A method, device, equipment, medium and product for monitoring a generator stator cooling water leakage fault

Abstract

The present disclosure relates to a generator stator cooling water leakage fault monitoring method, device, equipment, medium and product, the method comprising: obtaining pressure pulsation data at multiple preset monitoring nodes in the stator cooling water circuit of the generator to be monitored; the pressure pulsation data is used to represent the fluctuation characteristics of the transient pressure wave generated in the stator cooling water circuit under the working condition of water flow rate change; based on the pressure pulsation data, the propagation characteristics of the transient pressure wave in the stator cooling water circuit are determined, and the pressure pulsation data at the multiple preset monitoring nodes is coherently synthesized based on the propagation characteristics to obtain target fluctuation characteristics; the phase space trajectory is obtained by phase space reconstruction of the target fluctuation characteristics, and the phase space feature vector is determined based on the wave mode mutation characteristics of the phase space trajectory; the monitoring result is determined based on the phase space feature vector; the monitoring result is used to represent the leakage position of the stator cooling water circuit. The present disclosure can realize high-sensitivity online accurate positioning of the leakage position in the stator cooling water circuit.

Description

Technical Field

[0001] This disclosure relates to the field of generator condition monitoring and fault diagnosis technology, and in particular to a method, device, equipment, medium and product for monitoring generator stator cooling water leakage faults. Background Technology

[0002] In the operation of large generators, the sealing performance of the stator cooling water circuit is crucial. Once leakage occurs, it can easily lead to insulation damage or even serious safety accidents. Therefore, it is urgent to monitor the operating status of the stator cooling water circuit in real time to achieve early warning of faults and ensure the safe and stable operation of the equipment.

[0003] In existing technologies, conventional monitoring schemes that combine distributed pressure sensors with threshold alarms are commonly used. However, this detection mechanism is relatively simple and is easily affected by fluctuations in normal operation under complex generator conditions, resulting in false alarms or missed alarms. Furthermore, the leakage location accuracy of existing monitoring methods is extremely low, making it difficult to meet the requirements for high-reliability online diagnosis. Summary of the Invention

[0004] This disclosure provides a method, device, equipment, medium, and product for monitoring generator stator cooling water leakage faults.

[0005] According to a first aspect of this disclosure, a method for monitoring generator stator cooling water leakage faults is provided, the method comprising: Acquire pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored; the pressure pulsation data is used to characterize the fluctuation characteristics of transient pressure waves generated in the constant cooling water circuit under the condition of water flow rate variation. Based on the pressure pulsation data, the propagation characteristics of the transient pressure wave in the constant cooling water circuit are determined, and the pressure pulsation data at the multiple preset monitoring nodes are coherently synthesized based on the propagation characteristics to obtain the target fluctuation characteristics. The phase space trajectory is obtained by reconstructing the target wave characteristics in phase space, and the phase space feature vector is determined based on the wave mode abrupt change characteristics of the phase space trajectory. The monitoring results are determined based on the phase space feature vectors; the monitoring results are used to characterize the leakage location of the constant cooling water circuit.

[0006] Furthermore, acquiring pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored includes: Pressure signals are captured by multi-channel resonant sensors installed at each of the preset monitoring nodes; Based on the real-time water flow rate change of the constant cooling water circuit, the resonant frequency of the multi-channel resonant sensor is determined, and the target monitoring frequency is obtained. The pressure signals are captured in real time by the target monitoring frequency to obtain the pressure pulsation data.

[0007] Further, determining the propagation characteristics of the transient pressure wave in the constant cooling water circuit based on the pressure pulsation data includes: Calculate the pressure change rate of the pressure pulsation data, and determine the fluctuation evolution characteristics of the transient pressure wave based on the pressure change rate; wherein, the fluctuation evolution characteristics include the first derivative and the second derivative of the pressure change rate; Based on the first derivative, the second derivative, and the pipe size parameters of the constant cooling water circuit, the propagation velocity components of the transient pressure wave on different propagation paths are determined. Based on the propagation velocity components and the topology of the constant cooling water circuit, a pressure wave propagation velocity matrix is ​​constructed; wherein, the pressure wave propagation velocity matrix is ​​used to characterize the propagation characteristics of the transient pressure wave in the constant cooling water circuit.

[0008] Furthermore, determining the fluctuation evolution characteristics of the transient pressure wave based on the pressure change rate includes: Identify the direction of pressure fluctuations in the pressure pulsation data; If the pressure fluctuation direction is positive, the forward difference algorithm is used to calculate the pressure change rate to obtain the second derivative characterizing the pressure acceleration trend. If the pressure fluctuation direction is negative, the backward difference algorithm is used to calculate the pressure change rate to obtain the second derivative characterizing the pressure deceleration trend.

[0009] Furthermore, the step of coherently synthesizing the pressure pulsation data at the multiple preset monitoring nodes based on the propagation characteristics to obtain the target fluctuation characteristics includes: The propagation direction of the transient pressure wave among the preset monitoring nodes is determined based on the propagation characteristics. Based on the propagation direction and the propagation speed component, the propagation time difference between the preset monitoring nodes is calculated; The pressure pulsation data is subjected to phase compensation processing based on the propagation time difference to obtain compensated pressure pulsation data. The compensated pressure pulsation data are superimposed to generate a synthetic wavefront signal, and the target fluctuation characteristics are determined based on the synthetic wavefront signal.

[0010] Further, the step of superimposing the compensated pressure pulsation data to generate a synthetic wavefront signal includes: Based on the propagation direction, the signals corresponding to different propagation directions in the compensated pressure pulsation data are respectively divided into forward propagation datasets and reverse propagation datasets; Based on the frequency response characteristics of the multi-channel resonant sensor, the forward propagation dataset and the backward propagation dataset are weighted by amplitude to obtain weighted multi-node component data. The weighted multi-node component data is superimposed to generate the synthetic wavefront signal.

[0011] Further, the step of reconstructing the phase space trajectory from the target wave characteristics includes: Identify the reference time interval of the fluctuation cycle in the target fluctuation feature, and divide the time into continuous time windows based on the reference time interval; Within each time window, the original state point is determined based on the peak time and amplitude of the target fluctuation characteristics, and a high-dimensional state point is constructed by combining the amplitude evolution components of each original state point within a preset span. By connecting the high-dimensional state points in chronological order, the phase space trajectory is obtained.

[0012] Further, determining the phase space feature vector based on the wave mode abrupt change characteristics of the phase space trajectory includes: The phase space trajectory is divided into multiple local trajectory segments, and multi-scale grid division is performed on each local trajectory segment to obtain the multi-scale grid distribution data corresponding to the local trajectory segment. Based on the multi-scale grid distribution data, the number of times each local trajectory segment crosses each multi-scale grid is counted, and the fractal dimension value of each local trajectory segment is calculated based on the number. The wave mode abrupt change feature is determined by detecting the change in the fractal dimension value of adjacent local trajectory segments; wherein the wave mode abrupt change feature is used to quantitatively characterize the spatial complexity anomaly of the phase space trajectory. The local trajectory segments that produce the abrupt change characteristics of the wave mode are identified as anomalous trajectory segments; High-dimensional state points are extracted from the abnormal trajectory segments, and the phase space feature vector is constructed based on the high-dimensional state points in the abnormal trajectory segments.

[0013] Further, determining the monitoring result based on the phase space feature vector includes: Calculate the spatial similarity between the phase space feature vector and the preset reference pattern vector to obtain the spatial distribution pattern encoding; The spatial distribution pattern encoding is matched with a pre-stored leakage location pattern library to obtain the overlap matching result; The monitoring results are determined based on the overlap matching results.

[0014] Further, determining the monitoring result based on the overlap matching result includes: The region identifiers of each of the fixed cooling water loops are sorted in descending order of the overlap matching results to obtain the loop region sorting sequence; Extract the overlap degree corresponding to the region identifier ranked first in the loop region sorting sequence to obtain the target overlap degree set; The values ​​in the target overlap set are subjected to probability transformation processing to obtain the probability distribution corresponding to each region identifier, and the monitoring results are constructed based on the probability distribution.

[0015] According to a second aspect of this disclosure, a monitoring device for generator stator cooling water leakage faults is provided, the device comprising: The acquisition module is used to acquire pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored; the pressure pulsation data is used to characterize the fluctuation characteristics of transient pressure waves generated in the constant cooling water circuit under the condition of water flow rate variation. The first determining module is used to determine the propagation characteristics of the transient pressure wave in the constant cooling water circuit based on the pressure pulsation data, and to coherently synthesize the pressure pulsation data at the multiple preset monitoring nodes based on the propagation characteristics to obtain the target fluctuation characteristics. The second determining module is used to reconstruct the phase space of the target wave characteristics to obtain the phase space trajectory, and to determine the phase space feature vector based on the wave mode abrupt change characteristics of the phase space trajectory. The monitoring module is used to determine the monitoring results based on the phase space feature vector; the monitoring results are used to characterize the leakage location of the constant cooling water circuit.

[0016] According to a third aspect of this disclosure, an electronic device is provided. The electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described above.

[0017] According to a fourth aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the methods described above.

[0018] According to a fifth aspect of this disclosure, a computer program product is provided. The computer program product includes a computer program that, when executed by a processor, implements the methods described above.

[0019] This disclosure provides a method, apparatus, equipment, medium, and product for monitoring generator constant cooling water leakage faults. First, pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored are acquired. The pressure pulsation data characterizes the fluctuation characteristics of transient pressure waves generated in the constant cooling water circuit under varying water flow rates. Then, based on the pressure pulsation data, the propagation characteristics of the transient pressure waves in the constant cooling water circuit are determined, and the pressure pulsation data at multiple preset monitoring nodes are coherently synthesized based on the propagation characteristics to obtain target fluctuation characteristics. Next, the target fluctuation characteristics are reconstructed in phase space to obtain a phase space trajectory, and the phase space feature vector is determined based on the wave mode abrupt change characteristics of the phase space trajectory. Finally, the monitoring result is determined based on the phase space feature vector. The monitoring result characterizes the leakage location in the constant cooling water circuit.

[0020] As described above, this embodiment of the present disclosure effectively captures the transient pressure wave fluctuation characteristics generated under varying water flow velocity conditions by acquiring pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator under test. This overcomes the limitations of traditional static pressure thresholds, which are sensitive to dynamic noise and extremely insensitive to early minor leaks. Furthermore, this embodiment of the present disclosure, based on in-depth analysis of pressure pulsation data, determines the dynamic propagation characteristics of transient pressure waves in complex pipelines. Accordingly, it coherently synthesizes the pressure pulsation data at multiple preset monitoring nodes, effectively overcoming pipeline propagation delay and phase deviation. While accurately analyzing the reflection and superposition effects of pressure waves, it filters background noise, thereby significantly enhancing the signal-to-noise ratio of the obtained target fluctuation characteristics. Subsequently, the phase space trajectory is obtained by reconstructing the target fluctuation characteristics in phase space, and the phase space feature vector is determined based on the wave mode mutation characteristics of the phase space trajectory. This process deeply explores the dynamic anomalies of the pressure wave in the nonlinear evolution process, and realizes the accurate separation of the real leakage signal from the fluctuations of the system under normal operating conditions. Finally, the monitoring results are determined based on the phase space feature vector, and the probability distribution representing the leakage location is output. This completely eliminates the traditional coarse differential alarm that relies on thresholds and leads to fuzzy positioning. The embodiments of this disclosure achieve high-sensitivity identification and high-precision spatial location tracing of leakage points in the fixed cooling water circuit without the need for extensive pre-training, and comprehensively improves the accuracy, timeliness and anti-interference capability of generator online diagnosis. Attached Figure Description

[0021] The above and other objects, features, and advantages of this disclosure will become more apparent from the more detailed description of the embodiments thereof in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this disclosure and form part of the specification. They are used together with the embodiments of this disclosure to explain the disclosure and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0022] Figure 1A flowchart of a method for monitoring generator stator cooling water leakage faults provided as an exemplary embodiment of this disclosure; Figure 2 One of the flowcharts for a method for monitoring generator stator cooling water leakage faults provided as another exemplary embodiment of this disclosure; Figure 3 A second flowchart of a method for monitoring generator constant cooling water leakage faults provided as another exemplary embodiment of this disclosure; Figure 4 A third flowchart of a method for monitoring generator stator cooling water leakage faults provided as another exemplary embodiment of this disclosure; Figure 5 A flowchart of a method for monitoring generator stator cooling water leakage faults provided as another exemplary embodiment of this disclosure; Figure 6 Fifth flowchart of a method for monitoring generator constant cooling water leakage faults provided as another exemplary embodiment of this disclosure; Figure 7 A flowchart of a method for monitoring generator stator cooling water leakage faults provided as another exemplary embodiment of this disclosure; Figure 8 A schematic diagram illustrating the entire process of a generator constant cooling water leakage monitoring method provided as an exemplary embodiment of this disclosure; Figure 9 A schematic block diagram of the functional modules of a generator constant cooling water leakage monitoring device provided as an exemplary embodiment of this disclosure; Figure 10 A structural block diagram of an electronic device provided as an exemplary embodiment of this disclosure; Figure 11 A structural block diagram of a computer system provided as an exemplary embodiment of this disclosure; Figure 12 A structural block diagram of a computer program product provided for an exemplary embodiment of this disclosure. Detailed Implementation

[0023] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0024] It should be understood that the steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0025] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below. It should be noted that the concepts of "first", "second", etc., used in this disclosure are only used to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0026] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more". The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0027] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0028] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this disclosed technical solution, based on the prompt message.

