A method and system for locating state abnormal points of a thermal power plant

Abstract

This invention relates to the field of anomaly detection and localization technology, and discloses a method and system for locating anomaly points in thermal power equipment. The method involves collecting furnace temperature field datasets, main steam pressure pulsation datasets, turbine shaft vibration datasets, and actual power load command sequences of the unit; screening benchmark samples under normal operating conditions to construct a multi-dimensional operating baseline; calculating the spatial distribution deviation index of the furnace temperature field dataset, the energy spectrum deviation index of the main steam pressure pulsation dataset, and the frequency response deviation index of the turbine shaft vibration dataset; mapping the spatial distribution deviation index, energy spectrum deviation index, and frequency response deviation index to generate an anomaly propagation probability matrix; performing threshold segmentation and connected component analysis on the anomaly propagation probability matrix to locate a set of candidate anomaly sources and outputting the spatial coordinates of the anomaly points. This achieves accurate localization and confidence level classification of candidate anomaly sources, improving anomaly localization accuracy and adaptability while ensuring real-time detection.

Description

Technical Field

[0001] This invention relates to the field of anomaly detection and location technology, and more specifically, to a method and system for locating anomaly points in thermal power equipment. Background Technology

[0002] Anomaly detection in the operation of thermal power equipment refers to the technology of monitoring and analyzing parameters such as temperature, pressure, and vibration of core equipment such as boilers and steam turbines to identify abnormal signs and locate the source of the anomaly. It is a key means to ensure the safe and stable operation of thermal power units and prevent major accidents.

[0003] Existing technologies often employ methods based on single-parameter threshold alarms or offline spectrum analysis. These methods rely on simple criteria such as upper temperature limits and vibration intensity limits, combined with human experience, to determine the cause of anomalies and locate the anomaly points. However, the parameter monitoring dimensions in these technologies are relatively isolated, making it difficult to capture complex anomaly characteristics such as spatial distribution distortion of the furnace temperature field, pressure pulsation energy spectrum migration, and shaft vibration mode coupling. Furthermore, the lack of correlation analysis between multiple parameters based on equipment topology makes it difficult to identify anomaly propagation paths, and the location results often lag behind the anomaly evolution process, affecting the timeliness of anomaly handling. Moreover, the lack of an adaptive update mechanism for the operating baseline makes it difficult to adapt to dynamic operating condition changes at different load stages, thus limiting the accuracy and robustness of anomaly detection. Summary of the Invention

[0004] This invention provides a method and system for locating abnormal points in thermal power equipment. By constructing a multi-dimensional operational baseline and mapping it to the equipment topology, an anomaly propagation probability matrix is ​​generated and spatial distribution, energy spectrum, and frequency response deviation index are integrated to achieve accurate location and confidence level classification of candidate points for anomaly sources. This improves the accuracy and adaptability of anomaly location while ensuring real-time detection.

[0005] To achieve the above objectives, the present invention provides a method for locating abnormal conditions in thermal power equipment, comprising:

[0006] Collect furnace temperature field data, main steam pressure pulsation data, and turbine shaft vibration data of the target thermal power unit during the monitoring period, and simultaneously obtain the actual power load command sequence of the unit. Based on the normal operating condition benchmark samples selected from the historical operating database, a multi-dimensional operating baseline is constructed that matches the actual power load command sequence. The operating baseline includes a temperature field gradient benchmark vector, a pressure pulsation frequency band benchmark vector, and a vibration mode benchmark vector. Calculate the spatial distribution deviation index of the furnace temperature field dataset relative to the temperature field gradient reference vector, calculate the energy spectrum deviation index of the main steam pressure pulsation dataset relative to the pressure pulsation frequency band reference vector, and calculate the frequency response deviation index of the turbine shaft vibration dataset relative to the vibration mode reference vector. The spatial distribution deviation index, energy spectrum deviation index, and frequency response deviation index are mapped based on the device topology to generate an anomaly propagation probability matrix. Threshold segmentation and connected component analysis are performed on the anomaly propagation probability matrix to locate the set of candidate anomaly sources, and the spatial coordinates and anomaly confidence levels of the anomaly points are output.

[0007] Furthermore, based on the normal operating condition benchmark samples selected from the historical operating database, a multi-dimensional operating baseline matching the actual generated power load command sequence is constructed, including: Extract continuous operating records from the historical operation database where the unit load rate is within the rated load range; The continuous operation records are subjected to a working condition stability assessment, and normal operating condition segments with load fluctuation rates less than a preset threshold and auxiliary machine operation modes unchanged are selected. The normal operating condition segment is divided into several load segments according to the actual generated power load command sequence; Statistical feature aggregation is performed on the furnace temperature field data, main steam pressure pulsation data and turbine shaft vibration data in each load segment to generate the temperature field gradient reference vector, pressure pulsation frequency band reference vector and vibration mode reference vector.

[0008] Further, the spatial distribution deviation index of the furnace temperature field dataset relative to the temperature field gradient reference vector is calculated, including: A thermocouple measuring point array is arranged in layers along the height of the furnace to obtain the circumferential temperature value of each layer; Fourier series expansion is performed on the circumferential temperature values ​​of each layer to extract the harmonic characteristic components of the temperature field. The harmonic characteristic components include a first-order eccentric component and a second-order asymmetric component. A thermal deviation evolution index is generated based on the amplitude change rate of the first-order eccentric component, and a combustion imbalance index is generated based on the phase jump amplitude of the second-order asymmetric component. The spatial distribution deviation index is generated by combining the thermal deviation evolution index and the combustion imbalance index.

