A wireless temperature measurement method and device for a reactor based on frequency conversion technology

Abstract

The application discloses a wireless temperature measurement method and device for a reactor based on a variable frequency technology, and belongs to the technical field of computers. The technical scheme provided by the embodiment of the application solves the problem that signals are prone to loss in a strong electromagnetic interference environment, and a circulating frequency conversion mechanism ensures reliable communication of a large-scale sensor node. A multi-parameter synchronous sensing and feature decoupling mechanism overcomes the defect that single temperature monitoring cannot comprehensively identify composite faults, and realizes accurate diagnosis of early faults of the reactor. Feedback of the result forms a closed loop, so that the monitoring strategy can be self-adaptive to a fault evolution process, and the intelligent level of state monitoring and the diagnosis reliability are significantly improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method and device for wireless temperature measurement of reactors based on frequency conversion technology. Background Technology

[0002] As a crucial component of the power system, the operating status of reactors directly impacts the stability and security of the power grid. Currently, the commonly used wireless temperature measurement technology primarily achieves single-parameter monitoring by deploying temperature sensors, but this method has significant limitations. In environments with strong electromagnetic interference, traditional fixed-frequency communication methods are prone to signal loss, and limited address capacity makes it difficult to support the deployment of large-scale sensor nodes.

[0003] Existing monitoring solutions are severely inadequate in fault identification capabilities. Early faults in reactors, such as inter-turn short circuits and core loosening, often manifest as changes in multi-dimensional physical characteristics, including abnormal temperature, altered vibration characteristics, and magnetic field distortion. A single temperature parameter cannot comprehensively reflect these complex fault states, hindering early warning and accurate diagnosis. Although the industry has attempted to introduce multi-parameter monitoring, the discrete sensor layout leads to asynchronous data acquisition and a lack of effective multi-parameter fusion analysis mechanisms.

[0004] Therefore, existing reactor monitoring technologies are unable to achieve synchronous perception and intelligent diagnosis of multi-dimensional state parameters in environments with strong electromagnetic interference, and cannot meet the needs of upgrading power equipment condition monitoring from simple parameter acquisition to intelligent diagnosis. Summary of the Invention

[0005] This application provides a wireless temperature measurement method and device for reactors based on frequency conversion technology, the technical solution of which is as follows:

[0006] On the one hand, a wireless temperature measurement method for reactors based on frequency conversion technology is provided, the method comprising:

[0007] Dynamic frequency conversion scanning is performed. Based on the electromagnetic environment of the reactor and the sensor topology, the working frequency band is divided into a basic inspection frequency band and a pre-diagnosis frequency band. In the basic inspection frequency band, cyclic frequency conversion addressing is performed on the reactor sensor nodes to obtain monitoring data. When abnormal spectral characteristics are identified, high-density sampling is triggered on the target node cluster in the pre-diagnosis frequency band to obtain a composite frequency signal modulated by temperature, vibration and magnetic field parameters.

[0008] The composite frequency signal selected based on the abnormal spectral features is subjected to feature decoupling to obtain temperature feature parameters, vibration spectral feature parameters, and magnetic field harmonic feature parameters;

[0009] Based on the temperature characteristic parameters, the vibration spectrum characteristic parameters, and the magnetic field harmonic characteristic parameters, the inter-turn short circuit identification result, the core loosening judgment result, and the insulation aging assessment result are generated.

[0010] The results are fed back to the dynamic frequency conversion scan, and the allocation of the pre-diagnostic frequency band and the scanning sequence of the target node cluster are dynamically adjusted based on the fault type. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of the implementation environment of a wireless temperature measurement method for a reactor based on frequency conversion technology provided in an embodiment of this application;

[0013] Figure 2 This is a flowchart of a wireless temperature measurement method for a reactor based on frequency conversion technology provided in an embodiment of this application;

[0014] Figure 3 This is a partial flowchart of a wireless temperature measurement method for a reactor based on frequency conversion technology provided in an embodiment of this application;

[0015] Figure 4 This is a partial flowchart of another wireless temperature measurement method for reactors based on frequency conversion technology provided in an embodiment of this application;

[0016] Figure 5 This is a partial flowchart of another wireless temperature measurement method for reactors based on frequency conversion technology provided in the embodiments of this application;

[0017] Figure 6 This is a partial flowchart of another wireless temperature measurement method for reactors based on frequency conversion technology provided in the embodiments of this application;

[0018] Figure 7 This is a schematic diagram of the structure of a wireless temperature measurement device for a reactor based on frequency conversion technology provided in an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0020] In this application, the terms "first," "second," etc., are used to distinguish identical or similar items with essentially the same function. It should be understood that there is no logical or temporal dependency between "first," "second," and "nth," nor are there any restrictions on quantity or execution order.

[0021] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0022] Figure 1 This is a schematic diagram illustrating the implementation environment of a wireless temperature measurement method for reactors based on frequency conversion technology, as provided in an embodiment of this application. (See attached diagram.) Figure 1 This implementation environment may include node 110 and server 140.

[0023] Node 110 is connected to server 140 via a wireless or wired network. Node 110 has an application installed and running that supports wireless temperature measurement of reactors based on frequency conversion technology.

[0024] Server 140 is a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms. Server 140 can provide background services for applications running on node 110.

[0025] In this embodiment of the application, since a large number of calculation processes are involved, the computing power of node 110 may not be sufficient to meet the requirements. Therefore, the relevant data will be sent to server 140, and server 140 will execute the technical solution provided in this embodiment of the application.

[0026] In the field of reactor condition monitoring, wireless temperature measurement technology faces the challenge of synchronous sensing and intelligent diagnosis of multi-dimensional condition parameters under strong electromagnetic interference environments. Electromagnetic interference leads to unstable communication links, and fixed-frequency communication methods are prone to signal loss. Furthermore, the limited address capacity of sensor nodes makes it difficult to meet the needs of large-scale deployment. In addition, single-parameter monitoring cannot comprehensively capture the multi-dimensional physical characteristic changes caused by complex faults such as inter-turn short circuits, core loosening, and insulation aging. These changes include the coupling effects of temperature anomalies, vibration characteristic alterations, and magnetic field distortion, resulting in insufficient fault identification capabilities and consequently affecting the reliability and diagnostic accuracy of the monitoring system.

[0027] For example, in the operation of reactors in a 500kV substation, when an initial inter-turn short-circuit fault occurs, the sudden change in the electromagnetic environment causes the fixed-frequency communication link to be interrupted. Although the temperature sensor data is within the normal threshold range, the vibration and magnetic field sensors have detected abnormal characteristics. Due to the discrete sensor layout causing asynchronous data acquisition and the lack of a multi-parameter fusion analysis mechanism, the monitoring system cannot correlate the covariant trend of vibration spectrum characteristics and magnetic field harmonic characteristics, resulting in the omission of fault characteristics. The diagnostic process relies solely on a single temperature parameter and cannot identify the essence of the fault.

[0028] If the above problems are not effectively resolved, early warning of potential reactor faults will be difficult to achieve, and the fault state may continue to evolve and spread, resulting in compromised monitoring data integrity and reduced reliability of fault results. This could lead to misjudgments of equipment operating status, threaten power grid stability, and increase safety risks during operation and maintenance.

[0029] In response, this application proposes a wireless temperature measurement method for reactors based on frequency conversion technology, see [link to relevant documentation]. Figure 2 Taking the server as the executing entity as an example, the method includes the following steps.

[0030] 201. Perform dynamic frequency conversion scanning. Based on the electromagnetic environment of the reactor and the sensor topology, the working frequency band is divided into the basic inspection frequency band and the pre-diagnosis frequency band. In the basic inspection frequency band, cyclic frequency conversion addressing is performed on the reactor sensor nodes to obtain monitoring data. When abnormal spectral characteristics are identified, high-density sampling is triggered on the target node cluster in the pre-diagnosis frequency band to obtain the composite frequency signal modulated by temperature, vibration and magnetic field parameters.

[0031] 202. Perform feature decoupling on the composite frequency signal selected based on abnormal spectrum features to obtain temperature feature parameters, vibration spectrum feature parameters, and magnetic field harmonic feature parameters.

[0032] 203. Based on temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters, generate inter-turn short circuit identification results, core loosening judgment results, and insulation aging assessment results.

[0033] 204. Feedback the results to the dynamic frequency conversion scan, and dynamically adjust the allocation of the pre-diagnostic frequency band and the scanning sequence of the target node cluster based on the fault type.

[0034] This application relates to a wireless temperature measurement method for reactors based on frequency conversion technology. Its core lies in achieving reliable monitoring under strong electromagnetic interference environments through dynamic frequency band management and a multi-physical quantity collaborative processing mechanism. In practical applications, dynamic frequency conversion scanning refers to the process of adjusting the communication frequency point according to real-time changes in the electromagnetic environment. This can be achieved by switching between preset frequency point sequences at fixed time intervals, or by dynamically selecting frequencies with less interference based on historical communication quality data. For example, by periodically scanning and recording the bit error rate of each frequency point and prioritizing frequencies with low bit error rates, the main purpose is to achieve adaptive adjustment of the operating frequency band to avoid persistent interference sources. Dividing the operating frequency band into a basic inspection band and a pre-diagnostic band refers to determining the frequency band range for different functions based on the electromagnetic interference distribution characteristics measured on-site and the spatial layout of sensor nodes. This can be achieved through manual division based on interference level thresholds. For example, continuous frequency bands with background noise below a preset value can be designated as the basic inspection band, while discrete low-interference frequency points can be allocated as pre-diagnostic bands. This is mainly to achieve differentiated allocation of spectrum resources to meet the different needs of routine monitoring and in-depth diagnosis. Specifically, cyclic frequency conversion addressing is the process of sequentially switching frequency points to query sensor nodes within the basic inspection frequency band. It can employ a polling mechanism to send addressing commands on multiple frequency points in turn, such as traversing all preset frequency points within a fixed time window to complete node data acquisition. Its main purpose is to achieve orderly communication among large-scale sensor nodes to avoid address conflicts. Furthermore, identifying abnormal spectral characteristics refers to detecting significant deviations in amplitude or frequency in the signal. This can be determined by setting amplitude thresholds; for example, triggering an anomaly marker when the signal amplitude exceeds the historical average plus three standard deviations. Its main purpose is to promptly capture sudden interference changes or signal distortions to initiate a deep monitoring process. As a preferred implementation, acquiring a composite frequency signal modulated by temperature, vibration, and magnetic field parameters refers to synchronously acquiring carrier signals modulated by multiple physical quantities through multi-channel sensors. This can be achieved using a single sensor fusion multi-parameter acquisition circuit, such as using piezoelectric elements to simultaneously sense vibration and magnetic field changes and modulate the temperature signal. Its main purpose is to achieve co-source acquisition of multi-dimensional state parameters to eliminate data asynchrony problems caused by discrete sensors. Feature decoupling is the process of separating the independent features of each parameter from a composite frequency signal. It can employ bandpass filter banks to separate signal components in different frequency bands; for example, a low-pass filter can be used for temperature gradient components, a bandpass filter for vibration modulation components, and a notch filter for magnetic field harmonic components. Its main purpose is to suppress cross-interference and obtain pure feature parameters. Specifically, the result generation is a process of determining the fault type based on the feature parameters. This can be done by comparing preset thresholds, such as comparing the standard deviation of the temperature feature parameter with a threshold to assess the insulation aging state. Its main purpose is to achieve quantified mapping of multi-dimensional fault features to support accurate diagnostic decisions.Therefore, feeding the results back to dynamic frequency conversion scanning refers to the mechanism of adjusting subsequent monitoring strategies according to the fault type. It can manually update the frequency band allocation table based on the urgency of the fault, such as increasing the allocation ratio of dedicated frequency resources for high-risk fault areas. Its main purpose is to achieve dynamic optimization of monitoring resources to adapt to the fault evolution process.

