A method and system for early warning diagnosis of hydropower plant equipment based on multi-source perception

Abstract

This application discloses a method and system for early warning and diagnosis of hydropower plant equipment based on multi-source sensing. The method includes: collecting multi-source sensing data of target equipment within a monitoring window to form a multi-source time series; obtaining the equivalent heat source intensity distribution inside the target equipment based on the spatial distribution data of the temperature field within the current monitoring window, generating thermal resistance parameters for each segment of the heat transfer path of the target equipment, and identifying abnormal thermal resistance segments; extracting thermal response delay parameters based on equipment operating status data; extracting the coupled operating condition feature vector of the current monitoring window, which is obtained by fusing the vibration signal data with temperature features including thermal resistance parameters and thermal response delay parameters; identifying the fault type based on the coupled operating condition feature vector; and outputting the diagnostic conclusion for the current window. This improves the accuracy of hydropower plant equipment condition monitoring and operation and maintenance efficiency.

Description

Technical Field

[0001] This invention relates to the field of auxiliary monitoring technology for hydropower plant equipment, and in particular to a method and system for early warning diagnosis of hydropower plant equipment based on multi-source sensing. Background Technology

[0002] Auxiliary equipment (such as oil pumps, water pumps, and gas turbines) in the oil, water, and gas systems of hydropower plants are crucial components ensuring the safe and stable operation of hydro-generator units. Compared to the generator units themselves, auxiliary equipment operates under more complex conditions, experiences more frequent start-ups and shutdowns, and is exposed to humid, dusty, and fluctuating environments, resulting in a higher probability and frequency of anomalies. The operational status of oil, gas, and water auxiliary equipment directly affects the continuity and safety of the entire hydropower plant's production; therefore, real-time monitoring and early warning of its status are of great significance. Thermal anomalies (such as blocked heat dissipation channels, localized overheating of motor windings, and bearing frictional heat) are common fault types leading to equipment performance degradation and even unplanned shutdowns. However, a singular anomaly assessment can easily overlook other fault issues. Currently, existing monitoring systems for auxiliary equipment in hydropower plants are typically limited to collecting single physical quantities (such as temperature or vibration) and setting fixed thresholds for alarms. When a monitored parameter exceeds the threshold, an alarm is triggered, achieving preliminary identification of abnormal equipment conditions.

[0003] Therefore, effective condition monitoring and early warning of equipment are of great significance for improving equipment reliability and reducing operation and maintenance costs. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by integrating multi-source sensing data to achieve a high-precision fault type identification and early warning diagnosis method, which significantly improves the intelligence level of equipment status monitoring and operation and maintenance efficiency.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: A method for early warning and diagnosis of hydropower plant equipment based on multi-source sensing includes: S101, Collect multi-source sensing data of the target equipment within the monitoring window to form a multi-source time series, including at least temperature field spatial distribution data, vibration signal data and equipment operating status data; S102, Based on the spatial distribution data of the temperature field within the current monitoring window, obtain the equivalent heat source intensity distribution inside the target device, generate the thermal resistance parameters of each segment of the heat transfer path of the target device, and identify the abnormal thermal resistance segments. S103, extract thermal response delay parameters based on equipment operating status data, wherein the thermal response delay is decomposed into startup phase response delay and steady-state thermal fluctuation delay. S104, extract the coupled operating condition feature vector of the current monitoring window, which is obtained by fusing the vibration signal data with temperature features including thermal resistance parameters and thermal response delay parameters. Based on the coupled operating condition feature vector, identify the fault type and output the diagnostic conclusion of the current window.

[0006] Preferably, the spatial distribution data of the temperature field is defined as: the temperature time series of each temperature measuring point of the target device. composition, , For discrete sampling times, , Let m be the set of temperature measurement points; The device operating status data is defined as follows: Where 0 represents the shutdown state, 1 represents the startup process state, and 2 represents the steady-state operation state.

[0007] Preferably, S102 specifically includes: A1. Within the current monitoring window, utilize temperature time series. As a constraint, the average heat source intensity distribution of this window is obtained; A2. Divide the heat transfer path into several segments, and calculate the equivalent thermal resistance of each segment based on the temperature at both ends of each segment and the heat flow through that segment. A3. Compare the equivalent thermal resistance of each segment with the preset healthy baseline to identify segments with abnormal thermal resistance.

[0008] Preferably, in A2, the equivalent resistance of each segment is calculated as follows: For the i-th segment, the equivalent thermal resistance within this window The ratio of the temperature difference between the two ends of this section to the heat flow passing through it:

[0009] in, Let i be the equivalent thermal resistance of the i-th segment. and These are the average temperatures at both ends of the segment within the monitoring window. The heat flow through this section is obtained by integrating the average heat source intensity of the window.

[0010] Preferably, the identification of abnormal thermal resistance segments includes: When the rate of change of thermal resistance in a certain segment When the following conditions are met: The system determines that there is an abnormal thermal resistance in this segment and outputs the abnormal segment number: , This is the equivalent thermal resistance of this section. This serves as the preset healthy baseline for the thermal resistance of the i-th segment. The preset abnormal thermal resistance threshold is set based on expert experience and historical fault data.

[0011] Preferably, the thermal response delay parameter is specifically: the steady-state thermal fluctuation delay of the current window. Startup phase response delay corresponding to the most recent startup event ; The startup phase response delay is the time difference between the target device starting from a cold state and the temperature at the key measuring point first reaching the thermal equilibrium threshold. The steady-state thermal fluctuation delay is the time difference corresponding to the peak value of the cross-correlation function between the load sequence and the temperature sequence during the steady-state operation phase.