[0009] Further, the energy spectrum deviation index of the main steam pressure pulsation dataset relative to the pressure pulsation frequency band reference vector is calculated, including: Variational mode decomposition of the pressure pulsation signal yields a finite number of eigenmode function components; Calculate the instantaneous frequency mean and instantaneous energy entropy of each intrinsic mode function component to generate mode feature pairs; The modal feature pairs are matched with the reference modes in the pressure pulsation frequency band reference vector to obtain a modal deviation sequence; The modal deviation sequence is subjected to an exponentially weighted moving average to generate the energy spectrum deviation index.

[0010] Further, the frequency response deviation index of the turbine shaft system vibration dataset relative to the vibration mode reference vector is calculated, including: A triaxial vibration acceleration sensor is installed in the bearing housing to collect the time-domain waveform of the shaft vibration. Perform cepstral analysis on the time-domain waveform to extract the characteristic sequence of the shaft system transfer function; The modal correlation coefficient is obtained by cross-correlation between the transfer function feature sequence and the vibration mode reference vector. The frequency response deviation index is calculated based on the modal correlation coefficient.

[0011] Furthermore, the spatial distribution deviation index, energy spectrum deviation index, and frequency response deviation index are mapped based on the device topology to generate an anomaly propagation probability matrix, including: The spatial distribution deviation index, energy spectrum deviation index, and frequency response deviation index are combined to obtain the comprehensive deviation index. Based on the comprehensive deviation index and correlation mapping results, an anomaly propagation probability matrix is ​​generated.

[0012] Furthermore, the spatial distribution deviation index, energy spectrum deviation index, and frequency response deviation index are comprehensively processed to obtain a comprehensive deviation index, including: The spatial distribution deviation index is normalized to obtain the temperature sub-index. The energy spectrum deviation index is logarithmically transformed to obtain the pressure component index; The frequency response deviation index is piecewise linearly mapped to obtain the vibration sub-index; The temperature, pressure, and vibration sub-indices are weighted and fused based on the distance attenuation factor in the equipment topology to generate a comprehensive deviation index.

[0013] Furthermore, based on the comprehensive deviation index and correlation mapping results, an anomaly propagation probability matrix is ​​generated, including: Construct a device connection topology diagram, wherein the nodes of the topology diagram are temperature measuring points, pressure measuring points and vibration measuring points, and the edges are the flow directions of the working fluid; An initial anomaly probability is assigned to each node based on the magnitude of the comprehensive deviation index; An anomaly propagation direction vector is established based on the working fluid flow direction, and the edge propagation probability is calculated in combination with the initial anomaly probability of the node. The initial anomaly probability is combined with the edge propagation probability to form the anomaly propagation probability matrix. The diagonal elements of the anomaly propagation probability matrix are the node self-sustaining probabilities, and the off-diagonal elements are the propagation impact probabilities.

[0014] Furthermore, threshold segmentation and connected component analysis are performed on the anomaly propagation probability matrix to locate the set of candidate anomaly sources, and the spatial coordinates and anomaly confidence levels of the anomaly points are output, including: The anomaly propagation probability matrix is ​​subjected to eigenvalue decomposition, and the eigenvector corresponding to the largest eigenvalue is extracted; Arrange the elements in the feature vector in descending order of their numerical values, and select the measurement point numbers corresponding to the first k elements; Based on the measurement point number, spatial mapping is performed in the three-dimensional model of the equipment to obtain the spatial coordinates of the abnormal point; Calculate the cumulative sum of the eigenvalue contribution rates of the first k elements. When the cumulative sum is greater than the preset confidence level, output the corresponding abnormal confidence level as high; otherwise, output as medium.

[0015] To achieve the above objectives, the present invention also provides a system for locating abnormal conditions in thermal power equipment, comprising: The data acquisition module is used to collect the furnace temperature field dataset, main steam pressure pulsation dataset, and turbine shaft vibration dataset of the target thermal power unit during the monitoring period, and simultaneously acquire the actual power load command sequence of the unit. The vector determination module is used to construct a multi-dimensional operating baseline that matches the actual power load command sequence based on normal operating condition benchmark samples selected from the historical operating database. The operating baseline includes a temperature field gradient benchmark vector, a pressure pulsation frequency band benchmark vector, and a vibration mode benchmark vector. The index calculation module is used to calculate the spatial distribution deviation index of the furnace temperature field dataset relative to the temperature field gradient reference vector, the energy spectrum deviation index of the main steam pressure pulsation dataset relative to the pressure pulsation frequency band reference vector, and the frequency response deviation index of the turbine shaft vibration dataset relative to the vibration mode reference vector. The matrix generation module is used to perform a correlation mapping between the spatial distribution deviation index, the energy spectrum deviation index, and the frequency response deviation index based on the device topology to generate an anomaly propagation probability matrix. The anomaly localization module is used to perform threshold segmentation and connected component analysis on the anomaly propagation probability matrix, locate the set of candidate points for anomaly sources, and output the spatial coordinates and anomaly confidence level of the anomaly points.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention discloses a method and system for locating anomaly points in thermal power equipment. The method involves collecting furnace temperature field datasets, main steam pressure pulsation datasets, turbine shaft vibration datasets, and actual power load command sequences of the unit. Normal operating condition benchmark samples are selected to construct a multi-dimensional operating baseline. The spatial distribution deviation index of the furnace temperature field dataset, the energy spectrum deviation index of the main steam pressure pulsation dataset, and the frequency response deviation index of the turbine shaft vibration dataset are calculated. The spatial distribution deviation index, energy spectrum deviation index, and frequency response deviation index are correlated and mapped to generate an anomaly propagation probability matrix. Threshold segmentation and connected component analysis are performed on the anomaly propagation probability matrix to locate a set of candidate anomaly sources and output the spatial coordinates of the anomaly points. This achieves accurate location and confidence level classification of candidate anomaly sources, improving anomaly location accuracy and adaptability while ensuring real-time detection. Attached Figure Description

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart illustrating a method for locating abnormal status points in thermal power equipment according to an embodiment of the present invention is shown. Figure 2 A schematic diagram of a system for locating abnormal status points of thermal power equipment according to an embodiment of the present invention is shown. Detailed Implementation

[0018] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0019] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application 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. Therefore, they should not be construed as limitations on this application.