[0035] This embodiment effectively solves the problem of synchronous sensing and intelligent diagnosis of multi-dimensional state parameters of reactors under strong electromagnetic interference environments through the organic combination of the above-mentioned technical features. Specifically, the dynamic frequency band allocation and cyclic frequency conversion addressing mechanism avoids signal loss caused by fixed-frequency communication and supports the deployment requirements of large-scale sensor nodes. High-density sampling triggered by abnormal spectrum characteristics and acquisition of composite frequency signals enable the synchronous acquisition of temperature, vibration, and magnetic field parameters, overcoming the limitation that a single temperature parameter cannot comprehensively identify complex faults such as inter-turn short circuits, core loosening, and insulation aging. Feature decoupling and the generation of multi-parameter results establish a direct correlation between physical characteristics and fault types, while the result feedback mechanism forms a closed-loop control of sensing-diagnosis-optimization, enabling the monitoring system to dynamically adjust frequency band allocation and scanning sequence according to fault type, thereby achieving reliable monitoring and accurate diagnosis of reactor status in complex electromagnetic environments.

[0036] This wireless temperature measurement method for reactors based on frequency conversion technology achieves accurate perception and diagnosis of reactor status under strong electromagnetic interference environments by constructing a closed-loop monitoring mechanism that integrates dynamic frequency conversion and multi-parameter fusion. In specific implementation, the operating frequency band is first divided into a basic inspection band and a pre-diagnosis band, based on the electromagnetic environment characteristics of the reactor and the spatial layout of sensor nodes. Within the basic inspection band, cyclic frequency conversion addressing is performed on the sensor nodes. This dynamically switches communication frequencies to avoid signal loss caused by interference at fixed frequencies. Simultaneously, the cyclic mechanism adapts to the needs of large-scale node deployment, ensuring stable acquisition of routine monitoring data. When the system detects abnormal spectral characteristics, it automatically triggers a high-density sampling mechanism in the pre-diagnosis band, performing in-depth data acquisition on a specific target node cluster to obtain a composite frequency signal modulated by temperature, vibration, and magnetic field parameters. The composite frequency signal is then subjected to feature decoupling processing, separating temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters. The feature decoupling process focuses on processing anomalous signals, obtaining pure parameters by suppressing cross-interference, thus ensuring that the temperature gradient component, vibration modulation component, and magnetic field harmonic component are each properly positioned. Based on the decoupled parameters, the system generates results for inter-turn short circuits, core loosening, and insulation aging. For example, it uses the correlation coefficient between vibration spectrum characteristics and magnetic field harmonic characteristics to identify inter-turn short circuits, determines core loosening based on the difference between vibration and temperature change rates, and assesses insulation aging based on multi-node temperature distribution patterns. Finally, the results are fed back to the dynamic frequency conversion scanning module for dynamically optimizing resource allocation in the pre-diagnostic frequency band and the scanning sequence of the target node cluster, forming an intelligent closed loop of perception-diagnosis-optimization.

[0037] In practical applications, such as in the monitoring of reactors in high-voltage substations, when an abnormal abrupt change in the amplitude of power frequency harmonics is detected in the spectrum during cyclic frequency conversion addressing, the system immediately switches to the pre-diagnostic frequency band and performs high-density sampling of the sensor node cluster at the reactor winding end. After feature decoupling, the acquired composite frequency signal shows a local temperature rise trend in temperature characteristic parameters, energy concentration in a specific frequency band in vibration spectrum characteristic parameters, and a significant enhancement of the third harmonic component in magnetic field harmonic characteristic parameters. Based on this comprehensive assessment, it is determined that there is a risk of inter-turn short circuit. Furthermore, this determination triggers the dynamic frequency conversion scanning module to adjust its resource strategy, allocate dedicated frequency points to the fault area, and reconstruct the scanning topology, thereby achieving precise allocation of monitoring resources.

[0038] Therefore, this method effectively solves the problem of signal loss under strong electromagnetic interference environments, and the cyclic frequency conversion mechanism ensures reliable communication of a large number of sensor nodes. The multi-parameter synchronous sensing and feature decoupling mechanism overcomes the deficiency of single temperature monitoring in comprehensively identifying complex faults, and realizes accurate diagnosis of early faults in reactors. The feedback loop of the results enables the monitoring strategy to adapt to the fault evolution process, significantly improving the intelligence level and diagnostic reliability of condition monitoring.

[0039] In some of the embodiments described above in this application, dynamic frequency conversion scanning is proposed to divide the basic inspection frequency band and the pre-diagnosis frequency band. However, in its implementation, the frequency band division lacks comprehensive consideration of the details of the electromagnetic environment around the reactor and the topology of the sensor node. This may result in the basic inspection frequency band being selected in a high-interference area, affecting the reliability of daily monitoring. The pre-diagnosis frequency band may be allocated in an unreasonable manner, failing to effectively support fault diagnosis. In particular, it may cause signal loss and resource waste in a strong electromagnetic interference environment.

[0040] In response, this application further proposes to perform dynamic frequency conversion scanning, dividing the operating frequency band into a basic inspection frequency band and a pre-diagnostic frequency band based on the reactor's electromagnetic environment and sensor topology. See [link to relevant documentation]. Figure 3 Taking the server as the executing entity as an example, the method includes the following steps.

[0041] 301. Generate an electromagnetic environment map of the environment around the reactor by performing a full-band spectrum scan of the operating frequency band and recording the background noise level and interference source distribution characteristics at each frequency point.

[0042] 302. Based on the electromagnetic environment map and sensor topology, the working frequency band is divided into the basic inspection frequency band and the pre-diagnosis frequency band. The basic inspection frequency band selects the continuous frequency band with the optimal electromagnetic environment, and the pre-diagnosis frequency band is allocated discrete diagnostic frequency points according to the number of sensor nodes.

[0043] 303. Based on the spatial and electrical relationships of each node in the sensor topology, configure corresponding scanning priorities and frequency allocation strategies for the basic inspection frequency band and the pre-diagnosis frequency band respectively.

[0044] In practical applications, an electromagnetic environment map refers to a data model characterizing the distribution of electromagnetic interference around a reactor. It can be implemented using matrix storage of spectrum scan data or visualization of heatmaps, aiming to provide real-time environmental data for frequency band allocation. The basic inspection frequency band can be understood as a low-interference continuous frequency band used for routine monitoring. It can select the frequency band area with the highest signal-to-noise ratio and good stability, aiming to ensure communication continuity and data integrity during daily inspections. The pre-diagnosis frequency band refers to a set of discrete frequency points reserved for fault diagnosis. It can dynamically allocate dedicated frequency resources based on the number of sensor nodes, aiming to support high-density sampling and avoid frequency conflicts. Scanning priority refers to the monitoring sequence determined based on node importance. It can calculate a comprehensive weight value based on node location and electrical relationships, aiming to optimize resource allocation efficiency. The frequency allocation strategy refers to the rule system for mapping frequency points to nodes. It can be implemented using static lookup tables or dynamic scheduling algorithms, aiming to ensure that key nodes receive priority communication guarantees.

[0045] Specifically, the proposed solution first performs a full-band spectrum scan of the operating frequency band to obtain the background noise level and interference source distribution characteristics at each frequency point, generating an electromagnetic environment map reflecting the electromagnetic environment status. This map, along with sensor topology data, serves as the input for frequency band allocation. The basic inspection frequency band selects the continuous frequency band with the optimal electromagnetic environment to ensure the stability of daily monitoring, while the pre-diagnostic frequency band allocates discrete diagnostic dedicated frequency points according to the number of sensor nodes to adapt to changes in node scale. Subsequently, the importance weight of the region is determined based on the spatial positional relationship of nodes in the sensor topology, and the electrical correlation weight is determined by combining the electrical connection relationship. The scanning priority is generated through weight fusion, and frequency point allocation strategies are configured for the basic inspection frequency band and the pre-diagnostic frequency band accordingly. This step-by-step execution mechanism realizes the organic linkage between environmental perception and topology analysis, enabling frequency band resource allocation to dynamically respond to electromagnetic environment fluctuations and node importance distribution, thereby maintaining the reliability of wireless temperature measurement communication under strong interference conditions.

[0046] In one specific implementation, the system performs a full-band scan of the reactor's operating frequency band to identify a continuous frequency band with the lowest interference level and highest stability. This continuous frequency band is designated as the basic inspection frequency band, while several discrete frequency points are assigned to the sensor nodes as preliminary diagnostic frequency bands. Based on the sensor topology, nodes in the winding end and core yoke regions are determined to have higher regional importance weights, and are assigned frequency points with lower background noise in the preliminary diagnostic frequency bands with higher scan priorities. Nodes in the outer casing region, on the other hand, adopt a shared frequency point strategy and are configured with lower scan priorities.

[0047] Through the above scheme, this application realizes the dynamic optimization allocation of frequency band resources, avoids the signal loss problem caused by the basic inspection frequency band entering the high interference area, and ensures that the reserved diagnostic frequency band is reasonably allocated to support fault diagnosis, effectively improving the monitoring reliability and diagnostic resource utilization efficiency of the wireless temperature measurement system under strong electromagnetic interference environment.

[0048] In some of the embodiments described above in this application, an electromagnetic environment map of the environment around the reactor is proposed to support the division of the operating frequency band. However, in its implementation, it only relies on basic spectrum scanning without fine classification of interference characteristics and mining of time-varying patterns. This results in a lack of specificity and dynamic adaptability in electromagnetic environment assessment, and it is impossible to accurately identify the differentiated effects of power frequency harmonics, impulse noise and continuous wave interference. Consequently, the basis for frequency band division is insufficient, and it is difficult to ensure the communication reliability and diagnostic accuracy of wireless temperature measurement in a strong electromagnetic interference environment.

[0049] In this regard, this application further proposes an electromagnetic environment map of the environment surrounding the reactor, including:

[0050] Perform a full-band spectrum scan on the operating frequency band, collect the background noise level, interference signal strength and spectrum occupancy at each frequency point, and record them as raw spectrum data.

[0051] Interference feature analysis is performed on the raw spectrum data to identify the spectral characteristics and time-varying patterns of power frequency harmonic interference sources, impulse noise sources, and continuous wave interference sources, and the interference feature analysis results are obtained.

[0052] Based on the interference characteristic analysis results, an electromagnetic environment quality assessment map is generated, which includes the signal-to-noise ratio distribution, interference source distribution characteristics, and time stability index of each frequency band.