[0029] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device. It is understood that the above notification and user authorization process is merely illustrative and does not constitute a limitation on the implementation of this disclosure; other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0030] In one embodiment, such as Figure 1 As shown, a method for monitoring generator stator cooling water leakage faults is provided, including the following steps: Step 101: Obtain pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored.

[0031] Here, the executing entity can acquire pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored. The pressure pulsation data is used to characterize the fluctuation characteristics of transient pressure waves generated in the constant cooling water circuit under the condition of water flow rate variation.

[0032] In one possible embodiment, such as Figure 2 As shown, acquiring pressure pulsation data at multiple preset monitoring nodes in the stator cooling water circuit of the generator to be monitored includes the following steps: Step 201: Capture pressure signals using multi-channel resonant sensors installed at each preset monitoring node.

[0033] Here, the executing entity can capture pressure signals through multi-channel resonant sensors set at each preset monitoring node.

[0034] In one possible embodiment, the executing entity can select and determine preset monitoring nodes on the critical path of pressure wave propagation based on the topology of the generator's chilled water circuit and the results of fluid dynamics simulation. Specifically, the preset monitoring nodes are locations in the chilled water circuit that are sensitive to the dynamic characteristics of the fluid, including the main inlet, the main return outlet, the branch flow points, and weak points such as sealing interfaces that are prone to leakage.

[0035] For example, the preset monitoring nodes can be specific key nodes such as the stator winding water inlet header (or stator water inlet main pipe) J1, the rotor cooling water branch tee (or rotor water inlet branch) J2 / J3, and the end cover return water collection pipe (or end cooling water jacket) J4. At these nodes, the execution subject captures pressure signals in real time through deployed multi-channel resonant sensors. The multi-channel resonant sensor is a pressure sensing device based on the mechanical resonance principle that can monitor multiple measuring points simultaneously. Its core is a piezoelectric crystal resonant cavity. The natural frequency of the cavity will shift with the fluctuation of external fluid pressure, thereby enabling the execution subject to capture the underlying physical pressure signal in real time.

[0036] For example, in a practical application, in the stator cooling water circuit of a 600MW coal-fired generator unit, technicians first selected four key monitoring nodes based on fluid simulation results, including the stator winding inlet water header J1, the rotor cooling water branch tee J2 / J3, and the end cover return water collection pipe J4. Multi-channel resonant sensors were installed at each node. The execution unit used the multi-channel resonant sensors set at each preset monitoring node (J1, J2, J3, J4) to capture the pressure signal inside the stator cooling water circuit in real time, providing a physical signal basis for subsequent frequency adaptive adjustment and pulsation data extraction.

[0037] Step 202: Based on the real-time water flow rate change of the constant cooling water circuit, determine the resonant frequency of the multi-channel resonant sensor to obtain the target monitoring frequency.

[0038] Here, after capturing pressure signals through multi-channel resonant sensors set at each preset monitoring node, the executing entity can determine the resonant frequency of the multi-channel resonant sensors based on the real-time water flow rate change of the constant cooling water circuit, and obtain the target monitoring frequency.

[0039] In one possible embodiment, the multi-channel resonant sensor integrates a miniature turbine velocity meter. The actuator can use this velocity meter to detect and obtain the cooling water flow velocity v(t) at the sensor's installation location in real time. Subsequently, the actuator, based on the obtained real-time water flow velocity v(t), applies a preset Strouhal number relationship. The target monitoring frequency is obtained by dynamically calculating and adaptively adjusting the resonant frequency, where k is a preset dimensionless coefficient and d is the inner diameter of the flow channel at the sensor installation location. By dynamically adjusting the excitation frequency of the piezoelectric crystal, the actuator ensures that the sensor always works in the sensitive range of the resonant peak, which solves the problem of decreased sensitivity or signal distortion of traditional fixed-frequency sensors under variable flow conditions and significantly improves the ability to capture transient pressure waves caused by minute leaks.

[0040] In practical applications, when the load of the aforementioned 600MW coal-fired power generating unit increases from 50% to 100%, the cooling water flow rate increases from 1.0m / s to 1.8m / s. At this time, the turbine flow meter built into the sensor of node J2 detects the change in flow rate in real time. The main controller controls its embedded processor to dynamically calculate the new frequency value of 23kHz according to the preset parameters k=0.23 and d=0.018m. Then, it drives the piezoelectric exciter to adjust the resonant cavity reference frequency from 12.78kHz when the load is 50% to 23kHz.

[0041] For example, in another operating condition, when the load of a generator set increases from 300MW to 450MW, the cooling water flow rate increases from 1.0m / s to 1.5m / s. After detecting the change in flow rate, the actuator dynamically calculates and drives the piezoelectric exciter to adjust the target monitoring frequency according to the preset parameters k=0.25 and d=0.02m. .

[0042] Step 203: Real-time capture of each pressure signal is performed using the target monitoring frequency to obtain pressure pulsation data.

[0043] Here, after obtaining the target monitoring frequency, the executing entity can capture each pressure signal in real time through the target monitoring frequency to obtain pressure pulsation data.

[0044] In one possible embodiment, after adjusting the sensor to the target monitoring frequency, the actuator captures the instantaneous offset of the cavity resonant frequency using its internal high-resolution frequency counter. Subsequently, the executing entity uses a pre-calibrated conversion coefficient K to reverse-calculate and convert the frequency offset into a specific pressure pulsation value P(t). The specific conversion formula is as follows: Finally, the execution entity generates and outputs a triplet sequence containing timestamps, pressure amplitudes, and resonant frequency offsets as pressure pulsation data to capture microsecond-level transient pressure fluctuations.

[0045] It should be noted that, in order to ensure that the spatiotemporal correlation of transient pressure events is fully preserved, the execution entity controls all channels to synchronously record multi-channel pressure pulsation time-series data at a sampling rate greater than 5kHz. This provides high-fidelity, time-domain aligned raw data for the subsequent construction of the pressure wave propagation matrix, laying the physical foundation for leak location.

[0046] For example, in practical applications, following the aforementioned frequency adjustment scenario of the 600MW unit, 3 seconds after the frequency adjustment, due to simulated leakage, the pressure fluctuation caused by a 0.5mm orifice leak downstream of node J3 is transmitted to node J2, and the sensor frequency counter captures the instantaneous shift in the resonant frequency. The implementing entity combines the factory calibration coefficient. Converting frequency offset into pressure pulsation value Ultimately, the executing unit transmitted the abnormal event along with a timestamp to the central processing unit at a sampling rate of 5kHz, fully recording the transient leakage characteristics under this drastic flow rate change condition. Similarly, in the example of the load increasing to 450MW mentioned above, 2 seconds after the adjustment, pressure fluctuations were caused by simulated leakage downstream of node J2, and the executing unit captured the instantaneous frequency shift through sensors. Combined with factory calibration coefficients The executing entity calculates the pressure pulsation value at this time. Finally, the abnormal event, along with its timestamp, is synchronously output as multi-channel data at a sampling rate of 5kHz.

[0047] In this embodiment, firstly, the executing entity captures pressure signals using multi-channel resonant sensors installed at each preset monitoring node; then, based on the real-time water flow rate change of the constant cooling water circuit, the executing entity determines the resonant frequency of the multi-channel resonant sensors to obtain the target monitoring frequency; finally, the executing entity captures each pressure signal in real time using the target monitoring frequency to obtain pressure pulsation data.

[0048] As described above, this embodiment captures pressure signals by setting up multi-channel resonant sensors at each preset monitoring node. Based on the real-time water flow rate changes in the constant cooling water circuit, the resonant frequency of the sensors is adaptively determined to obtain the target monitoring frequency. This ensures that the multi-channel resonant sensors always operate within the optimal sensitive range of the resonant peak. This adaptive frequency modulation mechanism effectively solves the defects of traditional fixed-frequency sensors, such as sensitivity decrease or signal distortion, that easily occur under variable flow rate conditions. Ultimately, this embodiment captures each pressure signal in real time through dynamically matched target monitoring frequencies, significantly improving the system's sensitivity to capturing microsecond-level transient pressure fluctuations caused by early minor leaks. This provides high-fidelity, time-domain aligned raw physical data support for subsequently constructing accurate pressure wave propagation characteristic models and achieving precise leak point location.

[0049] Step 102: Based on the pressure pulsation data, determine the propagation characteristics of the transient pressure wave in the constant cooling water circuit, and based on the propagation characteristics, coherently synthesize the pressure pulsation data at multiple preset monitoring nodes to obtain the target fluctuation characteristics.

[0050] Here, after obtaining pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored, the executing entity can determine the propagation characteristics of transient pressure waves in the constant cooling water circuit based on the pressure pulsation data, and coherently synthesize the pressure pulsation data at multiple preset monitoring nodes based on the propagation characteristics to obtain the target fluctuation characteristics.

[0051] In one possible embodiment, such as Figure 3 As shown, based on pressure pulsation data, the propagation characteristics of transient pressure waves in a constant cooling water circuit are determined, including the following steps: Step 301: Calculate the pressure change rate of the pressure pulsation data, and determine the fluctuation evolution characteristics of the transient pressure wave based on the pressure change rate.

[0052] Here, after the execution entity obtains the pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored, it can calculate the pressure change rate of the pressure pulsation data and determine the fluctuation evolution characteristics of the transient pressure wave based on the pressure change rate. The fluctuation evolution characteristics include the first derivative and the second derivative of the pressure change rate.

[0053] In one possible embodiment, determining the fluctuation evolution characteristics of transient pressure waves based on the rate of pressure change includes the following steps: Identify the direction of pressure fluctuations in pressure pulsation data; If the pressure fluctuation direction is positive, the forward difference algorithm is used to calculate the pressure change rate and obtain the second derivative characterizing the pressure acceleration trend. If the pressure fluctuation direction is negative, the backward difference algorithm is used to calculate the pressure change rate and obtain the second derivative characterizing the pressure deceleration trend.

[0054] Specifically, after acquiring pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored, the executing entity first identifies the pressure fluctuation direction of the pressure pulsation data. If the pressure fluctuation direction is positive, the forward difference algorithm is used to calculate the pressure change rate to obtain the second derivative representing the pressure acceleration trend. If the pressure fluctuation direction is negative, the backward difference algorithm is used to calculate the pressure change rate to obtain the second derivative representing the pressure deceleration trend.

