fault feature library and moving it to an archive area, while retaining its historical records for auditing purposes.

[0039] In practice, after removing outdated nodes, the system fills the gaps using recent correct case nodes. The goal of this gap-filling is to maintain the data distribution density of the multi-dimensional fault feature library in the vector space. The gap-filling method involves selecting spatially adjacent nodes from the recent correct case node set, performing feature interpolation, and generating new nodes to cover the sparse areas that may have been created by removing outdated nodes. The gap-filling algorithm can be described by the following formula:

[0040] in: This represents the feature vector of the new node generated to fill the gap. This represents the number of recently correct case nodes within a preset radius surrounding the stale nodes that were removed in the feature space. Representing the The feature vectors of the nearest recently correct case nodes. Optionally, recently correct case nodes refer to the nodes corresponding to cases whose diagnoses have been verified and confirmed as correct by on-site disassembly within the past 90 days. Optionally, after the purification and update process is completed, the system will reload the full data index of the purified multi-dimensional fault feature library from disk into the memory of the fault tracing engine to ensure that all subsequent diagnostic queries are based on the latest, purified feature library. The memory loading operation is usually performed automatically during periods of low system load to reduce the impact on online diagnostic services.

[0041] See Figure 5 This is a time-series chart showing the changes in the number of nodes in a multi-dimensional fault feature database, used to illustrate the changes in the number of historical fault nodes and the process of purging outdated nodes in the boiler fault diagnosis system. From March 13, 2024 to March 7, 2025, the number of nodes fluctuated between 10 and 35, showing a slow downward trend, reflecting the gradual accumulation of outdated nodes that had not been used for a long time, diluting the effective sample size. On March 7, 2025, the system performed database self-purification, removing outdated and misjudged nodes. After purification, the number of nodes remained in the range of 10 to 35, indicating that the purification operation mainly focused on "replacing outdated samples" and did not significantly reduce the total sample size. From March 8, 2025 to November 2, 2025, the number of nodes remained stable, and the sample database entered a stable update period, providing reliable basic data for fault tracing. After November 3, 2025, the number of nodes rapidly climbed to a peak of nearly 80, then fell back and remained in the range of 30-60, indicating that the system added a large number of new fault cases, the sample library was significantly expanded, and the diagnostic capabilities were further enhanced.

[0042] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for diagnosing the energy consumption of a gas-fired steam boiler, characterized in that, The method includes: During the operation cycle of the gas-fired steam boiler, flame radiation spectrum data inside the furnace and flue gas component concentration data in the flue gas duct are collected. The flame radiation spectrum data is sent to a combustion stability analyzer for processing. The combustion stability analyzer calculates the flame flashing frequency by extracting the frequency domain features of the spectral data. The flue gas component concentration data is sent to a heat loss calculator for processing. The heat loss calculator calculates the exhaust heat loss value based on the excess air coefficient and the specific heat capacity of the flue gas. The flame flicker frequency data output by the combustion stability analyzer will be synchronously transmitted to the heat loss calculator to correct the calculation deviation of the exhaust gas temperature. The heat loss calculator outputs the current boiler operating efficiency baseline value; The boiler operating efficiency benchmark value is compared with the preset theoretical energy efficiency threshold to generate an energy efficiency deviation vector; The energy efficiency deviation vector is used as input to drive the fault tracing engine to traverse the multi-dimensional fault feature library, which records the mapping relationship between various abnormal operating conditions and energy efficiency indicators. The fault tracing engine identifies several suspected faulty components and their corresponding confidence scores based on the traversal results.

2. The energy consumption diagnosis method for a gas-fired steam boiler as described in claim 1, characterized in that, The flame radiation spectrum data is fed into a combustion stability analyzer for processing. The combustion stability analyzer calculates the flame flashing frequency by extracting the frequency domain features of the spectral data, including: The flame radiation spectrum data is subjected to bandpass filtering to remove infrared interference noise from the environment; The filtered spectral data is converted into a grayscale image sequence, which reflects the changes in the transient morphology of the flame; The Fourier transform algorithm is applied to the grayscale image sequence to extract the spectral curve of the flame brightness changing over time; Find the characteristic peak with the largest amplitude on the spectrum curve, and determine the frequency corresponding to the characteristic peak as the flame flashing frequency; A combustion fluctuation coefficient is obtained by comparing the flame flicker frequency with the standard frequency range under the boiler's rated load.

3. The energy consumption diagnosis method for a gas-fired steam boiler as described in claim 2, characterized in that, The flue gas component concentration data is sent to a heat loss calculator for processing. The heat loss calculator calculates the exhaust heat loss value based on the excess air coefficient and the specific heat capacity of the flue gas, including: The volume fractions of oxygen and carbon monoxide were extracted from the flue gas component concentration data. Based on the volume fraction of oxygen, the excess air coefficient is derived by using the conversion relationship between oxygen content and air coefficient. Collect real-time exhaust temperature data in the exhaust duct and calculate the exhaust temperature difference by combining it with ambient temperature data; The actual exhaust heat loss value is calculated based on the excess air coefficient, exhaust temperature difference, and preset average specific heat capacity of the flue gas. The combustion fluctuation coefficient is incorporated into the calculation process of the exhaust heat loss value. When the combustion fluctuation coefficient exceeds the warning value, the exhaust temperature data is smoothed and noise-reduced.