[0020] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0021] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0022] The following is a description of preferred embodiments of the present invention in conjunction with the accompanying drawings.

[0023] like Figure 1 As shown, an embodiment of the present invention discloses a method for locating abnormal status points in thermal power equipment, including: S110: Collect furnace temperature field data, main steam pressure pulsation data, and turbine shaft vibration data of the target thermal power unit during the monitoring period, and simultaneously obtain the actual power load command sequence of the unit. S120: Based on the normal operating condition benchmark samples selected from the historical operating database, a multi-dimensional operating baseline is constructed that matches the actual power load command sequence. The operating baseline includes a temperature field gradient benchmark vector, a pressure pulsation frequency band benchmark vector, and a vibration mode benchmark vector. S130: Calculate the spatial distribution deviation index of the furnace temperature field dataset relative to the temperature field gradient reference vector, calculate the energy spectrum deviation index of the main steam pressure pulsation dataset relative to the pressure pulsation frequency band reference vector, and calculate the frequency response deviation index of the turbine shaft vibration dataset relative to the vibration mode reference vector. S140: Perform a correlation mapping based on the device topology of the spatial distribution deviation index, energy spectrum deviation index and frequency response deviation index to generate an anomaly propagation probability matrix; S150: Perform threshold segmentation and connected component analysis on the anomaly propagation probability matrix to locate the set of candidate anomaly sources, and output the spatial coordinates of the anomaly points and the anomaly confidence level.

[0024] In this embodiment, the monitoring cycle is set to 24 hours, and the sampling frequency is 1 minute. The furnace temperature field dataset is collected by 40 thermocouples at 8 measuring points per layer across 5 layers in the furnace height direction, including the temperature values ​​at each measuring point. The main steam pressure pulsation dataset is collected by two high-frequency pressure transmitters installed in the main steam pipeline, with a sampling frequency of 1000 Hz, extracting the effective value of the pulsation every 10 seconds. The turbine shaft vibration dataset is collected by installing one triaxial acceleration sensor in each of bearing housings 1 to 8, with a sampling frequency of 2000 Hz, extracting the vibration intensity value every minute. The actual power load command sequence is read from the DCS system, recording the unit's actual load command value every minute. The historical operation database contains the unit's operating data for the past three years, stored in the power plant's SIS system. The normal operating condition baseline sample selection criteria are: load rate of 60% to 100%, stable auxiliary equipment operation, and no abnormal operating conditions such as shutdown or load reduction, resulting in approximately 5000 hours of data. When constructing the multi-dimensional operating baseline, the load is divided into 12 load segments (180 to 660 MW, 40 MW each) according to the load command value. The average temperature gradient of each layer is calculated for the temperature field data in each segment to form a 40-dimensional temperature field gradient reference vector. Fourier transform is performed on the pressure pulsation data to extract the energy of the 10 to 100 Hz frequency band to form a 10-dimensional frequency band reference vector. Modal analysis is performed on the vibration data to extract the first three natural frequencies and damping ratios to form a 6-dimensional modal reference vector.

[0025] In some embodiments of this application, a multi-dimensional operating baseline matching the actual generated power load command sequence is constructed based on normal operating condition benchmark samples selected from the historical operating database, including: Extract continuous operating records from the historical operation database where the unit load rate is within the rated load range; The continuous operation records are subjected to a working condition stability assessment, and normal operating condition segments with load fluctuation rates less than a preset threshold and auxiliary machine operation modes unchanged are selected. The normal operating condition segment is divided into several load segments according to the actual generated power load command sequence; Statistical feature aggregation is performed on the furnace temperature field data, main steam pressure pulsation data and turbine shaft vibration data in each load segment to generate the temperature field gradient reference vector, pressure pulsation frequency band reference vector and vibration mode reference vector.

[0026] In this embodiment, the rated load is 660 MW, and the rated load range is 60% to 100%. All operating records within this load range over the past year are extracted. The load fluctuation rate is calculated as the absolute value of the load change rate per minute, with a preset threshold of 3 MW per minute. Segments with fluctuation rates less than 3 MW per minute are filtered out, for example, 5000 hours. "Auxiliary equipment operation mode unchanged" means that the number and combination of major auxiliary equipment such as coal mills, feedwater pumps, and induced draft fans remain stable. Segments with mode switching are eliminated based on auxiliary equipment status words, ultimately obtaining 4500 hours of normal operating conditions. The actual power load command sequence is divided into load segments of 40 MW each, for a total of 7 segments: 396-435 MW, 435-475 MW, 475-515 MW, 515-555 MW, 555-595 MW, 595-635 MW, and 635-660 MW. For each segment of temperature field data, the average temperature of each layer and the interlayer temperature difference are calculated to form a 10-dimensional temperature field gradient reference vector, such as [1250, 1280, 20, 25, ...]. For pressure pulsation data, the average energy of each frequency band is calculated to form a 10-dimensional frequency band reference vector, such as [120, 85, 95, ...]. For vibration data, the average of the first three modal frequencies and damping ratios is calculated to form a 6-dimensional modal reference vector, such as [18, 0.03, 45, 0.04, 78, 0.035]. Statistical feature aggregation uses the mean, standard deviation, and quartiles to ensure baseline representativeness. Load rate range screening ensures that the baseline covers the main operating conditions, and the volatility threshold is set according to the unit's primary frequency regulation performance.