[0012] Preferably, the coupled operating condition feature vector includes the operating state features of the current window. Vibration characteristics and temperature characteristics; the vibration characteristics include at least the effective velocity value extracted from the vibration signal data. kurtosis factor Frequency band energy Any one or more; the temperature characteristics include startup phase response delay. Steady-state thermal fluctuation delay Maximum thermal resistance change rate in all segments The running state characteristics of the current window are defined as follows: That is, the state of most sampling points within the window. This indicates taking the mode.

[0013] Preferably, the step of identifying the fault type based on the coupled operating condition feature vector and outputting the diagnostic conclusion of the current window specifically includes: S201, Input the coupled operating condition feature vector into the pre-trained fault identification model and output the fault type; the fault type includes pure thermal fault, pure mechanical fault and mechanical-thermal coupled fault; S202, if the fault type is a pure thermal fault or a mechanical-thermal coupling fault, and there is a thermal resistance abnormal segment, the fault location is based on the output thermal resistance abnormal segment number. S203, Generate structured diagnostic conclusions for the current monitoring window; S204, based on the structured diagnostic conclusions of the monitoring window, triggers an alert when an abnormal state is determined.

[0014] Preferably, the abnormal state is determined as follows: an abnormal state is determined when any of the following conditions are met: The identified fault types are either purely thermal faults or mechanical-thermal coupling faults; Any rate of change of thermal resistance >0.5; The thermal fluctuation delay under steady-state conditions showed a monotonically increasing trend within three consecutive monitoring windows.

[0015] This application also proposes a hydropower plant equipment early warning and diagnosis system based on multi-source sensing, the system comprising: an acquisition module, a feature extraction module, and a diagnosis module; The acquisition module is used to collect multi-source sensing data of the target device within the monitoring window and form a multi-source time series, which includes at least temperature field spatial distribution data, vibration signal data and device operating status data. The feature extraction module is used to obtain the equivalent heat source intensity distribution inside the target device based on the temperature field spatial distribution data within the current monitoring window, generate thermal resistance parameters for each segment of the heat transfer path of the target device, and identify abnormal thermal resistance segments; extract thermal response delay parameters based on the device operating status data, wherein the thermal response delay is decomposed into the response delay during the startup phase and the thermal fluctuation delay under steady-state conditions. The diagnostic module is used to extract the coupled operating condition feature vector of the current monitoring window, which is obtained by fusing the vibration signal data with temperature features including thermal resistance parameters and thermal response delay parameters. Based on the coupled operating condition feature vector, the fault type is identified and the diagnostic conclusion of the current window is output.

[0016] The beneficial effects of this invention are: By deploying a multi-source heterogeneous sensor network, spatial distribution data of temperature field, vibration signal data, and operating status data of the target equipment are acquired. Based on the sliding window mechanism, the equivalent heat source intensity inversion and segmented thermal resistance calculation of the heat transfer path are completed within the monitoring window, enabling precise location of thermal resistance anomaly segments. By decomposing the thermal response delay into the start-up response delay and the steady-state thermal fluctuation delay, the thermal capacity and thermal resistance characteristics of the equipment are characterized respectively, and the thermal anomaly latency period is identified before the temperature amplitude exceeds the threshold. Furthermore, by fusing vibration characteristics (effective velocity value, kurtosis factor, frequency band energy) and temperature characteristics (start-up response delay, steady-state fluctuation delay, thermal resistance change rate of each segment), a coupled operating condition feature vector is constructed, and a pre-trained fault identification model is used to accurately distinguish between pure thermal faults, pure mechanical faults, and machine-thermal coupled faults. By comprehensively applying temperature field spatiotemporal reconstruction, thermal response feature decomposition, and vibration-thermal coupling analysis, we have achieved early warning, spatial location, and fault type differentiation for thermal anomalies in hydropower plant auxiliary equipment. Compared with existing technologies that rely on single threshold alarms and isolated analysis of vibration and temperature, this approach significantly improves fault location accuracy and diagnostic accuracy. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a method for early warning and diagnosis of hydropower plant equipment based on multi-source sensing, according to an embodiment of the present invention. Figure 2This is a schematic diagram of the structure of a hydropower plant equipment early warning and diagnosis system based on multi-source sensing, according to an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] In the description of this application, 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 indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0020] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0021] Thermal failure is one of the main failure modes of auxiliary equipment in hydropower plant oil, water, and gas systems during long-term operation. Existing monitoring methods mostly rely on single temperature threshold alarms, only issuing warnings after the temperature exceeds the limit, failing to effectively identify thermal failures during their latency period. Traditional methods do not decompose the thermal response process mechanistically, making it difficult to distinguish between thermal capacity-type and thermal resistance-type anomalies, leading to ambiguous fault root cause identification and thus introducing blind spots into operation and maintenance. Furthermore, vibration monitoring and temperature monitoring are isolated from each other, making it difficult to accurately distinguish between purely thermal, purely mechanical, and machine-thermal coupled faults, resulting in insufficient reliability of diagnostic results.

[0022] Therefore, this application constructs a joint monitoring matrix of the equipment's temperature field and vibration field by deploying a multi-source heterogeneous sensor network. Based on this, a sliding window mechanism is used to reconstruct the temperature field in time and space and invert the equivalent heat source intensity, thereby calculating the equivalent thermal resistance of each segment of the heat transfer path and realizing the spatial location of thermal anomalies. At the same time, the thermal response delay is decomposed into the response delay during the start-up phase and the thermal fluctuation delay under steady-state conditions, which respectively characterize the thermal capacity and thermal resistance characteristics. Furthermore, vibration features are extracted and fused with thermal features across modes to construct a coupled operating condition feature vector, which is input into the fault identification model and outputs the classification results of pure thermal faults, pure mechanical faults, or machine-thermal coupled faults. Combined with the location of the thermal resistance anomaly segment, a structured diagnostic conclusion containing the fault type and fault location is output.