[0055] In one possible embodiment, the executing entity first identifies the pressure fluctuation direction at each sampling point in the pressure pulsation data. The pressure fluctuation direction refers to the dynamic characteristic of determining whether the pressure is in an upward or downward phase by comparing the pressure values ​​of adjacent sampling points, which is used to distinguish the propagation phase of the pressure wave. Specifically, the executing entity compares the current pressure value with the numerical relationship of the preceding and following sampling points point by point. If the executing entity determines that the pressure at a certain point is both higher than the previous point and lower than the following point, showing an upward trend, it is marked as a positive fluctuation; if the executing entity determines that the pressure at a certain point is both lower than the previous point and higher than the following point, showing a downward trend, it is marked as a negative fluctuation. For example, for the pressure sequence [100.5, 101.2, 101.0] kPa, the executing entity identifies that the second point 101.2 is higher than the previous value 100.5 but lower than the following value 101.0, so it is marked as a negative fluctuation.

[0056] Next, for different pressure fluctuation directions, the executing entity adopts differentiated differential calculation strategies: For positive pressure fluctuation data, the executing entity uses forward differencing to calculate the forward first-order differential value of the pressure change rate between the current sampling point and subsequent sampling points within a set time window. Forward differencing is a numerical calculation method for positive fluctuations. The executing entity calculates the pressure rise rate by dividing the pressure difference between the current point and points within the future time window by the time interval. Specifically, the executing entity calculates the pressure change rate between the current point and five future sampling points using the following formula: Forward first-order differential value = (average of n future points - current point) / time window width. For example, if the current point P(t) = 102.6 kPa and the average pressure of the next five points is 103.3 kPa, the executing entity calculates the forward first-order differential value = (103.3 - 102.6) / 0.005 = 140 kPa / s. Subsequently, based on the forward first-order differential value of the positive pressure fluctuation data, the executing entity further calculates the second-order differential value indicating an accelerating pressure change trend. For example, when the differential value is 140 kPa / s at time t, and rises to 155 kPa / s at time t+0.2 ms, the executing entity calculates the second-order differential value with a single sampling interval of 0.2 ms based on the time step. The corresponding second-order differential value is (155-140) / 0.0002 = 75,000 kPa / s².

[0057] In one possible embodiment, for negative pressure fluctuation data, the executing entity uses backward difference to calculate the backward first derivative of the pressure change rate between the current sampling point and the sampling points within a previously set time window. Backward difference is for negative fluctuations. The executing entity calculates the pressure decrease rate by the pressure difference between the current point and the points within the historical time window. The executing entity takes the current point and the previous 5 sampling points for calculation. The calculation formula is: backward first derivative = (current point - average of the previous n points) / time window width. For example, if the current point P(t) = 99.8 kPa and the average of the previous 5 points is 100.5 kPa, the executing entity calculates the backward first derivative = (99.8 - 100.5) / 0.005 = -140 kPa / s. The negative sign indicates a downward trend. Then, based on the backward first derivative of the negative pressure fluctuation data, the executing entity calculates the second derivative of the pressure change trend, which shows a decelerating trend. For example, when the differential value changes from -140 kPa / s to -130 kPa / s, causing the absolute value to decrease, the execution entity calculates the corresponding second-order differential value = [-130 - (-140)] / 0.0002 = 50,000 kPa / s², where a positive value indicates that the deceleration trend is weakening.

[0058] For example, in practical applications, when processing the pressure data of pipe segment J1-J2, when the execution unit monitors that the pressure value of node J2 in the generator stator cooling water circuit drops from 101.8 kPa to 101.2 kPa and then to 100.9 kPa at t=3.205 seconds, the execution unit identifies the intermediate point of 101.2 kPa as a negative fluctuation point. Subsequently, the execution unit performs backward differential calculation on this point, taking the average pressure of the first 5 sampling points as 102.1 kPa, and calculates the backward first derivative value = (101.2-102.1) / 0.005 = -180 kPa / s. Then, the execution... The execution entity calculates the acceleration based on the differential value sequence. Given that the differential value at the first 0.2 milliseconds is -190 kPa / s, the execution entity calculates the second derivative = [-180 - (-190)] / 0.0002 = 50,000 kPa / s². At the same time, the execution entity performs forward calculation on the forward fluctuation data of the J2-J1 pipe segment during the same time period, calculating the forward first derivative value = (101.5 - 100.8) / 0.005 = 140 kPa / s, and its second derivative value = (145 - 140) / 0.0002 = 25,000 kPa / s².

[0059] Step 302: Based on the first-order differential, second-order differential, and pipe size parameters of the constant cooling water circuit, determine the propagation velocity components of the transient pressure wave on different propagation paths.

[0060] Here, after determining the fluctuation evolution characteristics of the transient pressure wave based on the pressure change rate, the executing entity can determine the propagation velocity components of the transient pressure wave on different propagation paths based on the first-order differential, the second-order differential, and the pipe size parameters of the constant cooling water circuit.

[0061] In one possible embodiment, the propagation velocity components of transient pressure waves along different propagation paths are determined based on the first-order differential, the second-order differential, and the pipe size parameters of the constant cooling water loop, including the following steps: Extract the equivalent hydraulic diameter of the pipes between adjacent key nodes in the constant cooling water circuit; Multiply the forward first derivative of the forward pressure fluctuation data by the equivalent hydraulic diameter of the corresponding pipe segment to obtain the forward pressure propagation intensity factor, and divide the second derivative of the pressure acceleration trend by the equivalent hydraulic diameter of the corresponding pipe segment to obtain the forward pressure acceleration factor. Divide the forward pressure propagation intensity factor by the forward pressure acceleration factor to obtain the propagation velocity component of the pressure wave on the forward propagation path. The negative pressure wave propagation intensity factor is obtained by multiplying the first derivative of the negative pressure wave data by the equivalent hydraulic diameter of the corresponding pipe segment. The negative pressure acceleration factor is obtained by dividing the second derivative of the pressure deceleration trend by the equivalent hydraulic diameter of the corresponding pipe segment. The propagation velocity component of the pressure wave on the reverse propagation path is obtained by dividing the negative pressure wave propagation intensity factor by the negative pressure acceleration factor.

[0062] Specifically, the executing entity first calls the internally stored database of pipe physical parameters to extract the equivalent hydraulic diameter of the pipes between adjacent key nodes in the constant cooling water circuit. The equivalent hydraulic diameter is a key parameter reflecting the influence of the pipe's cross-sectional shape on the fluid throughput capacity. For non-circular pipes, the calculated value is 4 times the cross-sectional area divided by the perimeter of the pipe wall in contact with the fluid. For example, for a rectangular cooling water pipe with a width of 30mm and a height of 20mm, the effective hydraulic diameter of the pipe is calculated to be 0.024m by the executing entity according to the formula.

[0063] Next, the executing entity integrates the physical parameters with the differential values ​​obtained from the aforementioned calculations to calculate the propagation velocity component: For positive waves, the executing entity multiplies the forward first-order differential value by the equivalent hydraulic diameter of the pipe to obtain the positive pressure propagation intensity factor, divides the second-order differential value by the equivalent hydraulic diameter of the pipe to obtain the positive pressure acceleration factor, and finally divides the propagation intensity factor by the acceleration factor to obtain the positive propagation velocity component. The propagation intensity factor is obtained by multiplying the first-order differential value by the hydraulic diameter, representing the energy of the pressure wave propagating in the pipe, and the acceleration factor is obtained by dividing the second-order differential value by the hydraulic diameter. The propagation velocity component describes the rate of change of the pressure wave. It is calculated by dividing the propagation intensity factor by the acceleration factor and directly reflects the actual propagation speed of the pressure wave in a specific direction. It is the core physical quantity for constructing the pressure wave propagation matrix. For negative waves, the calculation logic is the same. The execution subject multiplies the absolute value of the backward first-order differential by the equivalent hydraulic diameter of the pipe to obtain the reverse pressure propagation intensity factor, divides the second-order differential by the equivalent hydraulic diameter of the pipe to obtain the reverse pressure acceleration factor, and finally divides the propagation intensity factor by the acceleration factor to obtain the reverse propagation velocity component.

[0064] For example, in a practical application, in a generator stator cooling water circuit, the executing entity calls the pipeline database to confirm that the J2-J3 section is a DN40 circular pipe with an equivalent hydraulic diameter of 0.04m. Then, the executing entity calculates the negative propagation component, and calculates the propagation intensity factor as 7.2kPa·m / s and the acceleration factor as 1,250,000kPa / (s²·m). Finally, the velocity component is obtained as 7.2 / 1,250,000 = 5.76m / s. At the same time, the executing entity performs positive calculation on the positive fluctuation data of the J2-J1 pipe section in the same time period. Combining the equivalent hydraulic diameter of 0.05m, the positive velocity is obtained as (140×0.05) / (25,000 / 0.05) = 14m / s.

[0065] Step 303: Based on the propagation velocity components and the topology of the constant cooling water loop, construct the pressure wave propagation velocity matrix.

[0066] Here, after determining the propagation velocity components of the transient pressure wave along different propagation paths, the executing entity can construct a pressure wave propagation velocity matrix based on the propagation velocity components and the topology of the constant cooling water circuit. The pressure wave propagation velocity matrix is ​​used to characterize the propagation characteristics of the transient pressure wave in the constant cooling water circuit.

[0067] In one possible embodiment, the pressure wave propagation velocity matrix is ​​constructed based on the propagation velocity components and the topology of the constant cooling water loop, including the following steps: Based on the spatial distribution constraints of the generator's stator cooling water circuit, the propagation velocity components of the forward and reverse propagation paths between all adjacent key nodes are combined to form a pressure wave propagation velocity matrix.

[0068] Specifically, after obtaining the forward and reverse pressure wave propagation velocity components of the pipes between all adjacent nodes, the executing entity constructs a pressure wave propagation velocity matrix based on the actual physical connection relationship of the generator stator cooling water circuit. The executing entity determines the water flow direction reference for each pipe segment according to the circuit topology diagram. For example, if node J1 to J2 is defined as the forward path, then J2 to J1 automatically corresponds to the reverse path. Then, the executing entity fills the forward propagation velocity component into the "forward velocity" column of the "J1-J2" row in the matrix, and fills the reverse propagation velocity component into the "reverse velocity" column of the same row. For multi-branch circuits, the executing entity processes each connection path in turn, and finally generates an N-row × 2-column matrix.

[0069] For example, in practical applications, the executing entity fills the positive velocity of 14 m / s of pipe segment J2-J1 into the "positive" column of the matrix according to the topology diagram, and fills the negative velocity of 5.76 m / s of pipe segment J2-J3 into the "negative" column, forming a local matrix. Through this scheme, the propagation velocity matrix constructed by the executing entity accurately characterizes the transmission efficiency of pressure waves in different pipe segments and different flow directions, providing key spatial dynamic characteristics for tracing the source of leaks, while avoiding dependence on absolute pressure thresholds and significantly enhancing the ability to analyze microsecond-level transient signals.

[0070] In this embodiment, firstly, the execution entity calculates the pressure change rate of the pressure pulsation data and determines the fluctuation evolution characteristics of the transient pressure wave based on the pressure change rate; then, based on the first-order differential, second-order differential, and pipe size parameters of the constant cooling water loop, the execution entity determines the propagation velocity components of the transient pressure wave on different propagation paths; finally, based on the propagation velocity components and the topology of the constant cooling water loop, the execution entity constructs a pressure wave propagation velocity matrix; wherein, the pressure wave propagation velocity matrix is ​​used to characterize the propagation characteristics of the transient pressure wave in the constant cooling water loop.