[0027] The beneficial effects of the above technical solution are: load rate range screening focuses on the main operating segments, load fluctuation rate threshold eliminates unstable operating conditions, auxiliary equipment commissioning mode judgment ensures operating condition consistency, load segmentation achieves fine matching, statistical feature aggregation refines benchmark parameters, multi-dimensional baseline comprehensively represents the normal state, provides a reliable reference for deviation detection, and improves the benchmark scientificity of anomaly identification.

[0028] In some embodiments of this application, calculating the spatial distribution deviation index of the furnace temperature field dataset relative to the temperature field gradient reference vector includes: A thermocouple measuring point array is arranged in layers along the height of the furnace to obtain the circumferential temperature value of each layer; Fourier series expansion is performed on the circumferential temperature values ​​of each layer to extract the harmonic characteristic components of the temperature field. The harmonic characteristic components include a first-order eccentric component and a second-order asymmetric component. A thermal deviation evolution index is generated based on the amplitude change rate of the first-order eccentric component, and a combustion imbalance index is generated based on the phase jump amplitude of the second-order asymmetric component. The spatial distribution deviation index is generated by combining the thermal deviation evolution index and the combustion imbalance index.

[0029] In this embodiment, the furnace adopts a tangential combustion configuration at the four corners, with a height of 40 meters. It is arranged in five layers along the height at 10 meters, 20 meters, 30 meters, 35 meters, and 38 meters. Each layer has two measuring points on each of the four walls, totaling 40 thermocouples. The circumferential temperature values ​​refer to the temperature data of eight measuring points in the same layer arranged circumferentially within the furnace. For example, the data for the third layer are [T1=1250℃, T2=1280℃, T3=1260℃, T4=1290℃, T5=1270℃, T6=1240℃, T7=1265℃, T8=1285℃]. Fourier series expansion treats the eight temperature values ​​as periodic functions, calculating the Fourier coefficients. The first-order eccentric component is the amplitude of the first harmonic, reflecting the overall flame center offset; for example, amplitude A1=25 degrees Celsius. The second-order asymmetric component consists of the amplitude and phase of the second harmonic, reflecting combustion asymmetry on the left and right sides. For example, amplitude A2 = 15 degrees Celsius, phase φ2 = 45 degrees Celsius. The amplitude change rate is calculated by the amplitude difference between two consecutive sampling points (1 minute apart). For example, if A1 decreases from 25 to 20, the rate is -5 degrees Celsius / minute. The thermal deviation evolution index = |amplitude change rate| ÷ normal change rate threshold (threshold 5 degrees Celsius / minute), resulting in 1.0. The phase jump amplitude is calculated by the number of times the phase difference between two consecutive points exceeds 30 degrees. For example, if φ2 jumps from 45 degrees to 100 degrees, the jump amplitude is 55 degrees, exceeding the threshold. The combustion imbalance index = jump amplitude ÷ 360 degrees = 0.153. The thermal deviation evolution index and the combustion imbalance index are weighted and summed to obtain the spatial distribution deviation index, which is 0.661. A larger value indicates a more severe temperature field deviation. A positive correlation ensures that the index value is consistent with the degree of deviation. The normal rate of change threshold is set according to the boiler design specifications, and the phase jump threshold of 30 degrees is determined according to the combustion stability requirements.

[0030] The beneficial effects of the above technical solution are: the layered measuring point array fully covers the furnace space, the circumferential temperature value captures the tangential combustion characteristics, the Fourier series extracts harmonic features to achieve data dimensionality reduction, the first-order eccentric component reflects the flame center offset, the second-order asymmetric component reveals the uneven combustion on the left and right, the amplitude change rate and phase jump amplitude quantify the dynamic evolution, the thermal deviation and combustion imbalance index accurately characterize the combustion state, and the weighted fusion generates the spatial distribution deviation index, which significantly improves the physical meaning and sensitivity of furnace abnormal feature extraction.

[0031] In some embodiments of this application, calculating the energy spectrum deviation index of the main steam pressure pulsation dataset relative to the pressure pulsation frequency band reference vector includes: Variational mode decomposition of the pressure pulsation signal yields a finite number of eigenmode function components; Calculate the instantaneous frequency mean and instantaneous energy entropy of each intrinsic mode function component to generate mode feature pairs; The modal feature pairs are matched with the reference modes in the pressure pulsation frequency band reference vector to obtain a modal deviation sequence; The modal deviation sequence is subjected to an exponentially weighted moving average to generate the energy spectrum deviation index.

[0032] In this embodiment, the pressure pulsation signal is acquired by a piezoelectric pressure transmitter installed on the main steam pipeline, with 20,000 data points acquired over 10 seconds each time. Variational Mode Decomposition (VMD) decomposes the signal into five Intrinsic Mode Function (IMF) components, with a penalty factor set to 2000 and the number of modes K=5, separating five frequency bands: 20-50 Hz, 50-80 Hz, 80-120 Hz, 120-150 Hz, and 150-200 Hz. The instantaneous frequency of each IMF component is calculated using Hilbert transform; for example, the average instantaneous frequency of IMF1 is 35 Hz. The instantaneous energy entropy is calculated as the normalized information entropy of the energy distribution of that component, for example, 0.85. The modal feature pair is (35 Hz, 0.85). The pressure pulsation frequency band baseline vector is extracted from historical data under high-load normal operating conditions and contains 5 baseline modal feature pairs, such as (33 Hz, 0.82) and (55 Hz, 0.80). Similarity matching calculates the Euclidean distance; the distance between the IMF1 feature pair and the baseline mode is 2.03, and the normalized modal deviation is 0.203. The modal deviation sequence contains deviation values ​​for 5 components, such as [0.203, 0.156, 0.089, 0.234, 0.178]. The weight coefficient λ of the Exponentially Weighted Moving Average (EWMA) is set to 0.2. The energy spectrum deviation index = λ × current mean + (1-λ) × previous exponent. The current mean is 0.172, and the previous exponent is 0.165, resulting in 0.1664, which is proportional to the mean to ensure that the greater the deviation, the higher the exponent.