[0053] Full-band spectrum scanning refers to the systematic detection of all frequency points within the operating frequency band. This can be achieved using continuous scanning or step-by-step scanning methods, aiming to comprehensively capture potential interference sources in the reactor's electromagnetic environment and avoid missing key interferences due to partial scanning. Interference feature analysis involves pattern recognition of raw spectrum data to distinguish different interference types. This can be achieved using wavelet transform combined with feature extraction algorithms or deep learning-based classification models, aiming to accurately identify the physical characteristics and time-varying behavior of power frequency harmonics, impulse noise, and continuous wave interference. Electromagnetic environment quality assessment maps are visual representations integrating multi-dimensional assessment indicators. They can be implemented using heatmaps overlaid with contour maps or 3D surface maps, aiming to intuitively present frequency band quality characteristics and provide a basis for subsequent frequency band allocation decisions.

[0054] Specifically, the proposed solution first performs a full-band spectrum scan to acquire complete raw spectrum data, ensuring the electromagnetic environment sensing coverage is broad-spectrum. Then, interference characteristic analysis is performed on the raw data, decoupling the mixed interference signals into three typical interference sources: power frequency harmonics, impulse noise, and continuous waves. Power frequency harmonics originate from the inherent characteristics of the power system, impulse noise is related to transient interference caused by switching operations, and continuous wave interference is related to external communication equipment. By classifying and identifying their spectral characteristics and time-varying patterns, the traditional method avoids the problem of conflating all interference. Finally, an electromagnetic environment quality assessment map is generated based on the analysis results. This map guides the selection of low-noise frequency bands to improve communication quality through signal-to-noise ratio distribution, accurately locates high-interference areas to avoid risks through interference source distribution characteristics, and reflects the long-term availability of frequency bands through time stability indicators. The synergy of these three factors ensures that frequency band allocation meets both immediate communication needs and takes into account dynamic environmental changes, thus providing a high-precision environmental basis for dynamic frequency conversion scanning.

[0055] As a specific embodiment, the solution of this application is implemented as follows: A Rohde & Schwarz FSW series spectrum analyzer is used to perform a step-by-step scan of the 30MHz to 6GHz operating frequency band, with a step interval of 100kHz, to collect background noise levels, interference signal strength, and spectral occupancy at each frequency point. Wavelet packet decomposition algorithm is applied to the raw spectrum data for time-frequency analysis, identifying the characteristics of power frequency harmonic interference sources as energy accumulation at 50Hz and its integer multiples, impulse noise sources as transient spikes in the time domain with a broadband distribution in the frequency domain, and continuous wave interference sources as narrowband continuous signals. Based on the interference characteristic analysis results, an electromagnetic environment quality assessment map is generated using Python's Matplotlib library, where the signal-to-noise ratio distribution is presented as a heatmap, the interference source distribution is marked with icons, and time stability indicators are represented by layers of different transparency, thus intuitively demonstrating the applicability of each frequency band.

[0056] Through the above scheme, this application can achieve accurate identification and dynamic evaluation of different types of interference sources in the electromagnetic environment of reactors, expand the frequency band division basis from a single noise level to multi-dimensional quality indicators, and effectively improve the reliability and diagnostic accuracy of wireless temperature measurement communication in strong electromagnetic interference environment.

[0057] Traditional wireless temperature measurement technology for reactors generates electromagnetic environment quality assessment maps based solely on interference feature analysis results, lacking weighted fusion and spatial mapping of frequency parameters. This results in the maps failing to accurately characterize the spatial distribution of the electromagnetic environment, thus affecting the accuracy of frequency band division and the reliability of subsequent temperature measurements.

[0058] In response, this application further proposes to extract the interference signal strength, spectrum occupancy, and time-varying parameters for each frequency point based on the interference feature analysis results.

[0059] The interference signal strength, spectrum occupancy, and time-varying parameters are weighted and fused to obtain the interference strength weight and time-varying stability coefficient for each frequency point.

[0060] By inputting the interference intensity weight and time-varying stability coefficient into the electromagnetic environment spatial mapping model based on the sensor topology, spatial distribution data of interference intensity and stability are obtained.

[0061] An electromagnetic environment quality assessment map is generated based on spatial distribution data. The map simultaneously presents a spatial interference heatmap based on interference intensity weights and a stability stratification map based on time-varying stability coefficients.

[0062] In practical applications, interference signal strength refers to the energy level of electromagnetic noise at a frequency point. It can be measured using an absolute level or relative signal-to-noise ratio by a spectrum analyzer, aiming to quantify the severity of the impact of external interference sources on that frequency point. Spectrum occupancy can be understood as the probability of a frequency point being occupied over time. It can be achieved by statistically analyzing the proportion of active states of a frequency point per unit time, reflecting the dynamic competition for frequency band resources. Time-varying parameters specifically refer to the statistical characteristics of interference intensity changing over time. They can be achieved using the standard deviation or rate of change calculated with a sliding window, aiming to capture the periodic or sudden behavioral characteristics of interference sources. Interference intensity weighting refers to the quantitative representation of multi-dimensional interference indicators. It can be achieved by setting different weighting coefficients for linear or nonlinear fusion, aiming to form an objective assessment of the overall interference level at a frequency point. The time-varying stability coefficient can be understood as a time-domain consistency index of the interference characteristics at a frequency point. It can be achieved by calculating the time series variance or autocorrelation function of the interference intensity, aiming to distinguish between stable and transient interference. Electromagnetic environment spatial mapping models refer to mathematical frameworks that correlate frequency domain parameters with physical space. These models can be implemented by constructing inverse distance weighted interpolation models or radial basis function models based on the geometric coordinates of sensor nodes, aiming to establish a mapping relationship between frequency characteristics and spatial location. Spatial interference heatmaps are specifically a visual representation of interference intensity in physical space. They can be achieved by mapping interference intensity weights to color gradients and overlaying them with a geographic coordinate system, aiming to visually display interference hotspots. Stability stratification maps refer to the spatial hierarchical representation of time-varying stability. They can be achieved by setting stability thresholds to generate isotropic regions and using semi-transparent layers for overlay, aiming to identify the spatial distribution of different stability levels.

[0063] Specifically, the proposed solution first extracts interference signal strength, spectrum occupancy, and time-varying parameters for each frequency point from the interference characteristic analysis results. These parameters characterize the frequency point's features from three dimensions: energy level, resource competition, and temporal dynamics. Then, a weighted fusion mechanism integrates these three parameters into interference strength weights and time-varying stability coefficients. This fusion process, based on a pre-defined weight allocation strategy, highlights the contribution of key interference factors, enabling the interference strength weights to comprehensively reflect the overall interference level of the frequency point, while the time-varying stability coefficients accurately characterize the temporal stability of the frequency point. Next, these coefficients are input into an electromagnetic environment spatial mapping model constructed based on sensor topology. This model utilizes the spatial coordinate information of sensor nodes to map the frequency point characteristics from the frequency domain to physical space, generating a dataset reflecting the spatial distribution of interference strength and stability. Finally, based on this spatial distribution data, a spatial interference heatmap and a stability stratification map are simultaneously generated. The heatmap visually presents the spatial distribution pattern of interference strength through color gradients, while the stratification map uses transparency differences to identify the spatial characteristics of stability levels. The fusion of these two data forms a comprehensive electromagnetic environment assessment view that includes spatial dimensions, thus providing an accurate spatial distribution basis for frequency band division.

[0064] As a preferred embodiment, the solution of this application is implemented as follows: The electromagnetic environment spatial mapping model can be constructed using a Kriging interpolation algorithm based on sensor node coordinates. This algorithm utilizes the spatial autocorrelation between nodes to map the frequency interference intensity weights to the reactor's three-dimensional spatial grid. The spatial interference heatmap is achieved by normalizing the interference intensity weights and mapping them to a red-yellow-blue gradient color spectrum, where the red area indicates the high interference intensity region. The stability stratification map divides the space into three levels—high stability, medium stability, and low stability—by setting a stability threshold, with each level displayed as a superimposed semi-transparent layer with different transparency. In the reactor winding end region, due to the dense distribution of sensor nodes, the spatial mapping model can automatically enhance the interpolation accuracy of this region, enabling the heatmap to clearly present the local interference hotspots at the winding end.

[0065] Through the above scheme, this application can accurately characterize the spatial distribution characteristics of the electromagnetic environment around the reactor, so that the frequency band division accurately reflects the interference distribution law of the physical space, thereby improving the pertinence of the pre-diagnostic frequency band allocation and effectively ensuring the reliable transmission of wireless temperature measurement signals in a strong electromagnetic environment.

[0066] In some of the embodiments described above in this application, an electromagnetic environment quality assessment map based on spatial distribution data is proposed to intuitively display the state of the electromagnetic environment. However, in its implementation, the numerical spatial distribution data is difficult to use directly for on-site diagnosis. The lack of effective visualization means leads to low efficiency in interference source location and stability assessment. Operators cannot quickly identify high interference areas and weak points in stability, which in turn affects the accuracy of frequency band division and fault diagnosis in strong electromagnetic interference environments.

[0067] In this regard, this application further proposes the following steps for generating an electromagnetic environment quality assessment map based on spatial distribution data:

[0068] The spatial distribution data is normalized and gridded to obtain a standardized spatial distribution matrix of interference intensity and a spatial distribution matrix of stability.

[0069] Based on the spatial distribution matrix of interference intensity, a spatial interference heatmap is generated through color mapping. There is a correspondence between the interference intensity value and the color gradient in the spatial interference heatmap.

[0070] Based on the stability spatial distribution matrix, a stability stratification map is generated using a contour extraction algorithm. Different stability levels in the stability stratification map are represented by layers with different transparency.

[0071] The spatial interference heatmap and the stability stratification map are fused together, and the spatial location markers of the main interference sources are marked in the fused map to form an electromagnetic environment quality assessment map.

[0072] Specifically, normalization and gridding refer to the process of converting the original spatial distribution data into a unified dimension and dividing it into spatial grid units. This can be achieved using min-max normalization or Z-score standardization methods. The aim is to eliminate the influence of different physical dimensions, making interference intensity and stability indices comparable, while discretizing the continuous space into a regular matrix structure, providing a data foundation for subsequent visualization. Color mapping to generate spatial interference heatmaps can be understood as a technique that maps numerical interference intensity to visual color gradients. This can be achieved using linear or logarithmic color mapping methods, for example, using a red-yellow-green gradient spectrum to represent interference intensity from low to high. The purpose is to utilize the human eye's sensitivity to color changes to transform abstract numerical values ​​into intuitive visual representations, facilitating rapid identification of high-interference areas. In practical applications, contour extraction algorithms to generate stability layer maps specifically refer to extracting boundary lines with the same stability value using mathematical algorithms. This can be achieved using Marching Squares or Contour Tracing algorithms. The purpose is to accurately capture the spatial variation boundaries of stability indices, with different stability levels represented by layers of different transparency, allowing operators to overlay and view different stable regions. Specifically, layer fusion refers to the technical process of combining spatial disturbance heatmaps with stability layer maps. It can be achieved using alpha mixing or weighted averaging methods. Its purpose is to simultaneously retain both disturbance intensity and stability information, achieving information complementarity through layer overlay and avoiding the limitations of single-dimensional display.