[0023] Example 1: Figure 1 This is a flowchart illustrating the early warning and diagnosis method for hydropower plant equipment based on multi-source sensing, according to an embodiment of the present invention.

[0024] like Figure 1 As shown, a method for early warning and diagnosis of hydropower plant equipment based on multi-source sensing includes the following steps: S101, through a pre-deployed multi-source heterogeneous sensor network, continuously collects multi-source sensing data of the target device within the monitoring window at a fixed sampling period to form a multi-source time series, which includes temperature field spatial distribution data, vibration signal data and device operating status data.

[0025] For example, the target equipment may include any one of the following: an oil pump in a hydroelectric power plant's oil system, a water pump in a water system, or a gas turbine in a gas system. Sensors are deployed in the critical heat-affected zones and vibration-sensitive zones of the target equipment to form a multi-source heterogeneous sensor network.

[0026] The multi-source heterogeneous sensor network can include temperature sensors forming a spatial temperature measurement point matrix, wireless sensors for detecting vibration data, and current harmonic sensors, continuously acquiring data at a fixed sampling frequency to form a multi-source time series. Specifically, it can be configured as follows: Temperature sensors are deployed in the critical heat-affected zone of the target equipment. For example, the critical heat-affected zone may include the surface of the motor housing, bearing housing, media inlet / outlet channels (referring to temperature measurement points on the walls of the inlet / outlet pipes of the media (compressed air, water, oil, etc.) processed by the target equipment, typically located near the pump body or machine flange; for example, in pneumatic equipment, this refers to the temperature measurement points on the walls of the compressed air inlet / outlet pipes; in water pump equipment, it refers to the temperature measurement points at the pump body inlet / outlet flanges; in oil pump equipment, it refers to the temperature measurement points on the walls of the oil inlet / outlet pipes, including the oil inlet and outlet pipes near the pump body flanges), cooling duct outlets, and environmental reference points. These measurement points constitute a spatial temperature measurement point matrix, forming a spatial temperature measurement point set. m represents the total number of temperature measurement points on a single device, with each temperature sensor using a fixed sampling period. Temperature data is continuously collected (e.g., every 1 second) to form a temperature time series. , , To obtain spatial distribution data of the temperature field at discrete sampling times; A three-in-one wireless sensor is deployed in the vibration-sensitive area of ​​the target device. This vibration-sensitive area may include the motor's on-load end and the bearing housing. The three-in-one wireless sensor is used to sample at a frequency... (e.g., 2560Hz) Continuously acquire triaxial acceleration signals to form a vibration time series. As vibration signal data, it is used for subsequent extraction of vibration features; The effective value of the motor current is collected by a current harmonic sensor to form a current time series. The sampling frequency is synchronized with the temperature sensor, and the device operating status data is analyzed based on the changing trend of the current time series. The current harmonic sensor is used to output the raw current waveform to extract the fundamental frequency (i.e., the frequency conversion), which is used for subsequent frequency band energy calculations. 0 indicates a shutdown state (determined by the current at multiple consecutive sampling points being below the no-load threshold), 1 indicates a startup state (determined by the current rising from below the no-load threshold to a steady-state value), and 2 indicates a steady-state operating state (determined by the current fluctuating near the steady-state value for a duration exceeding the threshold).

[0027] It should be noted that the multi-source sensing data mentioned above are synchronized in the time dimension, aligned using a unified timestamp, and based on the temperature sampling period. As a benchmark; to achieve time-series matching between temperature and vibration data, the vibration signal data is not aligned point-by-point, but processed in units of the same monitoring window as the temperature data: within each monitoring window (the definition of the monitoring window is shown in S102), window statistical characteristics (including effective velocity value, kurtosis factor, frequency band energy, etc.) are calculated for the vibration sampling points within the window, and these window statistical characteristics are assigned to the end time of the window. As a time marker, it achieves time alignment with the temperature window characteristics; the temperature data retains the original sampling points within the window and is used for heat source intensity inversion and thermal resistance calculation within the window.

[0028] S102, based on the spatial distribution data of the temperature field within the current monitoring window, obtain the equivalent heat source intensity distribution inside the target device, generate the thermal resistance parameters of each segment of the heat transfer path of the target device, and locate the thermal resistance abnormal segment.

[0029] Specifically, a sliding monitoring window mechanism is adopted, and data processing is performed within each monitoring window. The length L of the monitoring window is determined according to the actual monitoring scenario requirements. In this embodiment, L = 300 seconds (i.e., 300 temperature sampling points), and the window sliding step size is... Seconds, window number w, corresponding time interval .

[0030] In some embodiments, step S102 specifically includes: A1. Within the current monitoring window w, utilize the temperature time series within the window. As a constraint, the equivalent heat source intensity distribution within the window is inverted to obtain the average heat source intensity distribution of the window.

[0031] For example, the heat conduction equation (a three-dimensional unsteady-state heat conduction equation describing the evolution of the temperature field in the temporal region of the device) is used:

[0032] in, For material density, is the specific heat capacity, k is the thermal conductivity, obtained from the actual material handbook of each component of the target equipment; T is the temperature, obtained through actual measurement by the temperature sensor in S101. The equivalent heat source intensity is denoted as , and is the unknown variable to be inverted. The equipment space is discretized into a finite element network (the motor, housing, and air duct are divided into several elements according to the equipment geometry). The measured temperature sequence within the window is used as the boundary condition, and the equivalent heat source intensity distribution at each moment is solved using the regularized least squares method. This is used to characterize the dynamic changes in the heat generation and dissipation capabilities inside the device.