[0033] The beneficial effects of the above technical solution are: variational mode decomposition adaptively separates the pressure pulsation frequency band, multi-component extraction avoids frequency band aliasing, instantaneous frequency mean and energy entropy constitute two-dimensional features, improving the ability to characterize anomalies, similarity matching quantifies and benchmark deviation, exponential weighted moving average smooths instantaneous fluctuations, and energy spectrum deviation index comprehensively reflects pressure pulsation spectrum anomalies, providing reliable pressure dimension input for anomaly propagation analysis.

[0034] In some embodiments of this application, calculating the frequency response deviation index of the turbine shaft system vibration dataset relative to the vibration mode reference vector includes: A triaxial vibration acceleration sensor is installed in the bearing housing to collect the time-domain waveform of the shaft vibration. Perform cepstral analysis on the time-domain waveform to extract the characteristic sequence of the shaft system transfer function; The modal correlation coefficient is obtained by cross-correlation between the transfer function feature sequence and the vibration mode reference vector. The frequency response deviation index is calculated based on the modal correlation coefficient.

[0035] In this embodiment, the triaxial vibration acceleration sensor is a triaxial piezoelectric accelerometer, installed in the X, Y, and Z directions of eight bearing seats in the high-pressure cylinder, low-pressure cylinder, and generator of the steam turbine, totaling 24 measurement points. The time-domain waveform sampling frequency is 2000 Hz, with each sampling lasting 60 seconds, resulting in 120,000 data points. Cepstrum analysis first performs a Fourier transform on the waveform, then takes the logarithm of the spectrum and performs an inverse Fourier transform to obtain the cepstrum diagram. The inverse frequency components representing the shaft transfer function are extracted, for example, the amplitude values ​​at the inverse frequencies of 15 ms and 20 ms constitute the transfer function characteristic sequence [0.85, 0.72]. The vibration modal reference vector is extracted from historical normal operating condition data, including the first three natural frequencies (e.g., first order 18 Hz, second order 45 Hz, third order 78 Hz) and corresponding damping ratios (e.g., 0.03, 0.04, 0.035), and converted into a characteristic sequence [0.82, 0.75, 0.68]. Cross-correlation is performed by sliding correlation between the characteristic sequence of the transfer function and the reference vector, and the correlation coefficient is calculated. For example, the correlation coefficient ρ = 0.68. The frequency response deviation index = 1 - ρ, which is 0.32 and is inversely proportional to the correlation coefficient. The lower the correlation coefficient, the greater the deviation.

[0036] The beneficial effects of the above technical solution are: the three-dimensional sensor comprehensively captures vibration information, the cepstral analysis extracts the inherent transmission characteristics of the shaft system, avoids interference from the excitation source, the characteristic sequence of the transfer function is reduced in dimension to characterize the mode, the cross-correlation operation quantifies the deviation from the benchmark, the correlation coefficient intuitively reflects the health status, the frequency response deviation index provides anomaly criteria for vibration dimension, and provides key input for the anomaly propagation probability matrix.

[0037] In some embodiments of this application, the spatial distribution deviation index, energy spectrum deviation index, and frequency response deviation index are mapped based on the device topology to generate an anomaly propagation probability matrix, including: The spatial distribution deviation index, energy spectrum deviation index, and frequency response deviation index are combined to obtain the comprehensive deviation index. Based on the comprehensive deviation index and correlation mapping results, an anomaly propagation probability matrix is ​​generated.

[0038] In some embodiments of this application, the spatial distribution deviation index, energy spectrum deviation index, and frequency response deviation index are comprehensively processed to obtain a comprehensive deviation index, including: The spatial distribution deviation index is normalized to obtain the temperature sub-index. The energy spectrum deviation index is logarithmically transformed to obtain the pressure component index; The frequency response deviation index is piecewise linearly mapped to obtain the vibration sub-index; The temperature, pressure, and vibration sub-indices are weighted and fused based on the distance attenuation factor in the equipment topology to generate a comprehensive deviation index.

[0039] In this embodiment, the spatial distribution deviation index is calculated to be 0.661. Normalization is performed using the maximum-minimum method, resulting in a normalized temperature sub-index of 0.33. The energy spectrum deviation index is 0.1664, logarithmically transformed to ln(0.1664+1)=0.154. Adding 1 avoids negative values, yielding a pressure sub-index of 0.154. The frequency response deviation index is 0.32. Piecewise linear mapping is set as follows: 0 to 0.1 is mapped to 0.05, 0.1 to 0.5 is mapped to 0.05+0.5×(x-0.1), and 0.32 belongs to the second segment. The vibration sub-index is 0.05+0.5×(0.32-0.1)=0.16. In the equipment topology, the boiler-to-turbine pipeline length is 30 meters, with a distance attenuation factor of 1-length / 100=0.7. The turbine-to-generator length is 10 meters, with a factor of 0.9. The distance attenuation factor between measuring points is taken as the product along the propagation path. The factor from temperature measuring point to pressure measuring point is 0.7, and the factor from pressure to vibration is 0.9. During weighted fusion, the weights are 0.4 for temperature, 0.3 for pressure, and 0.3 for vibration. The pressure sub-index is multiplied by the attenuation factor of 0.7 to get 0.108, and the vibration sub-index is multiplied by 0.9 to get 0.144. The overall deviation index is 0.207.