4. The energy consumption diagnosis method for a gas-fired steam boiler as described in claim 3, characterized in that, The heat loss calculator outputs the current boiler operating efficiency baseline value, including: Simultaneously collect boiler feedwater flow rate data and steam outlet pressure data to calculate the effective heat absorption per unit time; Dividing the effective heat absorption by the sum of the fuel input heat and the exhaust heat loss yields a preliminary efficiency calculation result. The preliminary efficiency calculation results are weighted and fused with the combustion fluctuation coefficient to generate the boiler operating efficiency benchmark value; The boiler operating efficiency benchmark value, along with the current excess air coefficient, is packaged and sent to the fault tracing engine.

5. The energy consumption diagnosis method for a gas-fired steam boiler as described in claim 4, characterized in that, The boiler operating efficiency benchmark value is compared with a preset theoretical energy efficiency threshold to generate an energy efficiency deviation vector, including: Retrieve the design thermal parameter table corresponding to the current boiler model and read the preset theoretical energy efficiency threshold from it; The difference between the boiler operating efficiency benchmark value and the preset theoretical energy efficiency threshold is calculated and defined as the instantaneous energy efficiency gap; Collect the current operating load rate data of the boiler, and normalize the instantaneous energy efficiency gap with the operating load rate data; The normalized data is combined into a multidimensional array, which is the energy efficiency deviation vector. The energy efficiency deviation vector is sent to the fault tracing engine as a retrieval key.

6. The energy consumption diagnosis method for a gas-fired steam boiler as described in claim 5, characterized in that, The energy efficiency deviation vector is used as input to drive the fault tracing engine to traverse a multi-dimensional fault feature library. This multi-dimensional fault feature library records the mapping relationship between various abnormal operating conditions and energy efficiency indicators, including: The energy efficiency deviation vector is mapped onto the coordinate system of the multidimensional fault feature library to determine its position in the multidimensional space; Retrieve several historical fault case nodes that are adjacent to the location in the multidimensional fault feature database; Calculate the Euclidean distance between the energy efficiency deviation vector and each historical fault case node, and select the nodes with the closest distance. Read the fault labels and feature weights corresponding to the selected nodes, and summarize them to generate the confidence score; The fault label with the highest confidence score is marked as the name of the suspected faulty component.

7. The energy consumption diagnosis method for a gas-fired steam boiler as described in claim 6, characterized in that, Also includes: For the suspected faulty component with the highest confidence score, a maintenance procedure template associated with the suspected faulty component is retrieved. The maintenance procedure template defines the disassembly steps and measurement points. According to the maintenance procedure template, disassembly and measurement operations are performed on the suspected faulty component on site, and the measured physical parameters are recorded. The measured physical parameters will be sent back to update the multidimensional fault feature database. For the suspected faulty component with the highest confidence score, a maintenance procedure template associated with the suspected faulty component is retrieved. This maintenance procedure template defines the disassembly steps and measurement points, including: Based on the name of the suspected faulty component, access an equipment asset management database to obtain the factory serial number and assembly drawing version number of the suspected faulty component; Based on the factory serial number and assembly drawing version number, download the corresponding maintenance procedure template from the cloud server; The structured data in the maintenance procedure template is analyzed to extract the standard torque parameters and sealing surface inspection standards for suspected faulty components. The standard torque parameters and sealing surface inspection standards are compiled into a work instruction card that can be executed on a mobile terminal, and the work instruction card is distributed to the handheld terminal of the on-site maintenance personnel.

8. The energy consumption diagnosis method for a gas-fired steam boiler as described in claim 7, characterized in that, According to the aforementioned maintenance procedure template, disassembly and measurement operations are performed on the suspected faulty component on-site, and the measured physical parameters are recorded. These measured physical parameters will be fed back to update the multi-dimensional fault feature database, including: According to the work instruction card, the on-site maintenance personnel disassembled the suspected faulty component and used measuring tools to measure the geometric dimensions of its key parts; Record the measured values ​​of the geometric dimensions and photos of the wear marks on the components, and package the information into a measured data package; The measured data packets are uploaded to the data center via a wireless network and associated with the previous energy efficiency deviation vector; In the database of the data center, the feature threshold range of the corresponding fault label in the multidimensional fault feature library is adjusted according to the numerical range of the measured data packet. After the adjustment is completed, the suspected faulty component is reassembled back into the boiler system.

9. The energy consumption diagnosis method for a gas-fired steam boiler as described in claim 8, characterized in that, Before retrieving several historical fault case nodes adjacent to the location from the multidimensional fault feature database, a database self-cleaning step is also included: Periodically scan all historical fault case nodes in the multidimensional fault feature library to identify obsolete nodes that have not been called for a long time; Retrieve the original operation logs associated with the old nodes and check for any misjudgment records caused by sensor drift; If there are misjudged records, the old nodes are removed, and the gaps are filled with recent correct case nodes. The purified multidimensional fault feature library index is then reloaded into the memory of the fault tracing engine.

10. A gas-fired steam boiler energy consumption diagnostic system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the energy consumption diagnosis method for gas-fired steam boilers as described in any one of claims 1 to 9.

Images