[0033] in, This is the measured temperature vector. The temperature distribution is calculated from the heat conduction equation. The pre-constructed regularization parameters are selected using the L-curve method to balance the goodness of fit of the data with the smoothness of the solution. For example, an L-curve is constructed for regularization. The x-axis represents the data fitting residuals. Using the vertical axis as the ordinate, for a series of candidates Value (e.g.) Calculate the corresponding points, draw the curve, and select the points corresponding to the inflection points of the L-curve. The value is taken as the optimal value, and the inflection point is defined as the point on the curve with the greatest curvature. The selection process is completed offline once at the initial stage of equipment commissioning and remains fixed during subsequent operation. If there are major changes to the equipment structure (such as replacing the motor or modifying the cooling system), the selection can be repeated. Since the temperature changes over time within the window, the inversion yields the Q distribution at each moment within the window. Further averaging over the window time is then performed. The average heat source intensity distribution of this window is used for subsequent thermal resistance calculations.

[0034] It should be noted that the specific application principles of the heat conduction equation and finite element network can be found in the relevant existing technology descriptions, and will not be elaborated upon in this invention.

[0035] A2. Divide the heat transfer path into several segments, and calculate the equivalent thermal resistance of each segment based on the temperature at both ends of each segment and the heat flow through that segment.

[0036] Specifically, based on the structural characteristics of the target equipment and the direction of heat flow, the heat transfer path is divided into several continuous segments, with temperature measuring points arranged at both ends of each segment.

[0037] In one exemplary implementation, for pneumatic equipment, the heat transfer path is divided into four segments: the first segment is from the motor winding to the motor housing; the second segment is from the motor housing to the cooling duct inlet; the third segment is from the cooling duct inlet to the cooling duct outlet; and the fourth segment is from the cooling duct outlet to the environment. For oil pumps and water pumps, similar path divisions can be made according to their specific structural characteristics. For example, for a water-cooled water pump, the heat transfer path can be divided into: the first segment is from the motor winding to the motor housing; the second segment is from the motor housing to the cooling water jacket inlet; the third segment is from the cooling water jacket inlet to the cooling water jacket outlet; and the fourth segment is from the cooling water jacket outlet to the environment. For an oil pump, the heat transfer path can be divided into: the first segment is from the motor winding to the motor housing; the second segment is from the motor housing to the oil tank wall; the third segment is from the oil tank wall to the cooling fins; and the fourth segment is from the cooling fins to the environment. The basic principle of path division is to segment the heat transfer path from the heat source (motor windings, bearing friction) to the final heat dissipation end (environment) according to the physical structural interfaces (such as the casing, radiator, and cooling medium channels). The core is to set temperature measuring points at both ends of each segment or obtain temperature values ​​through finite element interpolation. This invention will not elaborate on this.

[0038] For the i-th segment, its spatial region is Temperature at both ends of the i-th segment The average of the measured sequences from the temperature measurement points at both ends of this segment is obtained by taking the window time average. It should be noted that for the interior of path segments without deployed temperature measurement points (such as inside resistor windings), the temperature can be obtained through finite element inversion results. Obtained by interpolation with the heat conduction equation; The heat flux flowing through the i-th segment is obtained by integrating the window average heat source intensity (the volume integral of the equivalent heat source intensity within the spatial region of this segment, i.e., obtained by integrating the equivalent heat source intensity inversion result along the path): , This refers to the spatial region where this segment is located.

[0039] Specifically, the equivalent resistance of each segment is calculated as follows: For the i-th segment, the equivalent thermal resistance within this window The ratio of the temperature difference between the two ends of this section to the heat flow passing through it:

[0040] A3. Compare the equivalent resistance of each segment with the preset healthy baseline to identify segments with abnormal thermal resistance.

[0041] Among them, the preset healthy baseline of the thermal resistance of the i-th segment. Based on historical normal operating condition window data (without abnormal operation data) and management personnel determination, for example, taking normal operating condition window data from the initial stage of target equipment commissioning (or within 30 days of equipment restarting after maintenance), data segments within this period that were affected by drastic fluctuations in ambient temperature (e.g., daily temperature difference exceeding 15℃) or abnormal load fluctuations (current exceeding rated value ±20%) are removed. The thermal resistance of each segment within the remaining valid window is calculated, and the median is taken as the median. .

[0042] Specifically, when the rate of change of thermal resistance in a certain segment When the following conditions are met: The system determines that there is an abnormal thermal resistance in this segment and outputs the abnormal segment number: ; in, A preset healthy baseline for the thermal resistance of the i-th segment. The preset abnormal thermal resistance threshold is set based on expert experience and historical fault data. It can be determined by analyzing the fluctuation range of the target equipment's thermal resistance under healthy conditions. Three times the relative standard deviation of the healthy thermal resistance is taken as the abnormal threshold, which is the preferred value. That is, when the thermal resistance rises more than 20% relative to the baseline, anomaly identification is triggered, and the anomaly segment number is output. As a result of spatial location of thermal anomalies, in practical applications, maintenance personnel can adjust this threshold in the system configuration interface according to factors such as equipment importance and environmental conditions.

[0043] A4. Output the thermal resistance vector of the current window. , thermal resistance rate vector And the abnormal segment number (if it exists), where n is the number of segments.