[0040] The beneficial effects of the above technical solutions are: normalization eliminates the dimension of the temperature component, logarithmic transformation reduces the numerical range of the pressure component to avoid dominance, piecewise linear mapping limits the vibration component to a reasonable range, distance attenuation factor reflects the topological propagation characteristics, weighted fusion integrates the three-dimensional deviations, and the comprehensive deviation index fully reflects the overall health status of the equipment, providing a unified input for the abnormal propagation probability matrix and improving the comprehensiveness of abnormal analysis.

[0041] In some embodiments of this application, an anomaly propagation probability matrix is ​​generated based on the comprehensive deviation index and correlation mapping results, including: Construct a device connection topology diagram, wherein the nodes of the topology diagram are temperature measuring points, pressure measuring points and vibration measuring points, and the edges are the flow directions of the working fluid; An initial anomaly probability is assigned to each node based on the magnitude of the comprehensive deviation index; An anomaly propagation direction vector is established based on the working fluid flow direction, and the edge propagation probability is calculated in combination with the initial anomaly probability of the node. The initial anomaly probability is combined with the edge propagation probability to form the anomaly propagation probability matrix. The diagonal elements of the anomaly propagation probability matrix are the node self-sustaining probabilities, and the off-diagonal elements are the propagation impact probabilities.

[0042] In this embodiment, the equipment connection topology includes 40 temperature measuring points in the furnace, 2 pressure pulsation measuring points, and 24 vibration measuring points, totaling 66 nodes. Edges represent the flue gas flow direction (furnace → superheater → main steam pipe → turbine → condenser) and vibration transmission paths (high / medium pressure cylinder → low pressure cylinder → generator). Temperature measuring points are divided into 5 layers according to furnace height, with 8 points in each layer, and directed connections between layers. The initial anomaly probability is assigned as the normalized value of the comprehensive deviation index; for example, if the comprehensive deviation index of a certain temperature measuring point is 0.207, with a historical maximum of 1.0, the initial probability is 0.207. The working fluid flow direction vector is determined by the DCS system pipeline flow diagram, with the direction from high temperature and high pressure to low temperature and low pressure. Edge propagation probability = initial anomaly probability × direction factor 0.5, for example, 0.207 × 0.5 = 0.1035, representing the probability of anomaly propagating downstream. The anomaly propagation probability matrix is ​​a 66×66 square matrix. The diagonal elements = 1 - initial anomaly probability, for example, 0.793, representing the node's self-sustaining normal probability; the off-diagonal elements p... ij This represents the probability of anomaly propagating from node i to j; it is 0 if there is no direct path from i to j. The direction factor of 0.5 is set based on the experience that anomalous energy attenuates by 50% during propagation and is adjustable. Matrix combination is completed by filling diagonal elements and calculating off-diagonal elements to form a complete propagation model.

[0043] The beneficial effects of the above technical solution are: the topology graph construction mathematizes the device connection relationship, the node initialization is based on the actual degree of deviation, the anomaly propagation direction vector reflects the power transmission of the working fluid, the edge propagation probability calculation quantifies the possibility of transmission, the matrix form compactly expresses the anomaly propagation network, the diagonal and off-diagonal elements describe self-sustaining and propagation respectively, and the anomaly propagation probability matrix provides a probabilistic propagation model for anomaly localization, thereby improving the accuracy of anomaly tracing.

[0044] In some embodiments of this application, threshold segmentation and connected component analysis are performed on the anomaly propagation probability matrix to locate a set of candidate anomaly sources, and the spatial coordinates and anomaly confidence levels of the anomaly points are output, including: The anomaly propagation probability matrix is ​​subjected to eigenvalue decomposition, and the eigenvector corresponding to the largest eigenvalue is extracted; Arrange the elements in the feature vector in descending order of their numerical values, and select the measurement point numbers corresponding to the first k elements; Based on the measurement point number, spatial mapping is performed in the three-dimensional model of the equipment to obtain the spatial coordinates of the abnormal point; Calculate the cumulative sum of the eigenvalue contribution rates of the first k elements. When the cumulative sum is greater than the preset confidence level, output the corresponding abnormal confidence level as high; otherwise, output as medium.

[0045] In this embodiment, the anomaly propagation probability matrix is ​​a 66×66 real symmetric matrix. Eigenvalue decomposition uses the QR algorithm to calculate 66 eigenvalues. The largest eigenvalue, λ1, is 3.2, corresponding to the eigenvector v1, whose elements are the centrality scores of the 66 nodes in the anomaly propagation network. The elements are arranged in descending order of value. The first three elements correspond to the node numbers T15 (measuring point on the left wall of the third layer of the furnace), P2 (measuring point 2 of the main steam pipeline), and V7 (measuring point 7 of the low-pressure cylinder bearing). The k value is set to 3, representing the three most likely anomaly sources. The equipment 3D model uses a BIM model. The coordinates of measuring point T15 are (X=15.2 m, Y=8.5 m, Z=42.0 m), P2 is (X=45.6 m, Y=12.3 m, Z=38.5 m), and V7 is (X=78.9 m, Y=10.1 m, Z=15.2 m), outputting the spatial coordinates of the anomaly points. The eigenvalue contribution rate is calculated as λ1 / the sum of all eigenvalues. Since the sum of all eigenvalues ​​is 66, the λ1 contribution rate is approximately 3.2 ÷ 66 ≈ 4.85%. If the cumulative contribution rate of the first three elements exceeds the preset confidence level of 70% (actually, this needs to be calculated), the first k eigenvectors need to be accumulated due to insufficient contribution from a single eigenvalue. Here, this is simplified to a weighted sum of the first k elements. The preset confidence level is set to 70% based on the reliability requirements for anomaly diagnosis. If the cumulative contribution rate is 75% (greater than 70%), the output anomaly confidence level is high, indicating high certainty; 50% to 70% is medium; and less than 50% is low. Eigenvalue decomposition reveals the dominant propagation direction of the network, the size of the eigenvector elements reflects the importance of nodes, spatial mapping achieves physical location, and the confidence level provides a reliability assessment for decision-making, significantly improving the engineering practicality of anomaly location.