[0071] As described above, this embodiment determines the fluctuation evolution characteristics of transient pressure waves by calculating the first and second derivatives of pressure pulsation data. This accurately captures the dynamic acceleration and deceleration trends of transient pressure waves at different propagation stages, overcoming the limitations of traditional monitoring that relies solely on absolute static pressure amplitude. Furthermore, this embodiment deeply integrates this dynamic differential characteristic with the physical properties of the underlying cooling water loop to calculate the propagation velocity component, achieving a precise mapping between the abstract pressure signal and the actual physical structure of the pipe network. This accurately reflects the propagation energy and actual speed of the pressure wave along a specific path. Finally, the pressure wave propagation velocity matrix constructed based on the propagation velocity components of each path and the loop topology can globally and multidimensionally characterize the dynamic spatial propagation characteristics of transient pressure waves in complex cooling water pipe networks. This mechanism effectively avoids dependence on absolute pressure thresholds, providing solid and reliable spatial dynamic feature support for subsequent high-precision tracing and accurate location of leakage faults.

[0072] In one possible embodiment, such as Figure 4 As shown, pressure pulsation data from multiple preset monitoring nodes are coherently synthesized based on propagation characteristics to obtain target fluctuation characteristics, including the following steps: Step 401: Determine the propagation direction of the transient pressure wave between each preset monitoring node based on the propagation characteristics.

[0073] Here, after determining the propagation characteristics of the transient pressure wave in the constant cooling water circuit based on the pressure pulsation data, the executing entity can determine the propagation direction of the transient pressure wave between each preset monitoring node based on the propagation characteristics.

[0074] In one possible embodiment, the executing entity extracts the velocity direction characteristics of each pipe segment from the constructed pressure wave propagation velocity matrix, and combines this with the arrival time sequence of wavefront characteristics (such as the initial pressure abrupt change point) in the pressure pulsation data at each preset monitoring node to jointly determine the actual propagation direction of the transient pressure wave. Specifically, the executing entity sets a reference frame based on the reference water flow direction of the constant cooling water circuit. If the executing entity detects that the same characteristic wave peak arrives at the upstream node first and then the downstream node, it determines that the propagation direction of the transient pressure wave is downstream propagation (forward); conversely, if the executing entity detects that the characteristic wave peak arrives at the downstream node first and then the upstream node, or identifies a rebound signal of phase reversal of the pressure wave at the closed end, it determines that its propagation direction is countercurrent propagation (reverse).

[0075] For example, in actual monitoring, the executing entity, in conjunction with the propagation velocity matrix, discovers that an abnormal pressure wave peak reaches node J1 at t=2.100s and reaches the downstream node J2 at t=2.101s. Based on the reference water flow direction from J1 to J2 at this time, the executing entity determines that the propagation direction of the pressure wave in the J1-J2 pipe section is positive.

[0076] Step 402: Calculate the propagation time difference between preset monitoring nodes based on the propagation direction and propagation speed components.

[0077] Here, after determining the propagation direction of the transient pressure wave between each preset monitoring node based on the propagation characteristics, the executing entity can calculate the propagation time difference between the preset monitoring nodes based on the propagation direction and propagation velocity components.

[0078] In one possible embodiment, the executing entity first retrieves the three-dimensional physical space model of the generator constant cooling water circuit from the pre-existing database, obtains the actual physical length of the pipeline between adjacent preset monitoring nodes, and the physical distance of each node from the sealing structure in the circuit. Then, for downstream propagation (forward), the executing entity divides the physical length of the pipeline between the target node and the upstream node by the corresponding forward propagation velocity component to calculate the downstream propagation time difference. For example, if the physical length of the J1 to J2 pipe section is 1.8m and the forward propagation velocity component is 18m / s, then the executing entity calculates the downstream propagation time difference as 0.1s. For upstream propagation (reverse), the executing entity divides the round-trip physical path length from the target node to the leakage reflection point (sealing structure) by the reverse propagation velocity component to calculate the reverse propagation time difference. For example, if the distance of node J2 from the downstream end cap reflection point is 2.7m and the reverse propagation velocity component is 15m / s, then the executing entity calculates the reverse propagation time difference as 0.18s.

[0079] Step 403: Perform phase compensation processing on the pressure pulsation data based on the propagation time difference to obtain compensated pressure pulsation data.

[0080] Here, after calculating the propagation time difference between preset monitoring nodes based on the propagation direction and propagation speed components, the executing entity can perform phase compensation processing on the pressure pulsation data based on the propagation time difference to obtain the compensated pressure pulsation data.

[0081] In one possible embodiment, the executing entity uses the calculated downstream and upstream propagation time differences to perform directional differential translation operations on the raw pressure pulsation data of each preset monitoring node in the time domain. Specifically, for the forward propagation dataset, the executing entity performs phase delay compensation (i.e., translation backward on the time axis) on the pressure wave signal of the upstream node according to the downstream propagation time difference. For example, if the original signal J1 has a peak at t=3.000 seconds, the executing entity delays it by 0.1 seconds to t=3.100 seconds to align it with the signal of the downstream node J2 on the time axis. For the reverse propagation dataset, the executing entity performs phase advance compensation (i.e., forward on the time axis) on the pressure wave signal of the node associated with the reflection point according to the upstream propagation time difference. For example, if the original peak of J3 is at t=3.180 seconds, the executing entity advances it by 0.18 seconds to t=3.000 seconds to align it with the J2 signal. Ultimately, the data of all nodes achieve strict wavefront alignment in the feature space, resulting in compensated pressure pulsation data that eliminates propagation time difference errors.

[0082] Step 404: Superimpose the compensated pressure pulsation data to generate a synthetic wavefront signal, and determine the target fluctuation characteristics based on the synthetic wavefront signal.

[0083] Here, the executing entity performs phase compensation processing on the pressure pulsation data based on the propagation time difference. After obtaining the compensated pressure pulsation data, the compensated pressure pulsation data can be superimposed to generate a synthetic wavefront signal, and the target fluctuation characteristics can be determined based on the synthetic wavefront signal.

[0084] In one possible embodiment, the compensated pressure pulsation data are superimposed to generate a synthetic wavefront signal, including the following steps: Based on the propagation direction, the signals corresponding to different propagation directions in the compensated pressure pulsation data are divided into forward propagation datasets and reverse propagation datasets. Based on the frequency response characteristics of the multi-channel resonant sensor, the forward propagation dataset and the backward propagation dataset are weighted by amplitude to obtain weighted multi-node component data. The weighted multi-node component data are superimposed to generate a synthetic wavefront signal.

[0085] Specifically, after performing phase compensation processing on the pressure pulsation data based on the propagation time difference to obtain the compensated pressure pulsation data, the executing entity first divides the signals corresponding to different propagation directions in the compensated pressure pulsation data into forward propagation datasets and reverse propagation datasets based on the propagation direction. Then, based on the frequency response characteristics of the multi-channel resonant sensor, the executing entity performs amplitude weighting on the forward propagation datasets and the reverse propagation datasets respectively to obtain weighted multi-node component data. After that, the executing entity performs superposition processing on the weighted multi-node component data to generate a synthetic wavefront signal. Finally, the executing entity system automatically calculates the dynamic baseline threshold based on historical normal pressure wave data and extracts the wave crest feature points in the synthetic wavefront signal whose amplitude exceeds the dynamic baseline threshold. These points are then grouped and combined according to the propagation direction to form an enhanced wavefront feature vector as the determined target wave feature.

[0086] In one possible embodiment, continuing the example from the generator operation phase described above, the executing entity first identifies that the pressure wave at target node J3 includes both forward (J2 to J3) and reverse (reflected from sealed ends J3 to J4), and accordingly divides the forward and reverse datasets: For the forward dataset, the execution entity calculates the downstream time difference of approximately 0.114 seconds for the J2 to J3 pipe segment (1.6m) based on a forward propagation speed of 14m / s. The peak of the original J2 signal at t=3.000 seconds is then delayed by 114 milliseconds to t=3.114 seconds. For the reverse dataset, the distance from J3 to the pump reflector is measured to be 1.4m. Based on a reverse speed of 12m / s, the reverse time difference is calculated to be approximately 0.117 seconds. The J4 signal is then advanced by 117 milliseconds. After compensation, the execution entity analyzes the J2 delayed signal, assigning a weight of 0.85 to its main frequency of 130Hz, and corrects the amplitude from 2.15kPa to 1.83kPa. The J4 signal is then advanced. With a main frequency of 170Hz and a weight of 0.78, the amplitude of 1.5kPa is corrected to 1.17kPa. The main body coherently superimposes the generated composite wave amplitude = 1.83 + 1.17 = 3.0kPa. Subsequently, the main body dynamically calculates the baseline threshold, taking 1.8 times the average normal wave peak of the most recent 10 minutes (1.60kPa), i.e., 2.88kPa. When the composite wave amplitude reaches 3.0kPa at t=3.114 seconds, which is greater than the threshold of 2.88kPa, the main body accurately extracts this feature point and outputs an enhanced wavefront feature vector [3.114s, 3.0kPa], perfectly removing the interference of background noise and generating a clean input.

[0087] In this embodiment, firstly, the execution entity determines the propagation direction of the transient pressure wave between each preset monitoring node based on the propagation characteristics; then, the execution entity calculates the propagation time difference between the preset monitoring nodes based on the propagation direction and propagation velocity components; subsequently, the execution entity performs phase compensation processing on the pressure pulsation data based on the propagation time difference to obtain compensated pressure pulsation data; finally, the execution entity superimposes the compensated pressure pulsation data to generate a synthetic wavefront signal, and determines the target fluctuation characteristics based on the synthetic wavefront signal.

[0088] As described above, this embodiment calculates the propagation time difference of transient pressure waves between different preset monitoring nodes and performs precise phase compensation accordingly. This effectively eliminates the time delay and phase deviation caused by the physical distance of the pipeline and the transmission of the fluid medium, achieving strict wavefront alignment of the dispersed pressure wave signals at multiple nodes in the feature space. Based on this wavefront alignment, the compensated data is coherently superimposed, accurately analyzing the reflection and superposition effects of pressure waves in complex pipelines. This results in a significant enhancement of the abrupt peak caused by actual leakage, while random and chaotic normal water flow disturbances and background noise in the pipeline cancel each other out during the superposition process. This mechanism significantly improves the signal-to-noise ratio of the target fluctuation characteristics, successfully separating extremely weak early leakage transient signals from strong background noise, accurately characterizing the propagation events of abnormal pressure waves, and thus providing a high-fidelity input source with extremely pure features for subsequent phase space reconstruction and high-precision positioning.

[0089] Step 103: Reconstruct the phase space of the target wave characteristics to obtain the phase space trajectory, and determine the phase space feature vector based on the wave mode abrupt change characteristics of the phase space trajectory.

[0090] Here, the executing entity performs coherent synthesis of pressure pulsation data at multiple preset monitoring nodes based on propagation characteristics to obtain target fluctuation characteristics. After obtaining the target fluctuation characteristics, it can reconstruct the phase space of the target fluctuation characteristics to obtain the phase space trajectory, and determine the phase space feature vector based on the wave mode abrupt change characteristics of the phase space trajectory.

[0091] In one possible embodiment, such as Figure 5 As shown, the phase space trajectory is obtained by reconstructing the phase space of the target fluctuation characteristics, including the following steps: Step 501: Identify the reference time interval of the fluctuation cycle in the target fluctuation characteristics, and divide the continuous time window based on the reference time interval.

[0092] Here, the execution entity performs coherent synthesis of pressure pulsation data at multiple preset monitoring nodes based on propagation characteristics to obtain target fluctuation characteristics. After that, it can identify the reference time interval of the fluctuation cycle in the target fluctuation characteristics and divide the continuous time window based on the reference time interval.

[0093] In one possible embodiment, the execution entity identifies the time interval between adjacent wave crest feature points in the enhanced wavefront feature vector (i.e., target wave feature) through the phase space reconstruction module, calculates the average value of the time interval, and uses it as the base pressure wave cycle (i.e., reference time interval). Then, the execution entity continuously extracts the enhanced wavefront feature vector along the time axis with an integer multiple of the base pressure wave cycle as the time window length, thereby dividing the continuous time window in the form of waveform data segments.