[0044] S103, extract the thermal response delay based on the equipment operating status data. The thermal response delay is decomposed into the startup phase response delay and the steady-state thermal fluctuation delay. The startup phase response delay is the time difference between the equipment starting from a cold state and the temperature at the key measuring point first reaching the thermal equilibrium threshold. The steady-state thermal fluctuation delay is the load sequence (motor current sequence) during the steady-state operation phase. ) and temperature sequence (temperature sequence of key measuring points) The time difference corresponding to the peak of the cross-correlation function of ().

[0045] Specifically, based on temperature field spatial distribution data and equipment operating status data, thermal response delay parameters of the target equipment are extracted. These parameters include startup response delay and steady-state fluctuation delay, which respectively reflect the equipment's thermal capacity and thermal resistance characteristics. Specifically, the steady-state thermal fluctuation delay within the current window is defined as follows: Response delay during the startup phase corresponding to the most recent startup event (If no new startup event occurs, the previous value will be used.)

[0046] It should be noted that the startup response delay is used to characterize the heat source establishment and thermal capacity characteristics during the equipment startup phase, and the steady-state fluctuation delay is used to characterize the response characteristics of thermal resistance to load fluctuations during steady-state operation of the equipment.

[0047] In some embodiments, a response delay is initiated. It is an event-driven feature that is generated only when the device transitions from shutdown to startup, and is not output to every window. It is defined as: Detect startup events and determine the moment when the equipment starts from a cold state based on the start / stop status signals in the equipment operating status data. That is, from the equipment operating status The moment when it changes from 0 to 1 is determined; The key measuring point is defined as the on-load end temperature measuring point of the motor, and its temperature is... Calculate the temperature change rate sequence at key measuring points:

[0048] in The sampling interval; Define the end time of the startup phase. The moment when the following conditions are met for the first time:

[0049] in A preset temperature change rate threshold is used to represent the criterion for temperature stabilization. For example, for gas turbine equipment in hydropower plants, the temperature change rate distribution from startup to steady state is analyzed, and the 95th percentile of the temperature change rate after steady state is taken as the threshold. (For example, (Take 0.01. For different equipment types such as water pumps and oil pumps, the ε value can be calibrated separately according to the statistical distribution of the temperature change rate of the same type of equipment, and each equipment is stored independently). As a preset thermal balance temperature benchmark (thermal balance threshold), based on historical data statistics of the target equipment, the average steady-state operating temperature under healthy conditions is taken. For example, steady-state operating temperature data of the target equipment is collected within the first 30 days after commissioning (or within 30 days after equipment is restarted after maintenance). This period is considered the equipment health status window. Data segments within this period that are affected by drastic fluctuations in ambient temperature (such as daily temperature difference exceeding 15℃) or abnormal load fluctuations (current exceeding rated value ±20%) are excluded. The arithmetic mean of the remaining valid data is taken as the average. If 30 days of valid data cannot be obtained in the initial stage of operation of the target equipment, the statistical average value of the same model of equipment under similar working conditions shall be used as the initial value and updated gradually in subsequent operation. The response delay during the startup phase is:

[0050] This parameter reflects the time required for the equipment to establish thermal equilibrium from a cold start. An increase in this parameter indicates an increase in the equipment's heat capacity or an obstruction in the establishment of the initial heat source, pointing to a heat capacity-type anomaly. Update after each startup event and retain it until the next startup event occurs.

[0051] In some embodiments, steady-state fluctuation delay Defined as: During the steady-state operation phase of the target equipment (i.e.) ), calculate the steady-state ripple delay within each window: Within the monitoring window w, the time delay relationship between the load sequence and the temperature sequence is analyzed using a cross-correlation function, and the motor current sequence is extracted. Temperature sequence of key measuring points That is, the load sequence is defined as the effective value sequence of motor current, and the temperature sequence is the temperature sequence of key measuring points, and the discrete cross-correlation function of the two is calculated. The time delay corresponding to the maximum correlation value is taken as the steady-state fluctuation delay:

[0052]

[0053] in, The number of delay steps (corresponding to time delay) ), Search Scope Second, and These are the mean values ​​of the current and temperature sequences within the window, respectively. This parameter reflects the lag time of load change transmission to temperature response. Its increase indicates an increase in thermal resistance of the heat transfer path, pointing to thermal resistance anomalies, such as blockage of heat dissipation channels or surface dirt adhesion. Updated within each window, reflecting the thermal resistance response characteristics under the current window.

[0054] It should be noted that only when When =2, efficient( (This refers to the current window's running state characteristics); otherwise, Mark as invalid value (or reuse valid value from the previous steady-state run) to avoid misleading cross-correlation calculations in the non-steady-state segment.

[0055] S104, extract the coupled operating condition feature vector of the current monitoring window. The coupled operating condition feature vector is obtained by fusing the vibration signal data with temperature features including thermal resistance parameters and thermal response delay parameters. Based on the coupled operating condition feature vector, identify the fault type and output the diagnostic conclusion of the current window.

[0056] In some embodiments, temperature features, vibration features, and operating status features of the current monitoring window are extracted, and the coupled operating condition feature vector includes the operating status features of the current window. Vibration characteristics and temperature characteristics; vibration characteristics include at least the effective velocity value extracted from the vibration signal data. kurtosis factor Frequency band energy Any one or more of the following; temperature characteristics include startup phase response delay. Steady-state thermal fluctuation delay Maximum thermal resistance change rate in all segments The current window's running state characteristics are defined as follows: That is, the state of most sampling points within the window. This indicates taking the mode.