[0046] The beneficial effects of the above technical solution are: eigenvalue decomposition extracts the principal components of the anomaly propagation network, eigenvector elements quantify the anomaly centrality of nodes, descending order sorting automatically identifies key measurement points, k-value setting controls the number of candidate points, 3D model mapping realizes physical spatial positioning, contribution rate accumulation assesses positioning reliability, confidence level distinguishes diagnostic certainty, providing operation and maintenance personnel with intuitive anomaly location and credibility, and improving anomaly handling efficiency.

[0047] To further illustrate the technical concept of this invention, the technical solution of this invention will now be described in conjunction with specific application scenarios.

[0048] Correspondingly, such as Figure 2 As shown, this application also provides a system for locating abnormal conditions in thermal power equipment, including: The data acquisition module is used to collect the furnace temperature field dataset, main steam pressure pulsation dataset, and turbine shaft vibration dataset of the target thermal power unit during the monitoring period, and simultaneously acquire the actual power load command sequence of the unit. The vector determination module is used to construct a multi-dimensional operating baseline that matches the actual power load command sequence based on normal operating condition benchmark samples selected from the historical operating database. The operating baseline includes a temperature field gradient benchmark vector, a pressure pulsation frequency band benchmark vector, and a vibration mode benchmark vector. The index calculation module is used to calculate the spatial distribution deviation index of the furnace temperature field dataset relative to the temperature field gradient reference vector, the energy spectrum deviation index of the main steam pressure pulsation dataset relative to the pressure pulsation frequency band reference vector, and the frequency response deviation index of the turbine shaft vibration dataset relative to the vibration mode reference vector. The matrix generation module is used to perform a correlation mapping between the spatial distribution deviation index, the energy spectrum deviation index, and the frequency response deviation index based on the device topology to generate an anomaly propagation probability matrix. The anomaly localization module is used to perform threshold segmentation and connected component analysis on the anomaly propagation probability matrix, locate the set of candidate points for anomaly sources, and output the spatial coordinates and anomaly confidence level of the anomaly points.

[0049] In the description of the above embodiments, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0050] Although the invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the embodiments disclosed in this invention can be combined with each other in any way. The fact that not all of these combinations are described in this specification is merely for the sake of brevity and resource conservation.

[0051] It will be understood by those skilled in the art that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for locating abnormal conditions in thermal power equipment, characterized in that, include: Collect furnace temperature field data, main steam pressure pulsation data, and turbine shaft vibration data of the target thermal power unit during the monitoring period, and simultaneously obtain the actual power load command sequence of the unit. Based on the normal operating condition benchmark samples selected from the historical operating database, a multi-dimensional operating baseline is constructed that matches the actual power load command sequence. The operating baseline includes a temperature field gradient benchmark vector, a pressure pulsation frequency band benchmark vector, and a vibration mode benchmark vector. Calculate the spatial distribution deviation index of the furnace temperature field dataset relative to the temperature field gradient reference vector, calculate the energy spectrum deviation index of the main steam pressure pulsation dataset relative to the pressure pulsation frequency band reference vector, and calculate the frequency response deviation index of the turbine shaft vibration dataset relative to the vibration mode reference vector. The spatial distribution deviation index, energy spectrum deviation index, and frequency response deviation index are mapped based on the device topology to generate an anomaly propagation probability matrix. Threshold segmentation and connected component analysis are performed on the anomaly propagation probability matrix to locate the set of candidate anomaly sources, and the spatial coordinates and anomaly confidence levels of the anomaly points are output.

2. The method for locating abnormal status points of thermal power equipment according to claim 1, characterized in that, Based on normal operating condition benchmark samples selected from the historical operating database, a multi-dimensional operating baseline matching the actual generated power load command sequence is constructed, including: Extract continuous operating records from the historical operation database where the unit load rate is within the rated load range; The continuous operation records are subjected to a working condition stability assessment, and normal operating condition segments with load fluctuation rates less than a preset threshold and auxiliary machine operation modes unchanged are selected. The normal operating condition segment is divided into several load segments according to the actual generated power load command sequence; Statistical feature aggregation is performed on the furnace temperature field data, main steam pressure pulsation data and turbine shaft vibration data in each load segment to generate the temperature field gradient reference vector, pressure pulsation frequency band reference vector and vibration mode reference vector.

3. The method for locating abnormal status points of thermal power equipment according to claim 1, characterized in that, Calculate the spatial distribution deviation index of the furnace temperature field dataset relative to the temperature field gradient reference vector, including: A thermocouple measuring point array is arranged in layers along the height of the furnace to obtain the circumferential temperature value of each layer; Fourier series expansion is performed on the circumferential temperature values ​​of each layer to extract the harmonic characteristic components of the temperature field. The harmonic characteristic components include a first-order eccentric component and a second-order asymmetric component. A thermal deviation evolution index is generated based on the amplitude change rate of the first-order eccentric component, and a combustion imbalance index is generated based on the phase jump amplitude of the second-order asymmetric component. The spatial distribution deviation index is generated by combining the thermal deviation evolution index and the combustion imbalance index.