[0073] Specifically, the proposed solution first normalizes and grids the spatial distribution data, transforming the raw data into standardized spatial distribution matrices for interference intensity and stability. This step ensures that data of different dimensions are unified to a comparable scale and discretizes the continuous space into regular units, laying a structured foundation for subsequent visualization. Subsequently, a spatial interference heatmap is generated based on the spatial distribution matrix for interference intensity using color mapping. Color gradients are used to visually represent the interference intensity distribution, making high-interference areas readily apparent. Simultaneously, a stability stratification map is generated based on the spatial distribution matrix for stability using contour extraction algorithms. Different transparency layers represent different stability levels, accurately reflecting stability boundaries. Finally, the spatial interference heatmap and the stability stratification map are fused, and spatial location markers for major interference sources are added. This fusion not only preserves both interference intensity and stability information but also directly links physical locations through spatial location markers, enabling the map to simultaneously possess global situational awareness and local detail positioning capabilities. This solves the problem that numerical data is difficult to use intuitively for on-site diagnosis.

[0074] As a specific implementation method, the scheme of this application is implemented as follows: In the reactor monitoring system, general image processing software such as MATLAB or Python's Matplotlib library is used to process the spatial distribution data. The normalization process uses min-max normalization to map the interference intensity values ​​to a standard range, and gridding divides the reactor space into regular small grid cells. Color mapping uses a red-yellow-green gradient spectrum to represent the interference intensity, where red corresponds to high interference areas. Contour extraction uses the Marching Squares algorithm to generate stability boundaries, and different stability levels are distinguished by layer transparency. During layer fusion, alpha blending technology is used to combine the heatmap and layered maps, and the physical location identifiers of the main interference sources are marked in the fused map.

[0075] Through the above scheme, this application realizes the intuitive expression of electromagnetic environment characteristics, enabling operators to quickly identify high interference areas and weak points in stability, significantly improving the efficiency of interference source location and stability assessment, thereby optimizing the frequency band division and fault diagnosis accuracy in strong electromagnetic interference environments.

[0076] In some of the embodiments described above in this application, a strategy based on sensor topology configuration scanning priority and frequency point allocation is proposed to optimize frequency band usage. However, in this process, if the spatial importance and electrical correlation of nodes are not distinguished, it may lead to insufficient allocation of monitoring resources for critical fault areas. Under strong electromagnetic interference, it may be impossible to prioritize the data acquisition quality of fault-prone areas such as winding ends and core yokes, thereby delaying the identification of early faults such as inter-turn short circuits and core loosening.

[0077] To address this, this application further proposes a strategy for configuring corresponding scanning priorities and frequency allocation strategies for the basic inspection frequency band and the pre-diagnostic frequency band, based on the spatial and electrical relationships of the nodes in the sensor topology. This includes: determining regional importance weights based on the spatial relationships of the nodes in the sensor topology, where nodes located at the ends of reactor windings, the core yoke, and the heat dissipation duct area have higher regional importance weights than nodes located in the reactor casing and the middle. Determining electrical correlation weights based on the electrical connections of the nodes in the sensor topology, where nodes directly connected to the winding input, responsible for inter-turn voltage monitoring, and located on the main magnetic flux path have higher electrical correlation weights than nodes used only for environmental monitoring. Determining the comprehensive scanning priority of each node based on the regional importance weights and electrical correlation weights, and allocating frequencies with lower background noise levels in the pre-diagnostic frequency band to nodes with the highest comprehensive scanning priority (by a predetermined proportion). Generating a scan sequence lookup table based on the comprehensive scanning priority, and establishing frequency allocation mapping tables for the basic inspection frequency band and the pre-diagnostic frequency band respectively, thus completing the configuration of the scanning priority and frequency allocation strategies.

[0078] Specifically, regional importance weight refers to the evaluation index assigned based on the spatial location of sensor nodes within the reactor's physical structure. This can be achieved using a predefined weight matrix based on historical fault statistics or a dynamically calculated spatial risk distribution map via a machine learning model. The aim is to prioritize the allocation of limited monitoring resources to high-risk areas with concentrated electromagnetic stress and frequent heat exchange. Electrical correlation weight can be understood as a quantitative parameter reflecting the close correlation between sensor nodes and the reactor's core electrical state. This can be implemented using graph theory algorithms based on electrical topology analysis or a pre-defined weight rule base based on the node monitoring task type. The aim is to differentiate the direct contribution of nodes to fault feature capture. Comprehensive scan priority refers to the node monitoring priority ranking that integrates both spatial location and electrical correlation dimensions. This can be achieved by normalizing and fusing regional importance weight and electrical correlation weight using a weighted summation method or an analytic hierarchy process. The aim is to establish a comprehensive and objective node importance evaluation system. The scan sequence lookup table can be understood as a database structure storing node scan order and time parameters. This can be implemented using a hash table or a binary search tree data structure. The aim is to quickly retrieve scan sequences across different frequency bands. A frequency allocation mapping table is a configuration table that records the correspondence between nodes and frequency points. It can be implemented using key-value pairs or a two-dimensional matrix, with the aim of realizing dynamic mapping and management of spectrum resources.

[0079] This application's solution first identifies fault-prone areas based on the spatial relationship of sensor topology, assigning higher importance weights to areas such as winding ends and core yokes. Simultaneously, it differentiates node functional attributes based on electrical connections, giving nodes undertaking core monitoring tasks higher electrical correlation weights. Subsequently, the dual weights are fused to generate a comprehensive scanning priority, ensuring key nodes have a dominant position in resource allocation. Furthermore, high-priority nodes are preferentially allocated high-quality frequencies with low background noise levels in the pre-diagnostic frequency band, directly avoiding strong electromagnetic interference. Finally, through differentiated configuration of the scan sequence lookup table and frequency allocation mapping table, the basic inspection frequency band achieves efficient routine monitoring coverage, while the pre-diagnostic frequency band precisely focuses on high-density sampling of abnormal areas. These steps form a closed-loop optimization mechanism, deeply coupling spectrum resource allocation with the physical structure characteristics and electrical state characteristics of the reactor, enabling the monitoring strategy to dynamically adapt to the fault evolution pattern.

[0080] As a specific implementation method, the solution of this application is implemented as follows: In the reactor monitoring system, the processing unit uses a microcontroller with an ARM Cortex-M7 architecture as the core computing module, and obtains the spatial coordinates and electrical connection information of nodes by parsing the sensor topology database. For determining the regional importance weight, the system divides the reactor's three-dimensional model into multiple spatial regions, sets initial weight values ​​based on historical failure rate data of regions such as winding ends and core yokes, and dynamically corrects them using runtime environment parameters. For determining the electrical correlation weight, the system classifies and assigns values ​​based on attributes such as whether a node is directly connected to the winding input terminal and whether it undertakes inter-turn voltage monitoring tasks, using a preset rule base. The comprehensive scan priority calculation module uses a weight ratio of 0.6:0.4 to fuse the regional importance weight and the electrical correlation weight, generating a node priority ranking list. The frequency point allocation module selects the 10% of frequency points with the lowest background noise level from the pre-diagnostic frequency band and prioritizes them for the top 20% of high-priority nodes. The scan sequence lookup table is stored in an in-memory database; the basic inspection frequency band is configured as a cyclic scan sequence, while the pre-diagnostic frequency band is configured as a priority-sorted skip scan sequence. The frequency point allocation mapping table is implemented through a configuration file. The basic inspection frequency band adopts a fixed frequency point mapping, while the pre-diagnosis frequency band adopts a dynamic frequency point remapping mechanism.

[0081] Through the above technical solutions, this application has achieved precise allocation of monitoring resources for key fault areas in a strong electromagnetic interference environment, effectively ensuring the data acquisition quality of fault-prone areas such as winding ends and core yokes, avoiding the loss of key signals due to interference, and significantly improving the timeliness of identification and diagnostic accuracy of early faults such as inter-turn short circuits and core loosening. At the same time, through the differentiated allocation strategy of spectrum resources, the monitoring efficiency is maximized under limited frequency band conditions, meeting the technical requirements for synchronous perception and intelligent diagnosis of multi-dimensional state parameters of reactors.

[0082] In some embodiments of this application, feature decoupling of composite frequency signals is proposed to obtain temperature feature parameters, vibration spectrum feature parameters, and magnetic field harmonic feature parameters. However, in the implementation process, the cross-interference caused by the co-modulation of temperature, vibration, and magnetic field parameters in the composite frequency signal leads to inaccurate feature extraction, which reduces the reliability of fault diagnosis results.

[0083] For this, see Figure 4 Taking the server as the executing entity as an example, the following steps are included.

[0084] 401. Decompose the composite frequency signal into three signal components: temperature gradient component, vibration modulation component, and magnetic field harmonic component.

[0085] 402. By using an adaptive Wiener filter bank, the temperature gradient component, vibration modulation component, and magnetic field harmonic component are purified in the frequency band to obtain purified temperature signal, purified vibration signal, and purified magnetic field signal with suppressed cross-interference.

[0086] 403. Perform time-domain feature extraction on the purified temperature signal, purified vibration signal, and purified magnetic field signal respectively to obtain temperature feature parameters, vibration spectrum feature parameters, and magnetic field harmonic feature parameters.

[0087] The decomposition of a composite frequency signal into three components—temperature gradient component, vibration modulation component, and magnetic field harmonic component—is based on the differences in physical characteristics: slow temperature change, vibration exhibiting as a modulation signal, and magnetic field harmonics concentrated at power frequency harmonics. This separation can be achieved using time-frequency analysis methods such as empirical mode decomposition, wavelet packet decomposition, or short-time Fourier transform. The aim is to utilize the inherent laws of each parameter in the time-frequency domain to achieve preliminary decoupling and avoid feature confusion when directly processing mixed signals. Bandwidth purification of the temperature gradient component, vibration modulation component, and magnetic field harmonic component using an adaptive Wiener filter bank refers to applying optimal filtering techniques under the minimum mean square error criterion to suppress interference in the signal components. This can be achieved using adaptive filtering structures based on the minimum mean square algorithm, recursive least squares algorithm, or Kalman filtering. The goal is to dynamically adjust the filtering parameters according to the bandwidth characteristics of the signal components, effectively suppressing cross-interference and ensuring the purity of the purified signal. Performing time-domain feature extraction on purified temperature, vibration, and magnetic field signals refers to extracting key parameters characterizing the physical state from the time-domain waveform. This can be achieved using methods such as envelope analysis, statistical feature calculation, or time-frequency feature extraction. The aim is to obtain characteristic parameters that directly reflect the temperature, vibration, and magnetic field states based on the purity of the purified signals, thus avoiding contamination of fault diagnosis indicators by interference signals.

[0088] Specifically, the proposed solution achieves precise separation of multi-parameter features through a step-by-step decoupling mechanism: First, based on the inherent laws of temperature, vibration, and magnetic field parameters in the time-frequency domain, the composite frequency signal is decomposed into three signal components, and initial separation is achieved using the slow temperature change characteristic, vibration modulation characteristic, and discrete frequency point characteristic of magnetic field harmonics. Second, an adaptive Wiener filter bank is used to dynamically optimize the filtering parameters according to the dominant frequency band characteristics of each signal component, ensuring the purity of the purified signal by minimizing cross-correlation interference between components. Finally, time-domain features are extracted from the purified signal, allowing the feature parameters to directly reflect the essence of the physical state. This step-by-step processing logic forms a complete feature decoupling chain, with each step progressing sequentially and supporting each other, effectively solving the problem of feature extraction distortion caused by cross-interference of multiple parameters.