[0057] Specifically, in vibration characteristics, the effective value of velocity To reflect the overall vibration intensity, the effective value of velocity is calculated using the following formula: , The velocity sequence obtained by integrating acceleration. The number of vibration sampling points within the window; the kurtosis factor K reflects the impact failure and is defined as the ratio of the fourth-order central moment to the square of the second-order moment: , The average acceleration within the window. For vibration acceleration time-domain sequence; frequency band energy Defined as in frequency conversion The frequency band energy ratio near its harmonics is used to identify faults such as imbalance, misalignment, and looseness. As an example, the spectrum A(f) is obtained by performing a Fast Fourier Transform on the acceleration signal, and the bandwidth is defined as the frequency at the target frequency. ( Centered on, with a width of The narrow band range, i.e., the frequency band boundary Calculated according to the following formula: , , The frequency band boundary is approximately 1, 2, or 3 times the frequency shift (e.g.) bandwidth), To analyze the upper frequency limit, half of the sampling frequency is taken; where, the frequency conversion... Determined based on the target equipment's rotational speed: For constant-speed equipment, the rotational frequency is determined according to the rated speed on the equipment's nameplate, using the following formula: = n / 60, where n is the rated speed (r / min). For variable frequency speed control equipment, the rotational frequency is obtained by extracting the fundamental frequency from the current signal collected by the current harmonic sensor in the window, or by determining in real time the frequency corresponding to the maximum amplitude peak in the vibration signal spectrum. This embodiment will not elaborate on this.

[0058] It should be noted that, in order to unify the dimensions, each element in the coupled operating condition feature vector is normalized, for example, by using Z-score standardization. The mean and standard deviation are calculated from the training sample set, which will not be elaborated upon in this invention. Therefore, this coupled operating condition feature vector simultaneously includes dynamic characteristics of thermal response, spatial thermal resistance distribution characteristics, and vibration and shock characteristics, providing multi-dimensional information input for fault type identification.

[0059] In some embodiments, the fault type is identified based on the coupled operating condition feature vector, and a diagnostic conclusion including the fault type and fault location is output in combination with the thermal resistance anomaly segment location result, specifically including: S201 inputs the coupled operating condition feature vector into the pre-trained fault identification model and outputs the fault type.

[0060] Specifically, the fault types include purely thermal faults, purely mechanical faults, and machine-thermal coupling faults, for example: Pure thermal fault: Vibration signal is normal, but thermal response delay parameter or thermal resistance parameter is abnormal, indicating changes in heat capacity or deterioration of thermal resistance; Pure mechanical faults: abnormal vibration signal, normal thermal response delay parameters and thermal resistance parameters, mechanical faults such as bearing wear and rotor imbalance; Mechanical-thermal coupling fault: The vibration signal and thermal response parameters are abnormal at the same time, and there is a time correlation between the two. This indicates a coupling state in which mechanical faults lead to increased frictional heating or thermal anomalies cause mechanical performance degradation.

[0061] Specifically, the methods for constructing and training pre-trained fault identification models may include: Collect historical operating data (a large amount of historical data and a fault case library under various known fault conditions) of the target equipment and other equipment of the same type (distributed in different hydropower plants or under different operating conditions), covering at least the following fault types: pure thermal faults (overheating caused by blocked heat dissipation ducts, aging of motor winding insulation, and frictional heat generated by poor bearing lubrication), pure mechanical faults (bearing wear, rotor imbalance, misalignment, and loose foundation), and machine-thermal coupling faults (severe bearing wear leading to increased frictional heat generation, and short circuits between motor turns leading to local overheating and abnormal vibration). Historical operating data is processed according to methods S101 to S104 to generate coupled operating condition feature vectors for windows. Manually labeled fault types are used as supervision signals to construct a dataset containing coupled operating condition feature vectors and corresponding fault type labels. For example, at least two professionals, in conjunction with equipment maintenance records, infrared thermal imaging detection reports, and disassembly inspection records, jointly label the fault types. If the labeling results are inconsistent, a third-party expert will make the decision. The labeled dataset is divided into training, validation, and test sets. A pre-selected network structure (e.g., a multilayer perceptron (MLP) with an input layer dimension equal to the feature vector dimension, two hidden layers with 64 nodes each, and a three-node output layer using softmax activation) is used as the classification model. The classification model is trained using the training set, i.e., with the coupled working condition feature vector as input and the fault type probability as output, establishing a mapping relationship from the feature space to the fault type space. Cross-validation is used to evaluate the model's generalization ability, continuously optimizing the model parameters. The model parameters with the highest accuracy on the validation set are selected to obtain the final fault identification model, which outputs the probabilities of three fault types. , For pure thermal fault types, The probability of a purely mechanical failure type. The probability of machine-thermal coupling fault type.

[0062] Specifically, the pre-trained fault identification model is also used for: based on the output fault type probability, when When the fault type is selected, the corresponding category will be output.

[0063] S202, if the fault type is a pure thermal fault or a machine-thermal coupling fault, and there is an abnormal thermal resistance segment, then the fault location is determined by combining the abnormal thermal resistance segment results output by S102: Output the abnormal segment number, which corresponds to the specific faulty component, for management personnel to monitor and reference.

[0064] S203, Generate a structured diagnostic conclusion for the current monitoring window, which includes at least one of the following: Window time indicator ; Fault types (if a fault is identified): purely thermal fault, purely mechanical fault, or mechanical-thermal coupling fault; Fault location (if present): The name of the component corresponding to the section with abnormal thermal resistance (e.g., "heat dissipation duct inlet to outlet section", "motor housing to heat dissipation duct inlet section", etc.). Characteristic change: Rate of change of startup response delay relative to baseline The rate of change of steady-state fluctuation delay relative to the baseline The rate of change of thermal resistance and the rate of change of vibration characteristics in each segment; among them, , This is the baseline value under the healthy condition of the equipment, taken as the average value of the normal operating condition window during the initial stage of commissioning.