[0094] For example, in a practical application, continuing the aforementioned case of monitoring a generator's stator cooling water system, the executing entity first identifies the timestamp sequence of adjacent peaks in the enhanced wavefront feature vector, for example, [3.000s, 3.020s, 3.040s]. Based on this, the executing entity calculates the base pressure fluctuation period as 0.020s. Subsequently, the executing entity takes three times the period, i.e., 0.060s, as the time window length and extracts waveform data from the 3.000-3.060s segment along the time axis.

[0095] Step 502: Within each time window, determine the original state point based on the peak time and amplitude of the target fluctuation characteristics, and construct a high-dimensional state point by combining the amplitude evolution components of each original state point within a preset span.

[0096] Here, after the executing entity obtains continuous time windows based on the benchmark time interval, it can determine the original state point within each time window based on the peak time and amplitude of the target fluctuation characteristics, and construct a high-dimensional state point by combining the amplitude evolution components of each original state point within a preset span.

[0097] In one possible embodiment, the specific process by which the executing entity constructs high-dimensional state points includes: First, the executing entity extracts all peak feature points within the current time window and records the occurrence time of each peak feature point. and amplitude Combined into two-dimensional coordinate data to form the original set of state points. Where p is the number of peaks within the current window, the execution entity then calculates the arithmetic mean of the coordinates of all original state points within the time window, which are used as the reference time. and reference amplitude This generates the mean point. .

[0098] Subsequently, the executing entity, centered on the mean point M, divides the current time window into a forward evolution segment and a backward evolution segment. In the forward evolution segment (i.e., earlier than...),... At any given time, the execution entity screening meets the requirements. The peak points are identified and arranged in reverse chronological order. The amplitude changes (i.e., amplitude evolution components) of adjacent peak characteristic points are calculated. Generate forward amplitude sequence In the backward evolution stage (i.e., later than) At any given time, the execution entity screening meets the requirements. The peak points are arranged in ascending time sequence, and the amplitude change of adjacent peak characteristic points is calculated. Generate backward amplitude sequence Finally, the executing entity concatenates the forward amplitude variation sequence and the backward amplitude variation sequence end to end to construct the high-dimensional state point of the current time window. This vector fully encompasses the bidirectional evolution characteristics of the pressure wave within the local time window.

[0099] For example, in a practical application, during the monitoring of a generator stator cooling water system, the execution subject first identifies the adjacent peak times [3.000s, 3.020s, 3.040s] of the enhanced wavefront feature vector, calculates the basic pressure fluctuation period as 0.020s, takes 3 times the period as 0.060s as the time window to intercept the 3.000-3.060s segment, and detects four peak points within the window: P1 (3.010s, 3.0kPa), P2 (3.020s, 3.2kPa), P3 (3.030s, 3.5kPa), and P4 (3.050s, 2.8kPa).

[0100] The executing entity calculates the reference time. Reference amplitude ,by =3.0275s as the boundary: For the forward evolution segment, take the points (P1, P2) where ti < 3.0275s, arrange them in reverse chronological order to obtain P2 to P1, and calculate the amplitude variation. The backward evolution segment takes points (P3, P4) where ti > 3.0275s, arranges them in ascending order to obtain P3 to P4, and calculates the amplitude variation. Construct a high-dimensional state point Vk=(0.2,-0.7).

[0101] Next, the executing entity divides the trajectory into a local segment of 5 windows and sets a baseline scale. =0.2, calculate the decreasing scale. , At a scale of r0=0.2, count the number of grid cells traversed by this segment. , At this scale, the number of grid cells traversed is N(r1) = 17, and the fractal dimension is calculated by the main body. Given that D=0.95 in the previous section, calculate the change in fractal dimension. When the change exceeds the historical threshold At that time, the executing entity marks the segment as an abnormal segment. Finally, the executing entity extracts the state points of the abnormal segment to form a phase space feature vector. .

[0102] Step 503: Connect the high-dimensional state points in chronological order to obtain the phase space trajectory.

[0103] Here, after constructing high-dimensional state points, the executing entity can connect each high-dimensional state point in chronological order to obtain the phase space trajectory.

[0104] In one possible embodiment, the executing entity, following the sliding order of the time windows, generates high-dimensional state points corresponding to each time window. By sequentially connecting them in a high-dimensional space, a continuous high-dimensional phase space attractor trajectory is formed, where T is the total number of time windows. This attractor trajectory, through the continuous migration path of state points in the phase space, can visually present the evolution process of the fluctuation mode of the constant cooling water loop fluid system, laying a precise geometric model foundation for subsequent capture of nonlinear mode distortion caused by leakage faults through fractal dimension mutation detection.

[0105] In one possible embodiment, such as Figure 6 As shown, determining the phase space eigenvector based on the abrupt change characteristics of wave modes in the phase space trajectory includes the following steps: Step 601: Divide the phase space trajectory into multiple local trajectory segments, and perform multi-scale grid division for each local trajectory segment to obtain the multi-scale grid distribution data corresponding to the local trajectory segment.

[0106] Here, after the executing entity reconstructs the phase space trajectory by the target fluctuation characteristics, it can divide the phase space trajectory into multiple local trajectory segments and perform multi-scale grid division on each local trajectory segment to obtain the multi-scale grid distribution data corresponding to the local trajectory segment.

[0107] In one possible embodiment, the executing entity first divides the complete continuous high-dimensional phase space attractor trajectory along the time axis into multiple continuous local trajectory segments. For each local trajectory segment, the executing entity performs multi-scale grid analysis and sets a reference spatial scale. And generate a decreasing scale sequence. ,in For each scale in this decreasing scale sequence The executing entity divides the high-dimensional phase space into sections with sides of length [missing information]. A cubic spatial grid array is used to map the abstract trajectory to a specific spatial grid, thereby obtaining multi-scale grid distribution data of the local trajectory segment at different scales.

[0108] For example, in practical applications, continuing the aforementioned case of phase space reconstruction of a generator stator cooling water system, the executing entity divides the constructed attractor trajectory into continuous local trajectory segments according to the rule of five time windows per segment, and the executing entity sets a reference spatial scale. And calculate the decreasing scale of the next level according to the formula. The executing entity divides the grid arrays for these scales respectively, generating multi-scale grid distribution data for subsequent statistics.

[0109] Step 602: Based on the multi-scale grid distribution data, count the number of times each local trajectory segment crosses each multi-scale grid, and calculate the fractal dimension value of each local trajectory segment based on the number.

[0110] Here, after obtaining the multi-scale grid distribution data corresponding to the local trajectory segment, the executing entity can count the number of times each local trajectory segment crosses each multi-scale grid based on the multi-scale grid distribution data, and calculate the fractal dimension value of each local trajectory segment based on the number.

[0111] In one possible embodiment, the executing entity accurately calculates the current local trajectory segment at a specific scale based on the generated multi-scale grid distribution data. The actual number of spatial grids traversed The executing entity repeats the statistical process of the number of grid crossings described above in a decreasing scale sequence to obtain a scale quantity sequence. Based on this, the executing entity utilizes adjacent scales... and and the corresponding number of grids and According to fractal geometry formulas The fractal dimension value D of the current local trajectory segment is calculated. This fractal dimension value quantifies the complexity and space-filling characteristics of the attractor trajectory in phase space and usually remains stable under normal system conditions.

[0112] For example, in practical applications, the executing entity statistically determines the aforementioned local trajectory segment at a reference scale. Number of grids crossed , in scale Number of grids crossed The executing entity substitutes these values ​​into the calculation formula and calculates that the fractal dimension of the segment is D≈1.086.

[0113] Step 603: Determine the abrupt change characteristics of the wave mode by detecting the change in the fractal dimension value of adjacent local trajectory segments.

[0114] Here, after the executing entity calculates the fractal dimension value of each local trajectory segment based on the quantity, it can determine the wave mode mutation characteristics by detecting the change in the fractal dimension value of adjacent local trajectory segments. The wave mode mutation characteristics are used to quantitatively characterize the spatial complexity anomaly of the phase space trajectory.

[0115] In one possible embodiment, the executing entity detects the change ΔD in the fractal dimension of two temporally adjacent local trajectory segments based on the fractal dimension value of each local trajectory segment calculated sequentially. Specifically, the executing entity calculates the absolute value of the difference between the fractal dimension values ​​of the current segment and the previous segment (i.e., The change in fractal dimension was identified as a wave mode mutation feature, which was used to keenly capture trajectory morphological mutations caused by leakage.

[0116] For example, in practical applications, the executing entity retrieves the fractal dimension value of the preceding segment of the local trajectory segment. Combined with the fractal dimension calculated in the current segment The executing entity calculates the change in fractal dimension of adjacent local trajectory segments. The executing entity uses this change as a characteristic of wave pattern mutation in quantification anomalies.

[0117] Step 604: The local trajectory segment that generates abrupt change characteristics of wave mode is identified as an abnormal trajectory segment.

[0118] Here, after determining the wave mode mutation characteristics by detecting the change in the fractal dimension of adjacent local trajectory segments, the executing entity can identify the local trajectory segments that generate wave mode mutation characteristics as abnormal trajectory segments.

[0119] In one possible embodiment, the executing entity will determine the abrupt change characteristics of the wave mode, i.e., the change in fractal dimension value. Compared with the preset historical normal operating condition thresholds in the system When a comparison and detection is performed, if the execution entity detects that the change in fractal dimension value is greater than the threshold of the historical normal operating condition, the adjacent local trajectory segments that produce the abrupt change characteristics of the wave pattern will be directly marked and identified as abnormal trajectory segments.

[0120] For example, in practical applications, the system's preset historical normal operating condition threshold is 0.15, and the executing entity will use the change calculated above. The detection comparison with the threshold showed that it was greater than the threshold of 0.15. Based on this, the executing entity accurately marked the local trajectory segment and identified it as an abnormal trajectory segment.

[0121] Step 605: Extract high-dimensional state points from the abnormal trajectory segments and construct phase space feature vectors based on the high-dimensional state points from the abnormal trajectory segments.

[0122] Here, after identifying the local trajectory segment that generates abrupt wave mode change characteristics as an abnormal trajectory segment, the executing entity can extract high-dimensional state points in the abnormal trajectory segment and construct a phase space feature vector based on the high-dimensional state points in the abnormal trajectory segment.

[0123] In one possible embodiment, the executing entity extracts all high-dimensional state points contained in the identified abnormal trajectory segments. Subsequently, the executing entity arranges the extracted high-dimensional state points strictly according to the time order of the abnormal trajectory segments, thereby generating and constructing a phase space feature vector characterizing the abnormal fluctuation pattern.

[0124] For example, in practical applications, the executing entity extracts all state points within the marked abnormal segment, i.e., the high-dimensional state points generated in the previous steps. This forms a phase space feature vector [-0.6, 0.3], which will be used as a pure and highly discriminative input for the anomaly mode and transmitted to the subsequent fault classification model.

[0125] In this implementation, firstly, the executing entity divides the phase space trajectory into multiple local trajectory segments and performs multi-scale grid subdivision for each local trajectory segment to obtain multi-scale grid distribution data corresponding to the local trajectory segments. Then, based on the multi-scale grid distribution data, the executing entity counts the number of times each local trajectory segment crosses each multi-scale grid and calculates the fractal dimension value of each local trajectory segment based on the number. Next, by detecting the change in the fractal dimension value of adjacent local trajectory segments, the wave mode abrupt change characteristics are determined. Then, the local trajectory segments that generate wave mode abrupt change characteristics are determined as anomalous trajectory segments. Finally, the executing entity extracts high-dimensional state points from the anomalous trajectory segments and constructs phase space feature vectors based on the high-dimensional state points in the anomalous trajectory segments.