[0089] As a specific implementation method, the solution of this application is implemented as follows: The signal processing unit adopts a microcontroller based on the ARM Cortex-M7 core. When abnormal spectral characteristics are identified, the unit decomposes the composite frequency signal into temperature gradient components, vibration modulation components, and magnetic field harmonic components through wavelet transform. Subsequently, an adaptive Wiener filter bank based on the least mean square algorithm is applied to purify the frequency band of each component, wherein the filter coefficients are dynamically updated according to the real-time signal characteristics. Finally, envelope analysis is performed on the purified temperature signal to obtain the gradient trend, frequency domain transformation is performed on the vibration signal to calculate the spectral centroid, and harmonic analysis is performed on the magnetic field signal to determine the total harmonic distortion rate, thereby obtaining various characteristic parameters for subsequent fault diagnosis.

[0090] Through the above technical solution, this application effectively suppresses the cross-interference caused by the co-modulation of temperature, vibration and magnetic field parameters in composite frequency signals, improves the accuracy of feature extraction, and thus enhances the reliability of fault diagnosis results.

[0091] In some of the embodiments described above in this application, a frequency band purification method using an adaptive Wiener filter bank is proposed to suppress cross-interference between temperature, vibration, and magnetic field signal components. However, in its implementation, due to the spectral overlap and mutual modulation of multidimensional physical signals in the frequency domain, simple frequency band division cannot effectively separate the signal components, resulting in the purified signal still containing significant cross-interference. This affects the extraction accuracy of temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters, ultimately reducing the reliability of inter-turn short circuit identification, core loosening judgment, and insulation aging assessment.

[0092] In response, this application further proposes to perform frequency band purification of the temperature gradient component, vibration modulation component, and magnetic field harmonic component using an adaptive Wiener filter bank, obtaining purified temperature signal, purified vibration signal, and purified magnetic field signal with suppressed cross-interference, including:

[0093] The dominant frequency bands of the temperature gradient component, vibration modulation component, and magnetic field harmonic component are determined. The dominant frequency band of the temperature gradient component is the lowest frequency band, the dominant frequency band of the vibration modulation component is the mid-frequency band, and the dominant frequency band of the magnetic field harmonic component is the discrete frequency point where the power frequency and its harmonics are located.

[0094] Based on the dominant frequency bands of the temperature gradient component, vibration modulation component, and magnetic field harmonic component, the cross-correlation function between each pair of the three signal components is calculated, and a covariance matrix characterizing the interference intensity between components is constructed based on the peak value of the cross-correlation function.

[0095] With minimizing the off-diagonal elements in the covariance matrix as the optimization objective, the coefficients of the Wiener filter bank are iteratively updated through an adaptive algorithm. Inverse filtering is then applied to the temperature gradient component, vibration modulation component, and magnetic field harmonic component to obtain purified temperature signal, purified vibration signal, and purified magnetic field signal with suppressed cross-interference.

[0096] The adaptive Wiener filter bank refers to a signal processing device that can dynamically adjust filtering parameters according to signal characteristics. It can be implemented using a digital signal processor in conjunction with an adaptive algorithm software module, aiming to optimize filtering performance in real time to adapt to changing electromagnetic environments. The dominant frequency band can be understood as the frequency range where signal energy is mainly concentrated. It can be determined based on prior knowledge of physical characteristics or spectral scanning results, aiming to provide a precise frequency domain positioning basis for signal separation. The cross-correlation function is a mathematical tool for quantifying the similarity between two signals. It can be calculated using the Fast Fourier Transform algorithm, aiming to accurately characterize the interference intensity between signal components. The covariance matrix is ​​a matrix structure describing the linear correlation between multiple random variables. It can be constructed based on the peak values ​​of the cross-correlation function, aiming to systematically characterize the interference relationship between components. Minimizing the off-diagonal elements in the covariance matrix refers to the optimization process that brings the matrix closer to a diagonal state. It can be achieved using the gradient descent method, aiming to significantly reduce the cross-coupling between signal components. The adaptive algorithm is a calculation method that automatically adjusts parameters according to the error signal. It can be implemented using the least mean square algorithm or the recursive least squares method, aiming to continuously improve filtering accuracy. Inverse filtering refers to the process of eliminating signal distortion by applying an inverse system. It can be implemented using frequency domain inverse filtering or time domain deconvolution methods, with the aim of restoring the independent characteristics of each signal component.

[0097] Specifically, the scheme in this application employs a frequency domain partitioning mechanism guided by physical characteristics. First, based on the gradual change characteristics of the temperature signal, its dominant frequency band is positioned in the lowest frequency band, which aligns with the natural law that temperature changes concentrate in the low-frequency range. Simultaneously, based on the modulation characteristics of the vibration signal, its dominant frequency band is set in the mid-frequency band, stemming from the physical nature of mechanical vibration energy concentration in the mid-frequency range. Furthermore, based on the inherent relationship between magnetic field harmonics and the power frequency, its dominant frequency band is designated as discrete frequency points of the power frequency and its harmonics. On this basis, the cross-correlation function between each pair of signal components is calculated using the determined dominant frequency band, and a covariance matrix is ​​constructed using its peak values ​​to quantify the interference intensity. Finally, with minimizing the off-diagonal elements of the covariance matrix as the optimization objective, the Wiener filter coefficients are iteratively updated using an adaptive algorithm, and inverse filtering is applied to each signal component. This process, through frequency domain partitioning guided by physical characteristics, quantitative modeling of interference intensity, and adaptive filtering optimization, achieves high-fidelity separation of multidimensional signals in a strong electromagnetic environment, ensuring the independence of the purified temperature signal, vibration signal, and magnetic field signal.

[0098] As a specific implementation method, the solution of this application is implemented as follows: In the reactor monitoring system, a digital signal processor is used as the core processing unit, which is equipped with an adaptive Wiener filter bank module. The dominant frequency bands of the temperature gradient component, vibration modulation component, and magnetic field harmonic component are determined through spectrum scanning during the system initialization phase. The dominant frequency band of the temperature gradient component is set to the ultra-low frequency band, the dominant frequency band of the vibration modulation component is set to the mid-frequency range, and the dominant frequency band of the magnetic field harmonic component is set to the power frequency fundamental wave and its harmonic discrete points. The cross-correlation function calculation module is implemented using the fast Fourier transform algorithm, and the covariance matrix construction module performs statistical calculations based on the peak value of the cross-correlation function. The coefficient update of the Wiener filter bank adopts the least mean square algorithm, which minimizes the off-diagonal elements of the covariance matrix by continuously adjusting the filter parameters. After the inverse filtering processing module processes each signal component, the purified temperature signal, purified vibration signal, and purified magnetic field signal are sent to the subsequent feature extraction unit.

[0099] Through the above-mentioned solution, this application effectively solves the problem of incomplete signal separation caused by spectral overlap and mutual modulation of multidimensional physical signals in the frequency domain, significantly reduces the cross-interference between temperature, vibration and magnetic field signal components, and improves the extraction accuracy of temperature characteristic parameters, vibration spectrum characteristic parameters and magnetic field harmonic characteristic parameters, thereby enhancing the reliability of inter-turn short circuit identification, core loosening judgment and insulation aging assessment.

[0100] In some of the above implementations, time-domain feature extraction is proposed to generate feature parameters by performing time-domain feature extraction on the purified signal. However, in the implementation process, the purified signal still contains redundant information, and the physical characteristics of temperature, vibration and magnetic field signals are significantly different. It is difficult to extract key features in a targeted manner by directly using the original signal, resulting in feature parameters being greatly affected by noise and having insufficient characterization ability, which cannot accurately support subsequent fault diagnosis.

[0101] In response, this application further proposes to perform empirical mode decomposition on the purified temperature signal, extract the first intrinsic mode function as the temperature gradual trend term, and use the mean of the envelope of the temperature gradual trend term as the temperature feature parameter.

[0102] The purified vibration signal is subjected to frequency domain transformation, and the spectral centroid is determined based on the dominant frequency band of the vibration modulation component. The spectral centroid is then used as a characteristic parameter of the vibration spectrum.

[0103] Harmonic analysis was performed on the purified magnetic field signal. The total harmonic distortion rate was determined based on the fundamental frequency and its main harmonics, and the total harmonic distortion rate was used as a characteristic parameter of the magnetic field harmonics.

[0104] In practical applications, Empirical Mode Decomposition (EMD) is an adaptive signal processing method that can be implemented using the Hilbert-Huang transform algorithm. Its purpose is to decompose nonlinear and non-stationary signals into multiple intrinsic mode function (EMF) components to adapt to the time-frequency characteristics of different physical quantity signals. The EMF refers to a signal component that satisfies local symmetry and has an equal or slightly different number of extrema and zero-crossings. It can be understood as an inherent oscillation mode in the signal, and its purpose is to separate signal components at different time scales. Specifically, the temperature gradual change trend term refers to the signal component that characterizes the slow change trend of temperature. It can be implemented using the first EMF obtained from EMD, aiming to highlight the long-term temperature change characteristics and avoid high-frequency noise interference. In practical applications, the envelope mean refers to the average value of the signal envelope, which can be obtained by calculating the mean after obtaining the signal envelope through Hilbert transform. Its purpose is to quantify the stability level of the temperature gradual change trend and reduce the impact of instantaneous fluctuations. The spectral centroid refers to the centroid frequency of the signal spectral energy distribution, which can be calculated based on the spectral data after Fast Fourier Transform (FFT). Its purpose is to characterize the main energy concentration location of the vibration signal and reflect changes in the vibration state. Specifically, the total harmonic distortion rate (THD) is the ratio of the total effective value of harmonic components to the effective value of the fundamental frequency. It can be calculated by measuring the amplitude of the fundamental frequency and its main harmonic components. The purpose is to quantify the degree of harmonic distortion of the magnetic field signal and indicate fault characteristics.

[0105] Specifically, the proposed solution achieves accurate extraction of key features by employing feature extraction methods tailored to the physical characteristics of the purified temperature, vibration, and magnetic field signals. For the temperature signal, the adaptive properties of empirical mode decomposition are used to separate the first intrinsic mode function representing a slowly changing trend, and short-term fluctuations are smoothed through envelope averaging. For the vibration signal, the spectral centroid is calculated within the dominant frequency band based on its energy distribution characteristics, focusing on the fault-related frequency band. For the magnetic field signal, the total harmonic distortion rate is calculated based on the harmonic distortion characteristics, quantifying the distortion degree at fault-sensitive frequency points. This differentiated processing strategy ensures that each feature parameter can effectively characterize the state changes of the corresponding physical quantity while suppressing noise interference.