[0065] S204. Based on the diagnostic conclusions of the monitoring window, when an abnormal state is determined, the system triggers an early warning and pushes the diagnostic conclusions to the management personnel for inspection and reference, forming a feedback record and updating it.

[0066] Specifically, an abnormal state is determined when any of the following conditions are met: The identified fault types are either purely thermal faults or mechanical-thermal coupling faults; any >0.5 (i.e., thermal resistance exceeds the baseline by 50%) It shows a monotonically increasing trend within three consecutive monitoring windows (i.e.) ).

[0067] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: By employing a sliding window mechanism to reconstruct the spatiotemporal intensity of the equivalent heat source inside the equipment, the limitations of traditional temperature monitoring, which can only sense surface temperature, are overcome, laying the foundation for heat transfer path analysis. On this basis, by dividing the heat transfer path into several continuous segments and calculating the equivalent thermal resistance of each segment, the spatial approximate location of thermal anomalies is realized for the first time, and specific faulty component information is output, providing reliable and efficient monitoring auxiliary information for operation and maintenance personnel.

[0068] The thermal response delay is decomposed into startup response delay and steady-state fluctuation delay, pointing to thermal capacity anomalies and thermal resistance anomalies respectively, thus achieving a preliminary distinction of thermal fault mechanisms and providing more instructive diagnostic basis for maintenance personnel. Simultaneously, by extracting vibration features and fusing them with thermal features across modes, a coupled operating condition feature vector is constructed. A classification model is used to output classification results for pure thermal faults, pure mechanical faults, or mechanical-thermal coupled faults. When mechanical faults lead to increased frictional heating or thermal anomalies cause mechanical performance degradation, the coupled features can capture the correlation between the two, avoiding the misclassification of coupled faults as a single fault type by traditional methods, thereby improving diagnostic accuracy and reducing redundant or ineffective maintenance. This overcomes the shortcomings of traditional methods where vibration monitoring and temperature monitoring are isolated, achieving comprehensive diagnosis based on multi-physics information. The sliding window mechanism enables continuous output of diagnostic conclusions. Within each monitoring window, multi-dimensional feature parameters such as thermal resistance vector and thermal response delay parameters are output, which can construct a dynamic baseline of equipment health status. By monitoring the changing trends of each feature parameter, continuous tracking and quantitative assessment of equipment health status can be achieved, providing data support for equipment health management throughout its entire life cycle. Multi-dimensional early warning trigger conditions are set, including fault type identification, thermal resistance exceeding limits, and monotonically increasing trend of steady-state fluctuation delay, to ensure the timeliness and reliability of early warnings.

[0069] Therefore, by introducing the thermal fluctuation delay characteristics under steady-state operating conditions and utilizing the cross-correlation analysis of load and temperature sequences, the influence of different load conditions on thermal response characteristics can be adaptively eliminated, reducing false alarms caused by changes in operating conditions. Simultaneously, based on feedback from maintenance personnel regarding the early warning results, the health baseline can be updated regularly, fault identification model parameters optimized, and early warning judgment thresholds adjusted, enabling the early warning and diagnostic system to continuously improve its accuracy during actual operation.

[0070] Example 2: Figure 2 This is a schematic diagram of the structure of a hydropower plant equipment early warning and diagnosis system based on multi-source sensing, according to an embodiment of the present invention.

[0071] like Figure 2 As shown, a hydropower plant equipment early warning and diagnosis system based on multi-source sensing includes: an acquisition module, a feature extraction module, and a diagnosis module; The acquisition module is used to collect multi-source sensing data of the target device within the monitoring window and form a multi-source time series, which includes at least temperature field spatial distribution data, vibration signal data and device operating status data. The feature extraction module is used to obtain the equivalent heat source intensity distribution inside the target device based on the temperature field spatial distribution data within the current monitoring window, generate thermal resistance parameters for each segment of the heat transfer path of the target device, and identify abnormal thermal resistance segments; extract thermal response delay parameters based on the device operating status data, wherein the thermal response delay is decomposed into the response delay during the startup phase and the thermal fluctuation delay under steady-state conditions. The diagnostic module is used to extract the coupled operating condition feature vector of the current monitoring window, which is obtained by fusing the vibration signal data with temperature features including thermal resistance parameters and thermal response delay parameters. Based on the coupled operating condition feature vector, the fault type is identified and the diagnostic conclusion of the current window is output.

[0072] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0073] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0074] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0075] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0076] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0077] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0078] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for early warning and diagnosis of hydropower plant equipment based on multi-source sensing, characterized in that, include: S101, Collect multi-source sensing data of the target equipment within the monitoring window to form a multi-source time series, including at least temperature field spatial distribution data, vibration signal data and equipment operating status data; S102, Based on the spatial distribution data of the temperature field within the current monitoring window, obtain the equivalent heat source intensity distribution inside the target device, generate the thermal resistance parameters of each segment of the heat transfer path of the target device, and identify the abnormal thermal resistance segments. S103, extract thermal response delay parameters based on equipment operating status data, wherein the thermal response delay is decomposed into startup phase response delay and steady-state thermal fluctuation delay. S104, extract the coupled operating condition feature vector of the current monitoring window, which is obtained by fusing the vibration signal data with temperature features including thermal resistance parameters and thermal response delay parameters. Based on the coupled operating condition feature vector, identify the fault type and output the diagnostic conclusion of the current window.