[0126] As described above, this embodiment accurately characterizes the spatial complexity and filling characteristics of high-dimensional dynamic trajectories at different scales by dividing the phase space trajectory into multi-scale grids and counting the number of grid crossings. Furthermore, by calculating and detecting the changes in the fractal dimension values ​​of adjacent local trajectory segments, it can keenly capture and quantify the nonlinear mode distortion of pressure waves caused by early minor leaks, effectively removing background interference caused by fluctuations in normal generator operating conditions. Moreover, this embodiment successfully transforms the abnormal dynamic evolution mode that characterizes the essence of leakage faults into structured data by accurately locating abnormal trajectory segments and extracting high-dimensional state points to construct phase space feature vectors. This provides a denoised, highly discriminative abnormal mode fingerprint for subsequent fault classification models, thereby significantly improving fault detection sensitivity and the accuracy of feature representation.

[0127] Step 104: Determine the monitoring results based on the phase space feature vector.

[0128] Here, after determining the phase space feature vector based on the wave mode abrupt change characteristics of the phase space trajectory, the executing entity can determine the monitoring results based on the phase space feature vector. The monitoring results are used to characterize the leakage location of the constant cooling water circuit.

[0129] In one possible embodiment, such as Figure 7 As shown, determining the monitoring results based on phase space eigenvectors includes the following steps: Step 701: Calculate the spatial similarity between the phase space feature vector and the preset reference pattern vector to obtain the spatial distribution pattern code.

[0130] Here, after determining the phase space feature vector based on the wave mode abrupt change characteristics of the phase space trajectory, the executing entity can calculate the spatial similarity between the phase space feature vector and the preset reference mode vector to obtain the spatial distribution mode encoding.

[0131] In one possible embodiment, the executing entity inputs a phase space feature vector containing anomalous wave mode information into a feature mapping layer. The feature mapping layer contains multiple randomly generated nonlinear transformation units. In each nonlinear transformation unit, the executing entity calculates the spatial similarity between the phase space feature vector and a randomly generated reference mode vector. Specifically, the nonlinear transformation unit refers to a randomly generated mode matching component, which is essentially a microprocessor with preset random parameters. Spatial similarity is an index characterizing the directional consistency between the input feature vector and the random reference vector; it is essentially a cosine angle value, reflecting the similarity of the anomalous wave modes' distribution in phase space, and is used to trigger unit activation decisions for each high-dimensional state point of the input vector. The executing entity uses formulas Perform fractal space similarity calculation, where L is the vector dimension. yes The change in fractal dimension of the segment in question. This refers to the change in fractal dimension associated with the reference pattern. This design allows similarity calculation to simultaneously consider spatial distribution characteristics and fractal mutation intensity. Based on the comparison between spatial similarity and randomly set response thresholds, when the calculated similarity Sim exceeds the unit random threshold... When the unit outputs an activation state value of 1, it outputs 0 otherwise. The execution body aggregates the activation state values ​​of all nonlinear transformation units to form a binary form of fractal enhanced spatial distribution pattern encoding.

[0132] For example, in practical applications, the input phase space feature vector contains high-dimensional state points. =[0.4,-0.2,-0.4] and its associated fractal dimension changes =0.15, Unit 1 random reference vector R=[0.3,-0.1,-0.5], associated =0.18, random threshold =0.6, the executing entity first calculates the fractal space similarity. The threshold was compared and found that 0.93 > 0.6, so the unit output activation value was 1. Finally, the activation states of 200 units were used to form the code "101...1" (containing 120 1s) as the spatial distribution pattern code.

[0133] Step 702: Match the spatial distribution pattern encoding with the pre-stored leakage location pattern library to obtain the overlap matching result.

[0134] Here, after obtaining the spatial distribution pattern code, the executing entity can match the spatial distribution pattern code with the pre-stored leakage location pattern library to obtain the overlap matching result.

[0135] In one possible embodiment, the executing entity matches the spatial distribution pattern code with a pre-stored leakage location pattern library. By extracting the baseline distribution pattern corresponding to each loop region identifier in the pattern library, the overlap coverage between the spatial distribution pattern code and each baseline distribution pattern is calculated as the overlap matching result. The overlap coverage is a proportional parameter that quantifies the degree of matching between the current code and the baseline code. Essentially, it is the ratio of the same activation site to the effective number of bits in the baseline code, reflecting the degree of consistency between the current fault mode and the historical leakage location characteristics.

[0136] For each loop region identifier in the library Extract its baseline fractal pattern The executing entity calculates according to the formula. Calculate the overlap coverage, where M is the total number of cells. For the current encoding associated with mean For pre-stored leak location pattern library, with specific loop area identifiers The mean change in the associated baseline fractal dimension, z is the identifier, and this coefficient ensures that only patterns with similar fractal mutation intensities can achieve a high degree of matching.

[0137] For example, in practical applications, the execution subject matches the pattern library area. The base code "110...0" (containing 100 ones) corresponds to =0.16, according to the formula, the execution entity calculates the overlap coverage. Similarly, the executing entity calculates the region. coverage .

[0138] Step 703: Determine the monitoring results based on the overlap matching results.

[0139] Here, after matching the spatial distribution pattern encoding with the pre-stored leakage location pattern library to obtain the overlap matching result, the execution entity can determine the monitoring result based on the overlap matching result.

[0140] In one possible embodiment, determining the monitoring result based on the overlap matching result includes the following steps: The region identifiers of each fixed cooling water loop are sorted in descending order of overlap matching results to obtain the loop region sorting sequence; Extract the overlap degree corresponding to the region identifier that ranks first in the loop region sorting sequence to obtain the target overlap degree set; The numerical values ​​in the target overlap set are subjected to probability transformation to obtain the probability distribution corresponding to each region identifier, and the monitoring results are constructed based on the probability distribution.

[0141] Specifically, after matching the spatial distribution pattern encoding with the pre-stored leakage location pattern library to obtain the overlap matching result, the executing entity first sorts the area identifiers of each fixed cooling water loop in descending order of the overlap matching result to obtain the loop area sorting sequence; then, it extracts the overlap corresponding to the area identifiers ranked first in the loop area sorting sequence to obtain the target overlap set; finally, it performs probability transformation processing on the values ​​in the target overlap set to obtain the probability distribution corresponding to each area identifier, and constructs the monitoring results based on the probability distribution.

[0142] In one possible embodiment, after matching the spatial distribution pattern encoding with a pre-stored leakage location pattern library to obtain the overlap matching result, the executing entity first sorts the area identifiers of each fixed cooling water loop in descending order of the overlap matching result, generating a loop area identifier sequence. Then, it extracts the overlap coverage values ​​(i.e., the target overlap set) of the top N loop area identifiers in the loop area sorting sequence. Finally, the executing entity uses the formula... The numerical values ​​are processed by probability transformation to form a probability distribution vector of leakage locations, which intuitively displays the probability of failure in each area and constructs the final monitoring results.

[0143] In one possible embodiment, following the coverage results calculated above, the executing entity sorts the results according to the coverage size to obtain a sequence. Subsequently, the executing entity calculates the normalized probability, obtaining the total coverage = 0.5625 + 0.5 = 1.0625. Then, the executing entity converts this into a probability value: , Finally, the executing entity outputs a probability distribution vector. Based on this, the loop region identifier with the highest probability value (i.e., 53%) is determined. The corresponding physical pipe section is the target fault location most likely to leak. At the same time, the probability distribution is sorted as a priority guide for on-site maintenance personnel to carry out troubleshooting, thus successfully completing the accurate location of the leak and outputting the monitoring results.

[0144] In this embodiment, firstly, the executing entity calculates the spatial similarity between the phase space feature vector and the preset reference pattern vector to obtain the spatial distribution pattern code; then, the executing entity matches the spatial distribution pattern code with the pre-stored leakage location pattern library to obtain the overlap matching result; finally, the executing entity determines the monitoring result based on the overlap matching result.

[0145] As described above, this embodiment cleverly transforms high-dimensional phase space anomaly features into spatial distribution pattern codes by calculating the spatial similarity between phase space feature vectors and preset reference pattern vectors. Then, it uses a pre-stored leakage location pattern library to perform precise matching of overlap, achieving efficient mapping and association between abstract fault features and the underlying physical spatial location of the generator. Finally, this embodiment generates a probability distribution based on the overlap matching results, which intuitively shows the possibility of leakage in each constant cooling water circuit area in a quantitative way. This mechanism completely gets rid of the excessive dependence of traditional fault classification models on massive labeled data. Without the need for time-consuming model pre-training, it successfully achieves online, real-time and accurate spatial quantitative localization of minor leakage faults in generators.

[0146] This disclosure provides a method, apparatus, equipment, medium, and product for monitoring generator constant cooling water leakage faults. First, pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored are acquired. The pressure pulsation data characterizes the fluctuation characteristics of transient pressure waves generated in the constant cooling water circuit under varying water flow rates. Then, based on the pressure pulsation data, the propagation characteristics of the transient pressure waves in the constant cooling water circuit are determined, and the pressure pulsation data at multiple preset monitoring nodes are coherently synthesized based on the propagation characteristics to obtain target fluctuation characteristics. Next, the target fluctuation characteristics are reconstructed in phase space to obtain a phase space trajectory, and the phase space feature vector is determined based on the wave mode abrupt change characteristics of the phase space trajectory. Finally, the monitoring result is determined based on the phase space feature vector. The monitoring result characterizes the leakage location in the constant cooling water circuit.

[0147] As described above, this embodiment of the present disclosure effectively captures the transient pressure wave fluctuation characteristics generated under varying water flow velocity conditions by acquiring pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator under test. This overcomes the limitations of traditional static pressure thresholds, which are sensitive to dynamic noise and extremely insensitive to early minor leaks. Furthermore, this embodiment of the present disclosure, based on in-depth analysis of pressure pulsation data, determines the dynamic propagation characteristics of transient pressure waves in complex pipelines. Accordingly, it coherently synthesizes the pressure pulsation data at multiple preset monitoring nodes, effectively overcoming pipeline propagation delay and phase deviation. While accurately analyzing the reflection and superposition effects of pressure waves, it filters background noise, thereby significantly enhancing the signal-to-noise ratio of the obtained target fluctuation characteristics. Subsequently, the phase space trajectory is obtained by reconstructing the target fluctuation characteristics in phase space, and the phase space feature vector is determined based on the wave mode mutation characteristics of the phase space trajectory. This process deeply explores the dynamic anomalies of the pressure wave in the nonlinear evolution process, and realizes the accurate separation of the real leakage signal from the fluctuations of the system under normal operating conditions. Finally, the monitoring results are determined based on the phase space feature vector, and the probability distribution representing the leakage location is output. This completely eliminates the traditional coarse differential alarm that relies on thresholds and leads to fuzzy positioning. The embodiments of this disclosure achieve high-sensitivity identification and high-precision spatial location tracing of leakage points in the fixed cooling water circuit without the need for extensive pre-training, and comprehensively improves the accuracy, timeliness and anti-interference capability of generator online diagnosis.

[0148] In a specific end-to-end application scenario, combined with Figure 8 As shown in the figure, this embodiment provides a complete execution process of a generator stator cooling water leakage fault online monitoring method: In the stator cooling water circuit of a certain generator set, technicians first deployed multi-channel resonant sensors at key nodes (stator inlet main pipe J1, rotor inlet branch J2, and end cooling water jacket J3).

[0149] First, in the data capture phase (corresponding to step 101 and its sub-steps 201-203): when the unit load increases from 300MW to 450MW, the cooling water flow velocity in the stator cooling water circuit increases from 1.0m / s to 1.5m / s. At this time, after the sensor at node J2 detects the change in flow velocity, its embedded processor dynamically adjusts the resonant frequency according to the Strouhal relation, based on preset parameters k=0.25 and d=0.02m, and calculates the target monitoring frequency. Two seconds after the frequency adjustment was completed, a simulated leak occurred downstream of node J2, causing a pressure fluctuation. The actuator accurately detected the frequency shift at this moment using sensors. Combined with the calibration coefficients of the sensor The executing entity calculates the pressure pulsation value at this moment. The pressure pulsation data is then output synchronously at a sampling rate of 5 kHz.