[0106] As a specific embodiment, the solution of this application is implemented as follows: The purified temperature signal is subjected to empirical mode decomposition using Hilbert-Huang transform to obtain multiple intrinsic mode functions. The first intrinsic mode function is selected as the temperature gradual change trend term. The envelope of this trend term is obtained through Hilbert transform, and the mean of the envelope is calculated as a temperature characteristic parameter. A fast Fourier transform is performed on the purified vibration signal to obtain the spectrum. Spectral data is filtered according to the dominant frequency band range of the vibration modulation component, and the centroid of the spectral energy distribution within this frequency band is calculated as the spectral centroid. A discrete Fourier transform is performed on the purified magnetic field signal to extract the fundamental frequency and the 3rd and 5th harmonic components. The ratio of the total effective value of these harmonic components to the effective value of the fundamental frequency is calculated as the total harmonic distortion rate.

[0107] Through the above scheme, the key features in the purified signal of this application can be accurately extracted, the temperature feature parameter can effectively characterize the long-term temperature change trend, the vibration spectrum feature parameter can sensitively reflect the vibration state change, and the magnetic field harmonic feature parameter can accurately quantify the degree of magnetic field distortion, thereby improving the robustness and diagnostic pertinence of the feature parameters and providing a reliable foundation for subsequent multi-parameter fusion fault identification.

[0108] To provide a clearer explanation of the above implementation methods, more specific implementation methods are provided below.

[0109] I. Detailed Implementation Process of Temperature Feature Parameter Extraction

[0110] The core of this process is to use Empirical Mode Decomposition (EMD) to adaptively separate signal components and extract trends.

[0111] 1. Empirical Mode Decomposition (EMD) process:

[0112] Input: The purified discrete temperature signal sequence s_temp[n] (n=0,1,...,N-1).

[0113] Initialization: Set the original margin r0[n]=s_temp[n], and the modal index k=1.

[0114] Iterative filtering (extracting the k-th intrinsic mode function (IMF)):

[0115] a. Let h0[n]=r_{k-1}[n].

[0116] b. Identify all local maxima and minima of h_{j-1}[n].

[0117] c. Fit the upper and lower envelopes e_max^{j-1}[n] and e_min^{j-1}[n] respectively using cubic spline interpolation.

[0118] d. Calculate the mean envelope: m_{j-1}[n]=(e_max^{j-1}[n]+e_min^{j-1}[n]) / 2.

[0119] e. Obtain new details: h_j[n] = h_{j-1}[n] - m_{j-1}[n].

[0120] f. Determine if h_j[n] satisfies the IMF condition (the number of extreme points and zero-crossing points is similar, and the mean envelope is close to zero). If it satisfies the condition, then the k-th intrinsic modulus function IMF_k[n] = h_j[n]; otherwise, let j = j + 1 and return to step b.

[0121] Update and Loop: After obtaining IMF_k[n], calculate the new residual r_k[n] = r_{k-1}[n] - IMF_k[n]. If r_k[n] is non-monotonic and contains enough extreme points, then let k = k+1 to continue extracting the next eigenfunction; otherwise, the decomposition ends. Suppose that K eigenfunctions are obtained, and the residual is r_K[n].

[0122] 2. Generation of temperature characteristic parameters:

[0123] The first intrinsic mode function (IMF_1[n]) typically contains the fastest fluctuations and noise in the signal. The temperature gradient trend term T_trend[n] is defined as the margin r_K[n], which is obtained by subtracting all intrinsic mode functions from the original signal and represents the slowest gradient component after adaptive filtering.

[0124] Calculate the arithmetic mean of the temperature gradual change trend term T_trend[n] over a complete diagnostic cycle (e.g., 10.24 seconds), and use it as the temperature feature parameter T_feature:

[0125] T_feature=(1 / N)Σ_{n=0}^{N-1}T_trend[n]

[0126] In the experiment, after processing a measured signal as described above, three intrinsic mode functions were obtained. The mean value of the residual T_trend[n] was 72.8°C after calibration and conversion. This value is T_feature.

[0127] II. Detailed Implementation Process of Vibration Spectrum Feature Parameter Extraction

[0128] This process aims to calculate the centroid of the vibrational energy distribution within the dominant frequency band.

[0129] 1. Frequency domain transformation and determination of dominant frequency band:

[0130] The purified vibration signal s_vib[n] is processed by removing DC and adding a Hanning window, and then the spectrum is obtained by performing a Fast Fourier Transform (FFT).

[0131] Spectral data are filtered based on the dominant frequency band of the vibration modulation component (e.g., determined statistically or preset as [80Hz, 1500Hz]).

[0132] 2. Calculation of the spectral centroid (center of mass of energy distribution):

[0133] Within the dominant frequency band, the centroid f_centroid is calculated as a characteristic parameter of the vibrational spectrum based on the power spectral density P[f_l]=|V[f_l]|². This parameter is the centroid of the spectral energy distribution within this frequency band, and its calculation formula is:

[0134] f_centroid=(Σ_{l:f_low≤f_l≤f_high}(f_l P[f_l])) / (Σ_{l:f_low≤f_l≤f_high}P[f_l])

[0135] In the experiment, under normal conditions, the centroid of the frequency spectrum of the vibration signal of a certain reactor is about 420 Hz; when mechanical loosening occurs, this value can rise to more than 580 Hz.

[0136] III. Detailed Implementation Process of Magnetic Field Harmonic Characteristic Parameter Extraction

[0137] This process quantifies magnetic field waveform distortion by calculating the distortion rate of specific harmonics.

[0138] 1. Harmonic component extraction:

[0139] The purified magnetic field signal was subjected to discrete Fourier transform to accurately extract the fundamental frequency wave and the 3rd, 5th, 7th and 9th harmonic components.

[0140] Calculate the effective value A1 of the fundamental frequency wave and the effective values ​​A3, A5, A7, and A9 of each harmonic component.

[0141] 2. Calculation of Total Harmonic Distortion (THD):

[0142] Calculate the total effective value of harmonic components: Σ_Ah=sqrt(A3²+A5²+A7²+A9²).

[0143] Calculate the total harmonic distortion rate (THD_F) based on the selected characteristic harmonics, and use it as the magnetic field harmonic characteristic parameter H_feature:

[0144] H_feature=THD_F=(ΣAh / A1)100%

[0145] In the example experiment, under healthy conditions, THD_F is usually <5%; when an inter-turn short circuit occurs, THD_F can rise sharply to over 15%.

[0146] In some of the embodiments described above in this application, the generation of inter-turn short circuit identification results, core loosening judgment results, and insulation aging assessment results is proposed to realize reactor fault diagnosis. However, in the implementation process, there is a lack of a multi-parameter correlation feature extraction mechanism for different fault types, which makes it impossible to accurately distinguish specific fault states such as inter-turn short circuit, core loosening, and insulation aging, affecting the accuracy and timeliness of diagnosis.

[0147] To address this, this application further proposes methods for generating inter-turn short circuit identification results, core loosening judgment results, and insulation aging assessment results based on temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters. (See [link to relevant documentation]). Figure 5 Taking the server as the executing entity as an example, the following steps are included.

[0148] 501. Based on the correlation characteristics of vibration spectrum characteristic parameters and magnetic field harmonic characteristic parameters, generate inter-turn short circuit identification results. The correlation characteristics include the correlation coefficient and covariance trend between vibration spectrum characteristic parameters and magnetic field harmonic characteristic parameters.

[0149] 502. Based on the difference between the vibration spectrum characteristic parameters and the temperature characteristic parameters, generate the core loosening judgment result. The difference features include the degree of difference between the rate of change of the vibration spectrum characteristic parameters and the rate of change of the temperature characteristic parameters.

[0150] 503. Based on the distribution pattern of the temperature characteristic parameters of the reactor at multiple nodes, generate insulation aging assessment results. The distribution pattern includes the standard deviation and gradient distribution characteristics of the temperature characteristic parameters of each node.

[0151] Among them, correlation characteristics refer to the statistical dependence between vibration spectrum characteristic parameters and magnetic field harmonic characteristic parameters. This can be achieved using statistical indicators such as Pearson correlation coefficient, Spearman rank correlation coefficient, or mutual information, aiming to quantify the synchronicity and correlation of their changes. Dissimilarity characteristics can be understood as the degree of mismatch between the changing behavior of vibration spectrum characteristic parameters and temperature characteristic parameters. This can be achieved using indicators such as the absolute difference in the rate of change, relative difference ratio, or dynamic deviation index, aiming to capture the parameter decoupling phenomenon unique to faults. Distribution patterns specifically refer to the spatial dispersion and changing trend of the temperature characteristic parameters of multiple nodes in the reactor. This can be achieved using statistical quantities such as standard deviation, variance, gradient distribution, or spatial autocorrelation function, aiming to characterize the uniformity of the temperature field and the diffusion characteristics of local hot spots.

[0152] Specifically, the proposed solution extracts core correlation features among multiple parameters through differential extraction, enabling a deep match between the fault diagnosis mechanism and the physical mechanism of specific faults. For inter-turn short-circuit faults, identification results are generated using the correlation features of vibration spectrum characteristic parameters and magnetic field harmonic characteristic parameters. Since inter-turn short circuits simultaneously induce abnormal winding vibration and magnetic field harmonic distortion, the correlation coefficient quantifies the statistical dependence of their changes, and the covariance trend captures the dynamic law of the synchronous evolution of vibration and magnetic field in the time domain, thus effectively eliminating misjudgments caused by fluctuations in a single parameter. For core loosening faults, judgment results are generated based on the difference features between vibration spectrum characteristic parameters and temperature characteristic parameters. Since core loosening mainly causes abrupt changes in mechanical vibration while the temperature response is relatively slow, the difference index, by comparing the degree of mismatch in the rates of change of the two, can accurately identify the vibration-temperature decoupling phenomenon unique to loosening. In terms of insulation aging assessment, the distribution pattern of multi-node temperature characteristic parameters is used to generate the data. Since insulation aging often causes local hot spot diffusion and temperature field distortion, the standard deviation reflects the overall temperature dispersion and indicates the uniformity of aging. The gradient distribution reveals the propagation trend of aging along the winding. The combination of the two achieves a three-dimensional assessment of the degree of insulation degradation.

[0153] As a specific implementation method, the scheme of this application is implemented as follows: Wireless temperature sensor nodes are deployed at the ends of the reactor windings, the core yoke, and the heat dissipation duct area. The processing unit uses an ARM Cortex-M7 microcontroller to run a feature decoupling algorithm. For inter-turn short circuit identification, the processing unit calculates the Pearson correlation coefficient between the vibration spectrum characteristic parameter and the magnetic field harmonic characteristic parameter, and analyzes their covariance trend in the time domain. For core loosening determination, the processing unit calculates the time series change rate of the vibration spectrum characteristic parameter and the temperature characteristic parameter, and determines the difference index through a sliding window. For insulation aging assessment, the processing unit analyzes the standard deviation of the multi-node temperature characteristic parameter and the average value of the temperature difference between adjacent nodes.

[0154] The above scheme enables precise differentiation of different fault types, significantly improving the accuracy and timeliness of reactor fault diagnosis, and effectively identifying specific fault states such as early inter-turn short circuits, core loosening, and insulation aging.