2. The method for early warning and diagnosis of hydropower plant equipment based on multi-source sensing according to claim 1, characterized in that, The spatial distribution data of the temperature field is defined as: the temperature time series of each temperature measuring point of the target device. composition, , For discrete sampling times, , Let m be the set of temperature measurement points; The device operating status data is defined as follows: Where 0 represents the shutdown state, 1 represents the startup process state, and 2 represents the steady-state operation state.

3. The method for early warning and diagnosis of hydropower plant equipment based on multi-source sensing according to claim 2, characterized in that, S102 specifically includes: A1. Within the current monitoring window, utilize temperature time series. As a constraint, the average heat source intensity distribution of this window is obtained; A2. Divide the heat transfer path into several segments, and calculate the equivalent thermal resistance of each segment based on the temperature at both ends of each segment and the heat flow through that segment. A3. Compare the equivalent thermal resistance of each segment with the preset healthy baseline to identify segments with abnormal thermal resistance.

4. The method for early warning and diagnosis of hydropower plant equipment based on multi-source sensing according to claim 3, characterized in that, In section A2, the equivalent resistance of each segment is calculated as follows: For the i-th segment, the equivalent thermal resistance within this window The ratio of the temperature difference between the two ends of this section to the heat flow passing through it: in, Let i be the equivalent thermal resistance of the i-th segment. and These are the average temperatures at both ends of the segment within the monitoring window. The heat flow through this section is obtained by integrating the average heat source intensity of the window.

5. The method for early warning and diagnosis of hydropower plant equipment based on multi-source sensing according to claim 4, characterized in that, The identification of abnormal thermal resistance segments includes: When the rate of change of thermal resistance in a certain segment When the following conditions are met: The system determines that there is an abnormal thermal resistance in this segment and outputs the abnormal segment number: , This is the equivalent thermal resistance of this section. This serves as the preset healthy baseline for the thermal resistance of the i-th segment. The preset abnormal thermal resistance threshold is set based on expert experience and historical fault data.

6. The method for early warning and diagnosis of hydropower plant equipment based on multi-source sensing according to claim 5, characterized in that, The thermal response delay parameter is specifically: the steady-state thermal fluctuation delay of the current window. Startup phase response delay corresponding to the most recent startup event ; The startup phase response delay is the time difference between the target device starting from a cold state and the temperature at the key measuring point first reaching the thermal equilibrium threshold. The steady-state thermal fluctuation delay is the time difference corresponding to the peak value of the cross-correlation function between the load sequence and the temperature sequence during the steady-state operation phase.

7. The method for early warning and diagnosis of hydropower plant equipment based on multi-source sensing according to claim 6, characterized in that, The coupled operating condition feature vector includes the operating state features of the current window. Vibration characteristics and temperature characteristics; the vibration characteristics include at least the effective velocity value extracted from the vibration signal data. kurtosis factor Frequency band energy Any one or more; the temperature characteristics include startup phase response delay. Steady-state thermal fluctuation delay Maximum thermal resistance change rate in all segments The running state characteristics of the current window are defined as follows: That is, the state of most sampling points within the window. This indicates taking the mode.

8. The method for early warning and diagnosis of hydropower plant equipment based on multi-source sensing according to claim 7, characterized in that, The method of identifying fault types based on coupled operating condition feature vectors and outputting the diagnostic conclusion for the current window specifically includes: S201, Input the coupled operating condition feature vector into the pre-trained fault identification model and output the fault type; the fault type includes pure thermal fault, pure mechanical fault and mechanical-thermal coupled fault; S202, if the fault type is a pure thermal fault or a mechanical-thermal coupling fault, and there is a thermal resistance abnormal segment, the fault location is based on the output thermal resistance abnormal segment number. S203, Generate structured diagnostic conclusions for the current monitoring window; S204, based on the structured diagnostic conclusions of the monitoring window, triggers an alert when an abnormal state is determined.

9. The method for early warning and diagnosis of hydropower plant equipment based on multi-source sensing according to claim 8, characterized in that, The abnormal state is determined as follows: a state is determined to be abnormal when any of the following conditions are met: The identified fault types are either purely thermal faults or mechanical-thermal coupling faults; Any rate of change of thermal resistance >0.5; The thermal fluctuation delay under steady-state conditions showed a monotonically increasing trend within three consecutive monitoring windows.

10. A hydropower plant equipment early warning and diagnostic system based on multi-source sensing, characterized in that, The system of the early warning and diagnosis method for hydropower plant equipment based on multi-source sensing as described in any one of claims 1-9 includes: an acquisition module, a feature extraction module, and a diagnosis module; The acquisition module is used to collect multi-source sensing data of the target device within the monitoring window and form a multi-source time series, which includes at least temperature field spatial distribution data, vibration signal data and device operating status data. The feature extraction module is used to obtain the equivalent heat source intensity distribution inside the target device based on the temperature field spatial distribution data within the current monitoring window, generate thermal resistance parameters for each segment of the heat transfer path of the target device, and identify abnormal thermal resistance segments; extract thermal response delay parameters based on the device operating status data, wherein the thermal response delay is decomposed into the response delay during the startup phase and the thermal fluctuation delay under steady-state conditions. The diagnostic module is used to extract the coupled operating condition feature vector of the current monitoring window, which is obtained by fusing the vibration signal data with temperature features including thermal resistance parameters and thermal response delay parameters. Based on the coupled operating condition feature vector, the fault type is identified and the diagnostic conclusion of the current window is output.

Images