[0150] Next, in the stage of constructing the pressure wave propagation velocity matrix (corresponding to steps 301 to 303): when processing the pressure pulsation data of pipe segments J1 to J2, the execution subject accurately identifies a negative fluctuation point at t=3.205s, with a pressure sequence of [102.1, 101.2, 100.9] kPa. The execution subject extracts the mean of the first 5 sampling points before this point, 102.1 kPa, and calculates its backward first derivative as -180 kPa / s; subsequently, based on the derivative value of -190 kPa / s at the first 0.2 ms, the second derivative representing the deceleration trend is further calculated as... At this point, the executing entity calls the pipe size parameters of the fixed cooling water circuit, confirms that the equivalent hydraulic diameter of the pipe section is 0.04m, calculates that its reverse propagation velocity component is 5.76m / s, and combines the velocity component data of the other pipe sections to finally construct a complete pressure wave propagation velocity matrix.

[0151] Subsequently, in the stage of coherent synthesis and extraction of target fluctuation features (corresponding to steps 401 to 404): For target node J2, the execution subject calculates the propagation time difference based on the previously determined propagation direction and propagation speed components, compensates the signal of its upstream node J1 downstream, and compensates the signal of the node J3 associated with the leakage reflection point upstream. After obtaining the compensated pressure pulsation data, the execution subject performs amplitude weighting and superposition according to the frequency response characteristics of each sensor to generate a synthesized wavefront signal with an amplitude of 1.7 + 1.2 = 2.9 kPa. When t = 3.305 s, a wavefront peak of 3.5 kPa appears in the synthesized wavefront signal. The execution subject determines that the amplitude of the wavefront exceeds the dynamic baseline threshold, and then extracts the feature point and determines it as the target fluctuation feature (i.e., the enhanced wavefront feature vector: [positive, 3.305 s, 3.5 kPa]).

[0152] Furthermore, in the phase space reconstruction stage (corresponding to steps 501 to 503): Based on the peak time series [3.300s, 3.305s, 3.310s] of the target fluctuation characteristics, the executing entity calculates the reference time interval (basic period) as 5ms. The executing entity takes 3 times the period (i.e., 15ms) as the time window length, dividing it into a continuous time window from 3.300s to 3.315s. This time window contains three original state points: P1 (3.302s, 3.2kPa), P2 (3.305s, 3.5kPa), and P3 (3.312s, 2.9kPa). After calculating the mean point M, the executing entity combines it with the amplitude evolution component to construct the high-dimensional state point V. k =(-0.6,0.3), connect the high-dimensional state points in chronological order, and the executing entity finally obtains the phase space trajectory.

[0153] Then, in the phase space eigenvector determination stage (corresponding to steps 601 to 605): the executing entity divides the phase space trajectory into local trajectory segments of 5 time windows and performs multi-scale grid division: at the reference scale r_0=0.2, the number of grid crossings is counted. At a finer scale At that time, count the number of times the grid is crossed. Based on the above quantities, the executing entity calculates the fractal dimension of this local trajectory segment. Since the fractal dimension value D of the previous adjacent local trajectory segment is 0.93, the execution subject detects the change in the fractal dimension value. The change (wave pattern mutation feature) is greater than the historical normal operating condition threshold of 0.15. Therefore, the execution subject identifies the local trajectory segment as an abnormal trajectory segment and extracts the high-dimensional state points in it to construct the phase space feature vector [-0.6, 0.3].

[0154] Finally, in the stage of determining the monitoring results (corresponding to steps 701 to 703): the execution subject calculates the spatial similarity between the phase space feature vector [-0.6, 0.3] and the preset reference pattern vector. The spatial similarity Sim calculated by a certain unit is 0.70. Since 0.70 exceeds the response threshold of 0.68, the activation state value of the generating unit is 1. After aggregating multiple activation state values, the execution subject obtains the spatial distribution pattern code [1, 0, 1]. The spatial distribution pattern code is matched with the pre-stored leakage location pattern library, and the execution subject obtains the overlap matching result: the overlap CA with region A is 0.650, and the overlap CB with region B is 0.420. After probability transformation processing, the probability distribution corresponding to each region identifier is finally obtained: , The executing entity constructs and outputs the final monitoring results based on this probability distribution. Based on this, the physical pipe section corresponding to area A with the highest probability value (approximately 60.7%) was determined to be the most likely target fault location for leakage. At the same time, the probability distribution was ranked as a priority guide for on-site maintenance personnel to conduct troubleshooting, thus successfully achieving accurate location of leakage and closed-loop online monitoring.

[0155] As described above, the technical solution disclosed herein captures pressure pulsation data under changes in water flow velocity in real time through the adaptive frequency modulation mechanism of a multi-channel resonant sensor, overcoming the limitation of static threshold sensitivity to dynamic noise. Furthermore, this technical solution utilizes a pressure wave propagation velocity matrix constructed using first and second derivatives to accurately quantify the dynamic propagation characteristics of waves in complex pipelines. Combined with the coherent synthesis of multi-node pressure waves to generate enhanced wavefront feature vectors, it effectively analyzes reflection and superposition effects. Moreover, by using a sliding time window adaptive to the wave period to ensure that trajectory construction is synchronized with the system's inherent dynamics, and by utilizing asymmetric evolution segment analysis to accurately capture pressure wave propagation in... The asymmetry of spatiotemporal evolution under leakage disturbance breaks through the dependence of traditional phase space reconstruction on time symmetry. Based on this, the technical solution of this disclosure extracts the fractal dimension mutation feature to capture the attractor trajectory anomaly caused by leakage, which significantly improves the sensitivity of fractal dimension mutation detection. Finally, the fault classification model with randomly initialized hidden layer nodes outputs the probability distribution of leakage location, realizing the online identification of the unique fractal dynamic mode of leakage fault. The technical solution of this disclosure effectively improves the capture rate of small leakage and reduces the location error, fundamentally providing the generator stator cooling water circuit with high sensitivity, low false alarm and accurate location online fault diagnosis capability.

[0156] By dividing each function into corresponding functional modules, this disclosure provides a monitoring device for generator constant cooling water leakage faults. This monitoring device can be a server or a chip applied to a server. Figure 9 This is a schematic block diagram of the functional modules of a generator stator cooling water leakage monitoring device provided as an exemplary embodiment of this disclosure. Figure 9 As shown, the monitoring device for generator stator cooling water leakage includes: The acquisition module 901 is used to acquire pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored; the pressure pulsation data is used to characterize the fluctuation characteristics of transient pressure waves generated in the constant cooling water circuit under the condition of water flow rate change. The first determining module 902 is used to determine the propagation characteristics of the transient pressure wave in the constant cooling water circuit based on the pressure pulsation data, and to coherently synthesize the pressure pulsation data at the multiple preset monitoring nodes based on the propagation characteristics to obtain the target fluctuation characteristics. The second determining module 903 is used to reconstruct the phase space of the target wave characteristics to obtain the phase space trajectory, and to determine the phase space feature vector based on the wave mode abrupt change characteristics of the phase space trajectory. The monitoring module 904 is used to determine the monitoring results based on the phase space feature vector; the monitoring results are used to characterize the leakage location of the constant cooling water circuit.

[0157] In one embodiment, the acquisition module 901 includes: The first acquisition unit is used to capture pressure signals through multi-channel resonant sensors set at each of the preset monitoring nodes; The first determining unit is used to determine the resonant frequency of the multi-channel resonant sensor based on the real-time water flow rate change of the constant cooling water circuit, and obtain the target monitoring frequency. The second acquisition unit is used to capture each of the pressure signals in real time through the target monitoring frequency to obtain the pressure pulsation data.

[0158] In one embodiment, the first determining module 902 includes: The first calculation unit is used to calculate the pressure change rate of the pressure pulsation data and determine the fluctuation evolution characteristics of the transient pressure wave based on the pressure change rate; wherein, the fluctuation evolution characteristics include the first derivative and the second derivative of the pressure change rate; The second calculation unit is used to determine the propagation velocity components of the transient pressure wave on different propagation paths based on the first derivative, the second derivative, and the pipe size parameters of the constant cooling water circuit. The first construction unit is used to construct a pressure wave propagation velocity matrix based on the propagation velocity component and the topology of the constant cooling water circuit; wherein the pressure wave propagation velocity matrix is ​​used to characterize the propagation characteristics of the transient pressure wave in the constant cooling water circuit.

[0159] In one embodiment, the first determining module 902 includes: The first identification unit is used to identify the pressure fluctuation direction of the pressure pulsation data; The third calculation unit is used to calculate the pressure change rate using a forward difference algorithm if the pressure fluctuation direction is positive, and obtain the second derivative characterizing the pressure acceleration trend. The fourth calculation unit is used to calculate the pressure change rate using a backward difference algorithm if the pressure fluctuation direction is negative, and obtain the second derivative characterizing the pressure deceleration trend.

[0160] In one embodiment, the first determining module 902 includes: The second determining unit is used to determine the propagation direction of the transient pressure wave between each of the preset monitoring nodes based on the propagation characteristics. The fifth calculation unit is used to calculate the propagation time difference between the preset monitoring nodes based on the propagation direction and the propagation speed component; The compensation unit is used to perform phase compensation processing on the pressure pulsation data based on the propagation time difference to obtain compensated pressure pulsation data. The third determining unit is used to superimpose the compensated pressure pulsation data to generate a synthetic wavefront signal, and determine the target fluctuation characteristics based on the synthetic wavefront signal.

[0161] In one embodiment, the first determining module 902 includes: The first division unit is used to divide the signals in the compensated pressure pulsation data corresponding to different propagation directions into forward propagation datasets and reverse propagation datasets based on the propagation direction. The weighting unit is used to perform amplitude weighting on the forward propagation dataset and the backward propagation dataset based on the frequency response characteristics of the multi-channel resonant sensor, respectively, to obtain weighted multi-node component data. The superposition unit is used to superimpose the weighted multi-node component data to generate the synthetic wavefront signal.

[0162] In one embodiment, the second determining module 903 includes: The second identification unit is used to identify the reference time interval of the fluctuation cycle in the target fluctuation feature, and to divide the continuous time window based on the reference time interval; The second construction unit is used to determine the original state point based on the peak time and amplitude of the target fluctuation characteristics within each time window, and to construct a high-dimensional state point by combining the amplitude evolution components of each original state point within a preset span. A connection unit is used to connect the high-dimensional state points in chronological order to obtain the phase space trajectory.

[0163] In one embodiment, the second determining module 903 includes: The second partitioning unit is used to divide the phase space trajectory into multiple local trajectory segments, and to perform multi-scale grid partitioning for each local trajectory segment to obtain multi-scale grid distribution data corresponding to the local trajectory segment. The statistical unit is used to count the number of times each local trajectory segment crosses each multi-scale grid based on the multi-scale grid distribution data, and to calculate the fractal dimension value of each local trajectory segment based on the number. The fourth determining unit is used to determine the wave mode abrupt change feature by detecting the change in the fractal dimension value of adjacent local trajectory segments; wherein the wave mode abrupt change feature is used to quantitatively characterize the spatial complexity anomaly of the phase space trajectory. The fifth determining unit is used to determine the local trajectory segment that generates the abrupt change feature of the wave mode as an abnormal trajectory segment; The third construction unit is used to extract high-dimensional state points in the abnormal trajectory segment and construct the phase space feature vector based on the high-dimensional state points in the abnormal trajectory segment.

[0164] In one embodiment, the monitoring module 904 includes: The sixth calculation unit is used to calculate the spatial similarity between the phase space feature vector and the preset reference pattern vector to obtain the spatial distribution pattern code; A matching unit is used to match the spatial distribution pattern encoding with a pre-stored leakage location pattern library to obtain an overlap matching result; The sixth determining unit is used to determine the monitoring result based on the overlap matching result.