[0155] In some of the embodiments described above in this application, a correlation feature based on vibration spectrum characteristic parameters and magnetic field harmonic characteristic parameters is proposed to generate inter-turn short circuit identification results. However, in its implementation, the specific implementation method of the correlation feature is not clearly defined. In particular, the calculation object of the correlation coefficient is not limited to specific harmonic components, and the time-domain quantification method of the covariance relationship is not specified. This makes the inter-turn short circuit identification results susceptible to interference from irrelevant harmonics and time asynchrony, making it difficult to achieve high-precision fault state determination and confidence assessment.

[0156] In response, this application further proposes a correlation feature based on vibration spectrum characteristic parameters and magnetic field harmonic characteristic parameters to generate inter-turn short circuit identification results, including:

[0157] Determine the correlation coefficients between the amplitude of the vibration spectrum characteristic parameter and the amplitudes of the third and fifth harmonics in the magnetic field harmonic characteristic parameter.

[0158] Determine the covariant relationship in the time domain between the variation trends of vibration spectrum characteristic parameters and magnetic field harmonic characteristic parameters.

[0159] Based on the comparison results of the correlation coefficient and the first preset threshold, and the comparison results of the covariance relationship and the second preset threshold, an inter-turn short circuit identification result containing the inter-turn short circuit status identifier and confidence level is generated.

[0160] The correlation coefficient is a quantitative indicator of the linear correlation between the amplitude of the vibration spectrum characteristic parameter and the amplitude of a specific harmonic component in the magnetic field harmonic characteristic parameter. It can be implemented using Pearson correlation coefficient or Spearman rank correlation coefficient, aiming to focus on the unique electromagnetic response law of inter-turn short-circuit faults in reactors and effectively isolate interference from the power frequency fundamental wave and other higher harmonics. The covariance relationship can be understood as the synchronicity of the changing trends of the vibration spectrum characteristic parameter and the magnetic field harmonic characteristic parameter in the time dimension. It can be quantified using cross-correlation functions or dynamic time warping algorithms, aiming to capture the physical consistency of fault characteristics during their development and suppress instantaneous misjudgments caused by asynchronous sensor sampling or environmental noise. The dual-threshold judgment mechanism specifically refers to the decision logic that generates fault state identification and confidence level based on the comparison results of the correlation coefficient and covariance relationship. It can be implemented using logical AND gates or weighted fusion rules, aiming to combine quantitative features with preset criteria, giving the results both deterministic and probabilistic evaluation capabilities.

[0161] Specifically, the proposed solution accurately reflects the characteristic coupling strength of inter-turn short-circuit faults by limiting the calculation of the correlation coefficient to the amplitudes of the third and fifth harmonics. Simultaneously, it ensures the synchronization of vibration and magnetic field change trends by quantifying the time-domain covariance relationship. Finally, a dual-threshold collaborative judgment mechanism is employed to comprehensively generate an identification result including state indicators and confidence levels by integrating the comparison results of the correlation coefficient and covariance relationship, thus forming a complete inter-turn short-circuit identification process.

[0162] As a specific implementation method, the solution of this application is implemented as follows: the processing unit adopts an ARM Cortex-M4 microcontroller, the correlation coefficient is calculated using the Pearson correlation coefficient method, the covariance is quantized in the time domain through the cross-correlation function, the dual threshold determination is implemented using a logic AND gate, and when the correlation coefficient exceeds the first preset threshold and the covariance exceeds the second preset threshold, a high-confidence inter-turn short circuit status identifier is generated.

[0163] Through the above scheme, this application can effectively reduce irrelevant harmonic interference and the impact of time asynchrony, improve the accuracy and reliability of inter-turn short circuit identification, and achieve high-precision fault state determination and confidence assessment.

[0164] In some of the embodiments described above in this application, a method for generating a core loosening determination result based on the difference between vibration spectrum characteristic parameters and temperature characteristic parameters is proposed. However, in its implementation, the method for calculating the rate of change lacks robustness, and the definition of the difference index is imprecise, which leads to a decrease in the reliability of core loosening determination under strong electromagnetic interference environment.

[0165] In response, this application further proposes to generate a core loosening determination result based on the difference between vibration spectrum characteristic parameters and temperature characteristic parameters, including:

[0166] The time series rate of change of the vibration spectrum characteristic parameter and the time series rate of change of the temperature characteristic parameter are determined. The time series rate of change is obtained by calculating the first derivative through a sliding window.

[0167] Determine the difference index between the rate of change of the vibration spectrum characteristic parameter and the rate of change of the temperature characteristic parameter. The difference index is the ratio of the absolute difference between the two rates of change to the rate of change of the vibration spectrum characteristic parameter.

[0168] Based on the comparison results between the difference index and the third preset threshold, a core loosening judgment result is generated, which includes the core loosening status indicator and the confidence level.

[0169] The time-series rate of change refers to the instantaneous rate at which the parameter changes over time. It can be achieved using a sliding window combined with the central difference method or by taking the derivative after polynomial fitting. The aim is to suppress electromagnetic noise interference on instantaneous changes through local data smoothing, ensuring the stability of the rate of change estimation under strong interference environments. The difference index can be understood as a quantitative representation of the inconsistency between vibration and temperature change trends. It can be implemented using normalized difference or ratio forms, aiming to eliminate interference from the absolute amplitude of the parameter and highlight the relative difference characteristic of core loosening faults: abnormally aggravated vibration with insignificant temperature changes. The third preset threshold is specifically a benchmark value used to determine the core loosening state. It can be determined based on statistical analysis of historical fault data or offline calibration methods, aiming to provide a reliable decision boundary to distinguish between normal and fault states.

[0170] Specifically, the proposed solution uses a sliding window to locally smooth the time series of vibration spectrum and temperature characteristic parameters, effectively suppressing random noise introduced by strong electromagnetic interference and thus obtaining a stable time series rate of change. Subsequently, the absolute difference between the vibration and temperature rates of change is calculated and normalized into a ratio-based difference index. This index directly reflects the inherent contradiction between abnormal vibration characteristics and relatively stable temperature characteristics under core loosening faults. Finally, the difference index is compared with a third preset threshold, and a judgment output including a status identifier and confidence level is generated based on the comparison result, forming a complete closed-loop logic from data acquisition to fault decision-making.

[0171] As a specific implementation method, the scheme of this application is implemented as follows: During the monitoring process, a sliding window of appropriate width is applied to the vibration spectrum characteristic parameter and the temperature characteristic parameter respectively, and the first derivative is calculated using the numerical differentiation method to obtain the time series change rate. The absolute difference between the vibration change rate and the temperature change rate is calculated and divided by the absolute value of the vibration change rate to obtain the difference index. This index is compared with a threshold determined through offline calibration. If the difference index is greater than the threshold, it is determined to be a loose core state, and three confidence levels (high, medium, and low) are assigned according to the degree to which it exceeds the threshold.

[0172] Through the above scheme, this application effectively improves the reliability of core loosening judgment in strong electromagnetic interference environment, avoids misjudgment caused by unstable change rate calculation and inaccurate definition of difference index, and ensures the accuracy and robustness of fault results.

[0173] In some of the embodiments described above in this application, it is proposed to generate insulation aging assessment results based on the distribution pattern of temperature characteristic parameters of multiple nodes of reactor. However, in its implementation, the calculation method of gradient distribution characteristics is not clearly defined, and the generation of assessment results lacks a specific threshold comparison mechanism, resulting in inaccurate identification of insulation aging status and insufficient reliability.

[0174] To address this, this application further proposes a method for generating insulation aging assessment results based on the distribution pattern of multi-node temperature characteristic parameters of a reactor. This includes determining the standard deviation and gradient distribution characteristics of the multi-node temperature characteristic parameters. The gradient distribution characteristics are obtained by determining the differences between the temperature characteristic parameters of adjacent nodes and calculating the average of these differences. Based on the comparison results between the standard deviation and a fourth preset threshold, and the comparison results between the gradient distribution characteristics and a fifth preset threshold, an insulation aging assessment result containing an insulation aging status identifier and a confidence level is generated.

[0175] Among them, standard deviation refers to the statistical quantification index of the dispersion of temperature characteristic parameters. It can be calculated mathematically using the square root of variance, aiming to objectively characterize the uneven temperature distribution caused by local overheating due to insulation aging. Gradient distribution characteristics refer to the comprehensive characterization parameter of the spatial variation trend of temperature characteristic parameters of adjacent nodes. It can be calculated by the arithmetic mean of the absolute values ​​of temperature differences between adjacent nodes, aiming to eliminate the interference of instantaneous fluctuations of individual abnormal nodes and accurately reflect the gradual change law of temperature in the reactor space. The preset threshold can be understood as a benchmark reference value used to judge the degree of abnormality in temperature distribution. It can be set based on historical operating data of the equipment or industry experience values, aiming to transform the fuzzy distribution pattern into a quantifiable objective judgment basis and avoid subjective arbitrariness in the evaluation process.

[0176] Specifically, the proposed solution first determines the standard deviation of multi-node temperature characteristic parameters to capture the overall discrete trend of temperature distribution, while simultaneously calculating gradient distribution characteristics to quantify the uniformity of spatial temperature changes. Then, the standard deviation is compared with a fourth preset threshold to determine whether the temperature distribution exceeds the normal range, and the gradient distribution characteristics are compared with a fifth preset threshold to detect the degree of anomaly in spatial temperature changes. Finally, the two comparison results are combined to generate an evaluation result including a status identifier and confidence level, forming a multi-dimensional quantitative evaluation mechanism for insulation aging status, thereby achieving accurate identification and early warning of insulation aging status.

[0177] As a specific implementation method, the solution of this application is implemented as follows: A temperature sensor network is arranged at the ends of the reactor windings, the iron core yoke, and the heat dissipation duct area, with each sensor node collecting temperature characteristic parameters. After receiving the temperature data from multiple nodes, the processing unit calculates the standard deviation of the temperature characteristic parameters of all nodes and sequentially calculates the average value of the temperature difference between adjacent nodes to obtain the gradient distribution characteristics. The calculation results are compared with preset thresholds. When the standard deviation exceeds a fourth preset threshold and the gradient distribution characteristics exceed a fifth preset threshold, a "severe insulation aging" status label is generated and assigned a high confidence level. When only the standard deviation exceeds the threshold, a "local overheating" label is generated and assigned a medium confidence level. The processing unit can specifically use an embedded microcontroller to implement the data processing function, and the sensor network is connected to the processing unit through a wireless communication module.

[0178] Through the above technical solution, this application solves the problems of unclear calculation methods and lack of evaluation mechanisms for gradient distribution characteristics, realizes objective quantitative evaluation of insulation aging status, avoids reliance on subjective experience, improves the accuracy and reliability of results, and effectively supports the accurate identification and early warning of reactor insulation status.

[0179] In some of the embodiments described above in this application, the results are fed back to dynamic frequency conversion scanning, and the allocation of the pre-diagnostic frequency band and the scanning sequence of the target node cluster are dynamically adjusted based on the fault type. However, in its implementation, the urgency and spatial diffusion characteristics of the fault are not quantitatively assessed and differentiated, resulting in a lack of targeted allocation of monitoring resources. Specifically, high-urgency faults (such as rapidly developing inter-turn short circuits) may fail to capture key data in a timely manner due to insufficient resources, while faults with high diffusion trends (such as spreading insulation aging) are difficult to effectively track propagation paths due to their fixed topology, leading to an increased risk of signal loss and decreased fault diagnosis accuracy in strong electromagnetic environments, which cannot meet the dynamic monitoring needs of reactors in multiple fault scenarios.