4. The method for locating abnormal status points of thermal power equipment according to claim 1, characterized in that, Calculating the energy spectrum deviation index of the main steam pressure pulsation dataset relative to the pressure pulsation frequency band reference vector includes: Variational mode decomposition of the pressure pulsation signal yields a finite number of eigenmode function components; Calculate the instantaneous frequency mean and instantaneous energy entropy of each intrinsic mode function component to generate mode feature pairs; The modal feature pairs are matched with the reference modes in the pressure pulsation frequency band reference vector to obtain a modal deviation sequence; The modal deviation sequence is subjected to an exponentially weighted moving average to generate the energy spectrum deviation index.

5. The method for locating abnormal status points of thermal power equipment according to claim 1, characterized in that, Calculating the frequency response deviation index of the turbine shaft system vibration dataset relative to the vibration mode reference vector includes: A triaxial vibration acceleration sensor is installed in the bearing housing to collect the time-domain waveform of the shaft vibration. Perform cepstral analysis on the time-domain waveform to extract the characteristic sequence of the shaft system transfer function; The modal correlation coefficient is obtained by cross-correlation between the transfer function feature sequence and the vibration mode reference vector. The frequency response deviation index is calculated based on the modal correlation coefficient.

6. The method for locating abnormal status points of thermal power equipment according to claim 1, characterized in that, The spatial distribution deviation index, energy spectrum deviation index, and frequency response deviation index are mapped based on the device topology to generate an anomaly propagation probability matrix, including: The spatial distribution deviation index, energy spectrum deviation index, and frequency response deviation index are combined to obtain the comprehensive deviation index. Based on the comprehensive deviation index and correlation mapping results, an anomaly propagation probability matrix is ​​generated.

7. The method for locating abnormal status points of thermal power equipment according to claim 6, characterized in that, The spatial distribution deviation index, energy spectrum deviation index, and frequency response deviation index are comprehensively processed to obtain a comprehensive deviation index, including: The spatial distribution deviation index is normalized to obtain the temperature sub-index. The energy spectrum deviation index is logarithmically transformed to obtain the pressure component index; The frequency response deviation index is piecewise linearly mapped to obtain the vibration sub-index; The temperature, pressure, and vibration sub-indices are weighted and fused based on the distance attenuation factor in the equipment topology to generate a comprehensive deviation index.

8. The method for locating abnormal status points of thermal power equipment according to claim 1, characterized in that, Based on the comprehensive deviation index and correlation mapping results, an anomaly propagation probability matrix is ​​generated, including: Construct a device connection topology diagram, wherein the nodes of the topology diagram are temperature measuring points, pressure measuring points and vibration measuring points, and the edges are the flow directions of the working fluid; An initial anomaly probability is assigned to each node based on the magnitude of the comprehensive deviation index; An anomaly propagation direction vector is established based on the working fluid flow direction, and the edge propagation probability is calculated in combination with the initial anomaly probability of the node. The initial anomaly probability is combined with the edge propagation probability to form the anomaly propagation probability matrix. The diagonal elements of the anomaly propagation probability matrix are the node self-sustaining probabilities, and the off-diagonal elements are the propagation impact probabilities.

9. The method for locating abnormal status points of thermal power equipment according to claim 1, characterized in that, The anomaly propagation probability matrix is ​​subjected to threshold segmentation and connected component analysis to locate the set of candidate anomaly sources, and the spatial coordinates and anomaly confidence levels of the anomaly points are output, including: The anomaly propagation probability matrix is ​​subjected to eigenvalue decomposition, and the eigenvector corresponding to the largest eigenvalue is extracted; Arrange the elements in the feature vector in descending order of their numerical values, and select the measurement point numbers corresponding to the first k elements; Based on the measurement point number, spatial mapping is performed in the three-dimensional model of the equipment to obtain the spatial coordinates of the abnormal point; Calculate the cumulative sum of the eigenvalue contribution rates of the first k elements. When the cumulative sum is greater than the preset confidence level, output the corresponding abnormal confidence level as high; otherwise, output as medium.

10. A system for locating abnormal state points of thermal power equipment, applied to the method for locating abnormal state points of thermal power equipment as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to collect the furnace temperature field dataset, main steam pressure pulsation dataset, and turbine shaft vibration dataset of the target thermal power unit during the monitoring period, and simultaneously acquire the actual power load command sequence of the unit. The vector determination module is used to construct a multi-dimensional operating baseline that matches the actual power load command sequence based on normal operating condition benchmark samples selected from the historical operating database. The operating baseline includes a temperature field gradient benchmark vector, a pressure pulsation frequency band benchmark vector, and a vibration mode benchmark vector. The index calculation module is used to calculate the spatial distribution deviation index of the furnace temperature field dataset relative to the temperature field gradient reference vector, the energy spectrum deviation index of the main steam pressure pulsation dataset relative to the pressure pulsation frequency band reference vector, and the frequency response deviation index of the turbine shaft vibration dataset relative to the vibration mode reference vector. The matrix generation module is used to perform a correlation mapping between the spatial distribution deviation index, the energy spectrum deviation index, and the frequency response deviation index based on the device topology to generate an anomaly propagation probability matrix. The anomaly localization module is used to perform threshold segmentation and connected component analysis on the anomaly propagation probability matrix, locate the set of candidate points for anomaly sources, and output the spatial coordinates and anomaly confidence level of the anomaly points.

Images