[0165] In one embodiment, the monitoring module 904 includes: The sorting unit is used to sort the area identifiers of each of the fixed cooling water loops in descending order of the overlap matching results to obtain the loop area sorting sequence. Extraction unit is used to extract the overlap degree corresponding to the region identifier ranked first in the loop region sorting sequence, and obtain the target overlap degree set; The fourth construction unit is used to perform probability transformation processing on the values ​​in the target overlap set to obtain the probability distribution corresponding to each region identifier, and to construct the monitoring result based on the probability distribution.

[0166] Figure 10 This is a schematic diagram of the structure of an electronic device provided as an exemplary embodiment of this disclosure. For example... Figure 10 As shown, the electronic device 1000 includes at least one processor 1001 and a memory 1002 coupled to the processor 1001. The processor 1001 can perform the corresponding steps in the methods disclosed in the embodiments of this disclosure.

[0167] The processor 1001 described above can also be called a central processing unit (CPU), which can be an integrated circuit chip with signal processing capabilities. Each step in the method disclosed in this embodiment can be implemented by the integrated logic circuitry in the processor 1001 or by software instructions. The processor 1001 can be a general-purpose processor, a digital signal processor (DSP), an ASIC, a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this embodiment can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software modules can be located in the memory 1002, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The processor 1001 reads information from the memory 1002 and, in conjunction with its hardware, completes the steps of the method described above.

[0168] Furthermore, various operations / processes according to this disclosure, implemented via software and / or firmware, can be transmitted from a storage medium or network to a computer system with a dedicated hardware architecture, such as... Figure 11 The computer system 1100 shown is equipped with the programs that constitute the software. When various programs are installed, the computer system is able to perform various functions, including functions such as those described above. Figure 11 A block diagram of a computer system provided for an exemplary embodiment of this disclosure.

[0169] Computer system 1100 is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0170] like Figure 11As shown, the computer system 1100 includes a computing unit 1101, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 1102 or a computer program loaded into random access memory (RAM) 1103 from a storage unit 1108. The RAM 1103 may also store various programs and data required for the operation of the computer system 1100. The computing unit 1101, ROM 1102, and RAM 1103 are interconnected via a bus 1104. An input / output (I / O) interface 1105 is also connected to the bus 1104.

[0171] Multiple components in computer system 1100 are connected to I / O interface 1105, including: input unit 1106, output unit 1107, storage unit 1108, and communication unit 1109. Input unit 1106 can be any type of device capable of inputting information into computer system 1100. Input unit 1106 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of the electronic device. Output unit 1107 can be any type of device capable of presenting information and may include, but is not limited to, a monitor, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 1108 may include, but is not limited to, hard disks and optical disks. Communication unit 1109 allows computer system 1100 to exchange information / data with other devices via a network such as the Internet, and may include, but is not limited to, modems, network cards, infrared communication devices, wireless communication transceivers, and / or chipsets, such as Bluetooth™ devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0172] The computing unit 1101 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 1101 performs the various methods and processes described above. For example, in some embodiments, the methods disclosed in this disclosure can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 1108. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 1000 via ROM 1102 and / or communication unit 1109. In some embodiments, the computing unit 1101 can be configured to perform the methods disclosed in this disclosure by any other suitable means (e.g., by means of firmware).

[0173] This disclosure also provides a computer-readable storage medium, wherein when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is able to perform the methods disclosed in this disclosure.

[0174] The computer-readable storage medium in this disclosure can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. The aforementioned computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specifically, the aforementioned computer-readable storage medium may include electrical connections based on one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0175] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0176] Figure 12 A computer program product 1200 is provided as an exemplary embodiment of the present disclosure. The computer program product 1200 includes a computer program 1201, wherein the computer program 1201, when executed by a processor, implements the methods disclosed in the embodiments of the present disclosure.

[0177] In embodiments of this disclosure, computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include, but are not limited to, object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network (including a local area network (LAN) or a wide area network (WAN)), or it can be connected to an external computer.

[0178] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0179] The modules, components, or units described in the embodiments of this disclosure can be implemented in software or hardware. The names of the modules, components, or units do not necessarily constitute a limitation on the module, component, or unit itself.

[0180] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary hardware logic components that can be used include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0181] The above description is merely an embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

[0182] While specific embodiments of this disclosure have been described in detail by way of example, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of this disclosure. Those skilled in the art should understand that modifications can be made to the above embodiments without departing from the scope and spirit of this disclosure. The scope of this disclosure is defined by the appended claims.

Claims

1. A method for monitoring generator stator cooling water leakage faults, characterized in that, The method includes: Acquire pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored; the pressure pulsation data is used to characterize the fluctuation characteristics of transient pressure waves generated in the constant cooling water circuit under the condition of water flow rate variation. Based on the pressure pulsation data, the propagation characteristics of the transient pressure wave in the constant cooling water circuit are determined, and the pressure pulsation data at the multiple preset monitoring nodes are coherently synthesized based on the propagation characteristics to obtain the target fluctuation characteristics. The phase space trajectory is obtained by reconstructing the target wave characteristics in phase space, and the phase space feature vector is determined based on the wave mode abrupt change characteristics of the phase space trajectory. The monitoring results are determined based on the phase space feature vectors; the monitoring results are used to characterize the leakage location of the constant cooling water circuit.

2. The method according to claim 1, characterized in that, The acquisition of pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored includes: Pressure signals are captured by multi-channel resonant sensors installed at each of the preset monitoring nodes; Based on the real-time water flow rate change of the constant cooling water circuit, the resonant frequency of the multi-channel resonant sensor is determined, and the target monitoring frequency is obtained. The pressure signals are captured in real time by the target monitoring frequency to obtain the pressure pulsation data.

3. The method according to claim 1, characterized in that, The determination of the propagation characteristics of the transient pressure wave in the constant cooling water circuit based on the pressure pulsation data includes: Calculate the pressure change rate of the pressure pulsation data, and determine the fluctuation evolution characteristics of the transient pressure wave based on the pressure change rate; wherein, the fluctuation evolution characteristics include the first derivative and the second derivative of the pressure change rate; Based on the first derivative, the second derivative, and the pipe size parameters of the constant cooling water circuit, the propagation velocity components of the transient pressure wave on different propagation paths are determined. Based on the propagation velocity components and the topology of the constant cooling water circuit, a pressure wave propagation velocity matrix is ​​constructed; wherein, the pressure wave propagation velocity matrix is ​​used to characterize the propagation characteristics of the transient pressure wave in the constant cooling water circuit.

4. The method according to claim 3, characterized in that, The determination of the fluctuation evolution characteristics of the transient pressure wave based on the pressure change rate includes: Identify the direction of pressure fluctuations in the pressure pulsation data; If the pressure fluctuation direction is positive, the forward difference algorithm is used to calculate the pressure change rate to obtain the second derivative characterizing the pressure acceleration trend. If the pressure fluctuation direction is negative, the backward difference algorithm is used to calculate the pressure change rate to obtain the second derivative characterizing the pressure deceleration trend.

5. The method according to claim 3, characterized in that, The method of coherently synthesizing the pressure pulsation data at the multiple preset monitoring nodes based on the propagation characteristics to obtain the target fluctuation characteristics includes: The propagation direction of the transient pressure wave among the preset monitoring nodes is determined based on the propagation characteristics. Based on the propagation direction and the propagation speed component, the propagation time difference between the preset monitoring nodes is calculated; The pressure pulsation data is subjected to phase compensation processing based on the propagation time difference to obtain compensated pressure pulsation data. The compensated pressure pulsation data are superimposed to generate a synthetic wavefront signal, and the target fluctuation characteristics are determined based on the synthetic wavefront signal.

6. The method according to claim 5, characterized in that, The step of superimposing the compensated pressure pulsation data to generate a synthetic wavefront signal includes: Based on the propagation direction, the signals corresponding to different propagation directions in the compensated pressure pulsation data are respectively divided into forward propagation datasets and reverse propagation datasets; Based on the frequency response characteristics of the multi-channel resonant sensor, the forward propagation dataset and the backward propagation dataset are weighted by amplitude to obtain weighted multi-node component data. The weighted multi-node component data is superimposed to generate the synthetic wavefront signal.

7. The method according to claim 1, characterized in that, The step of reconstructing the phase space trajectory of the target fluctuation characteristics includes: Identify the reference time interval of the fluctuation cycle in the target fluctuation feature, and divide the time into continuous time windows based on the reference time interval; Within each time window, the original state point is determined based on the peak time and amplitude of the target fluctuation characteristics, and a high-dimensional state point is constructed by combining the amplitude evolution components of each original state point within a preset span. By connecting the high-dimensional state points in chronological order, the phase space trajectory is obtained.

8. The method according to claim 7, characterized in that, The determination of the phase space feature vector based on the wave mode abrupt change characteristics of the phase space trajectory includes: The phase space trajectory is divided into multiple local trajectory segments, and multi-scale grid division is performed on each local trajectory segment to obtain the multi-scale grid distribution data corresponding to the local trajectory segment. Based on the multi-scale grid distribution data, the number of times each local trajectory segment crosses each multi-scale grid is counted, and the fractal dimension value of each local trajectory segment is calculated based on the number. The wave mode abrupt change feature is determined by detecting the change in the fractal dimension value of adjacent local trajectory segments; wherein the wave mode abrupt change feature is used to quantitatively characterize the spatial complexity anomaly of the phase space trajectory. The local trajectory segments that produce the abrupt change characteristics of the wave mode are identified as anomalous trajectory segments; High-dimensional state points are extracted from the abnormal trajectory segments, and the phase space feature vector is constructed based on the high-dimensional state points in the abnormal trajectory segments.

9. The method according to claim 1, characterized in that, The determination of monitoring results based on the phase space feature vector includes: Calculate the spatial similarity between the phase space feature vector and the preset reference pattern vector to obtain the spatial distribution pattern encoding; The spatial distribution pattern encoding is matched with a pre-stored leakage location pattern library to obtain the overlap matching result; The monitoring results are determined based on the overlap matching results.

10. The method according to claim 9, characterized in that, The determination of the monitoring result based on the overlap matching result includes: The region identifiers of each of the fixed cooling water loops are sorted in descending order of the overlap matching results to obtain the loop region sorting sequence; Extract the overlap degree corresponding to the region identifier ranked first in the loop region sorting sequence to obtain the target overlap degree set; The values ​​in the target overlap set are subjected to probability transformation processing to obtain the probability distribution corresponding to each region identifier, and the monitoring results are constructed based on the probability distribution.

11. A monitoring device for generator stator cooling water leakage fault, characterized in that, include: The acquisition module is used to acquire pressure pulsation data at multiple preset monitoring nodes in the constant cooling water circuit of the generator to be monitored; The pressure pulsation data is used to characterize the fluctuation characteristics of the transient pressure wave generated in the constant cooling water circuit under the condition of water flow rate variation. The first determining module is used to determine the propagation characteristics of the transient pressure wave in the constant cooling water circuit based on the pressure pulsation data, and to coherently synthesize the pressure pulsation data at the multiple preset monitoring nodes based on the propagation characteristics to obtain the target fluctuation characteristics. The second determining module is used to reconstruct the phase space of the target wave characteristics to obtain the phase space trajectory, and to determine the phase space feature vector based on the wave mode abrupt change characteristics of the phase space trajectory. The monitoring module is used to determine the monitoring results based on the phase space feature vector; the monitoring results are used to characterize the leakage location of the constant cooling water circuit.

12. An electronic device, characterized in that, include: At least one processor; Memory for storing the at least one processor-executable instruction; The at least one processor is configured to execute the instructions to implement the method as described in any one of claims 1-10.

13. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is enabled to perform the method as described in any one of claims 1-10.

14. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-10.

Images