[0180] To address this, this application further proposes feeding the results back to dynamic frequency conversion scanning, dynamically adjusting the allocation of pre-diagnostic frequency bands and the scanning sequence of the target node cluster based on the fault type. See [link to relevant documentation]. Figure 6 Taking the server as the executing entity as an example, the following steps are included.

[0181] 601. Based on the results of inter-turn short circuit identification, core loosening judgment, and insulation aging assessment, determine the fault time urgency and spatial diffusion trend. The fault time urgency is determined according to the fault development rate, and the spatial diffusion trend is determined according to the degree of propagation of fault characteristics in adjacent nodes.

[0182] 602. Based on the time urgency and spatial spread trend of the fault, allocate corresponding monitoring resources to the pre-diagnostic frequency band. Among them, faults with high time urgency are allocated exclusive frequency resources, and faults with high spatial spread trend are allocated continuous frequency band resources.

[0183] 603. Based on the time urgency and spatial spread trend of the fault, reconstruct the scanning topology of the target node cluster. Faults with high time urgency are monitored using a star topology, while faults with high spatial spread trend are tracked and monitored using a mesh topology.

[0184] 604. Based on the reconstructed scanning topology and allocated monitoring resources, update the frequency point mapping relationship of the pre-diagnostic frequency band and the scanning time series of the target node cluster.

[0185] Among these, fault time urgency refers to a quantitative indicator characterizing the urgency of fault development. It can be determined based on the time series change rate of temperature characteristic parameters, vibration spectrum characteristic parameters, or magnetic field harmonic characteristic parameters. Its purpose is to distinguish between high-risk faults that require immediate handling and low-risk faults that can be dealt with later. Spatial diffusion trend refers to a quantitative indicator characterizing the possibility of fault propagation in the spatial dimension of the reactor. It can be achieved by analyzing the correlation coefficient or gradient distribution characteristics of fault characteristic parameters between adjacent nodes. Its purpose is to identify fault types with propagation characteristics. Exclusive frequency resources refer to dedicated discrete frequency points allocated to specific fault areas. This can be achieved by using a single frequency point for exclusive use or a combination of a few discrete frequency points. Its purpose is to avoid signal conflicts caused by electromagnetic interference and ensure the stable transmission of critical fault data. Continuous frequency band resources refer to the continuous frequency range allocated to the fault area. This can be achieved by using a bandwidth-extended frequency band allocation method. Its purpose is to support high-density sampling and facilitate the capture of the continuous spatial changes of the fault. A star topology is a communication structure with the fault location as the central node and surrounding nodes as subordinate nodes. It can be implemented using a communication protocol with centralized scheduling by the central node, aiming to reduce communication latency and enhance real-time monitoring of emergency faults. A mesh topology is a communication structure where nodes establish peer-to-peer connections. It can be implemented using a distributed routing protocol, aiming to dynamically expand the monitoring range and effectively cover the propagation paths of spreading faults.

[0186] Specifically, the proposed solution achieves dynamic optimization of monitoring resources by feeding the results back to the dynamic frequency conversion scanning system. First, based on the generated inter-turn short-circuit identification results, core loosening judgment results, and insulation aging assessment results, the system quantifies the fault's temporal urgency and spatial propagation trend. This directly utilizes multi-dimensional diagnostic data, avoiding additional data acquisition delays. Subsequently, based on the quantified fault characteristics, the system allocates monitoring resources differentially: exclusive frequency resources are allocated to faults with high temporal urgency to ensure the reliability of critical data transmission, while continuous frequency band resources are allocated to faults with high spatial propagation trends to support continuous tracking of fault propagation paths. Simultaneously, the system reconstructs the scanning topology of the target node cluster, employing a star topology for high temporal urgency faults to achieve focused monitoring, and a mesh topology for high spatial propagation trend faults to achieve tracking monitoring, thereby optimizing communication paths. Finally, based on the reconstructed topology and allocated resources, the system updates the frequency mapping relationship and scanning time series to ensure that the resource optimization strategy takes effect immediately. This fault-characteristic-driven dynamic adjustment mechanism shifts the allocation of monitoring resources from static preset to adaptive optimization, significantly improving the targeting and robustness of reactor multi-fault monitoring in strong electromagnetic environments.

[0187] As a specific embodiment, the solution of this application is implemented as follows: When the system detects an inter-turn short circuit at the end of the reactor winding with a confidence level higher than the threshold, it first calculates the rate of change of temperature characteristic parameters and magnetic field harmonic characteristic parameters to determine that the fault urgency is high. Simultaneously, it analyzes the correlation coefficients of temperature characteristic parameters of adjacent nodes to determine that the spatial diffusion trend is low. Based on this, the system allocates three discrete dedicated frequency points from the pre-diagnostic frequency band as exclusive frequency point resources to the fault area and constructs a star topology centered on the fault location, with the fault location as the central node and surrounding nodes as subordinate nodes. The central node is allocated the first set of dedicated frequency points with a shorter scan time interval, while the subordinate nodes are allocated the second set of dedicated frequency points with a longer scan time interval. In this way, the system can focus on monitoring high-urgency faults, ensuring stable acquisition of critical data and rapid response.

[0188] Through the above-described scheme, this application achieves dynamic optimization of monitoring resources driven by fault characteristics, effectively solving the problem of critical data loss due to insufficient resources in high-urgency faults, and the monitoring blind spot problem caused by fixed topology in high-propagation-trend faults. In environments with strong electromagnetic interference, this scheme significantly improves the transmission reliability of critical fault data and enhances the ability to trace fault propagation paths, thereby improving the accuracy and timeliness of results in reactor multi-fault scenarios.

[0189] In some of the embodiments described above in this application, the urgency of the fault time and the spatial propagation trend are proposed to dynamically adjust the allocation of the pre-diagnostic frequency band and the scanning sequence of the target node cluster. However, in its implementation, there is a lack of a precise quantitative method for the fault development rate and propagation index, which makes it impossible to objectively distinguish the evolution speed and spatial propagation characteristics of the fault, thereby affecting the pertinence and timeliness of subsequent monitoring resource allocation and scanning topology reconstruction.

[0190] In response, this application further proposes methods for determining the urgency and spatial propagation trend of faults based on inter-turn short-circuit identification results, core loosening judgment results, and insulation aging assessment results, including:

[0191] Based on the time series data from the inter-turn short circuit identification results, core loosening judgment results, and insulation aging assessment results, the rate of change of temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters is calculated, and the maximum rate of change is taken as the fault development rate.

[0192] Based on the spatial distribution of multi-node temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters of the reactor, the correlation coefficient and gradient value of fault characteristic parameters between adjacent nodes are determined, and the weighted sum of the average correlation coefficient and gradient value is used as the propagation index.

[0193] The failure development rate is compared with a preset urgency threshold to determine the failure time urgency. When the failure development rate exceeds the preset urgency threshold, it is defined as high time urgency.

[0194] The spatial diffusion trend is determined by comparing the propagation index with a preset diffusion threshold. When the propagation index exceeds the preset diffusion threshold, it is defined as a high spatial diffusion trend.

[0195] Among them, the fault development rate is a core indicator characterizing the dynamic evolution speed of a fault over time. It can be achieved by using the rate of change of the most significant parameter among temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters. Its purpose is to highlight the most active fault dimension characteristics and avoid weakening key fault signals due to multi-dimensional parameter averaging. The propagation index is a comprehensive indicator quantifying the breadth and intensity of fault propagation in the spatial dimension. It can be achieved by weighted fusion of the correlation coefficient and gradient values ​​of fault characteristic parameters between adjacent nodes. Its purpose is to balance the assessment weights of the fault's lateral spread and vertical deterioration, ensuring the comprehensiveness of the propagation characteristic analysis. The preset urgency threshold and preset diffusion threshold are judgment criteria used to convert continuously changing fault development rates and propagation indices into discrete levels. They can be achieved by using fixed thresholds determined based on statistical analysis of historical fault data or thresholds dynamically adjusted in conjunction with equipment operating conditions. Their purpose is to provide objective and operable classification criteria for fault urgency and diffusion.

[0196] Specifically, the proposed solution calculates the rate of change of temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters based on time-series data from inter-turn short-circuit identification results, core loosening judgment results, and insulation aging assessment results. The maximum value is selected as the fault development rate, thereby accurately capturing the activity level and evolution trend of the fault. Simultaneously, based on the spatial distribution of temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters of multiple reactor nodes, the correlation coefficients and gradient values ​​of fault characteristic parameters between adjacent nodes are determined, and these are weighted and fused into a propagation index, effectively revealing the spatial correlation and propagation path of the fault. Furthermore, the fault development rate is compared with a preset urgency threshold to determine the fault's temporal urgency, and the propagation index is compared with a preset diffusion threshold to determine the spatial diffusion trend, forming an objective quantitative mechanism for the fault's urgency and diffusion. This provides a precise decision-making basis for dynamically adjusting the allocation of pre-diagnostic frequency bands and the scanning sequence of the target node cluster.

[0197] As a preferred embodiment, the solution of this application is implemented as follows: In the reactor monitoring system, after acquiring time-series data of temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters of multiple sensor nodes, the system uses the sliding window method to calculate the first derivative of each parameter to obtain the rate of change, and selects the maximum value as the fault development rate. Simultaneously, based on the spatial relationship of the sensor topology, the system performs correlation analysis and gradient calculation on the fault characteristic parameters of adjacent nodes, and fuses the average correlation coefficient and gradient value into a propagation index using preset weights. Subsequently, the system compares the fault development rate with a preset urgency threshold; if it exceeds the threshold, it is defined as high temporal urgency. Similarly, the propagation index is compared with a preset diffusion threshold; if it exceeds the threshold, it is defined as high spatial diffusion trend. This process can be executed by an embedded signal processing unit, wherein the fault development rate calculation module uses an adaptive filtering algorithm to suppress noise interference, and the propagation index calculation module dynamically adjusts the weight allocation based on the node topology relationship.

[0198] Through the above scheme, this application achieves accurate quantification of the time urgency and spatial propagation trend of reactor faults, and can objectively distinguish the evolution speed and spatial propagation characteristics of faults, thereby improving the pertinence and timeliness of subsequent monitoring resource allocation and scanning topology reconstruction.

[0199] In some of the embodiments described above in this application, the allocation of pre-diagnostic frequency bands is dynamically adjusted based on the urgency of the fault time and the spatial diffusion trend. However, in its implementation, there is a lack of refined resource allocation strategies for different fault scenarios, resulting in the inability to adaptively configure monitoring resources according to fault characteristics. Specifically, when the urgency of the fault time and the spatial diffusion trend present different combinations, the existing solutions struggle to balance the real-time acquisition of key parameters with the coverage of diffusion area monitoring. This may cause high-urgency faults to be diagnosed late due to insufficient frequency resources, or high-diffusion-trend faults to fail to capture spatial evolution details due to bandwidth limitations, thereby affecting the accuracy of reactor fault diagnosis and resource utilization efficiency under strong electromagnetic interference environments.

[0200] In response, this application further proposes allocating corresponding monitoring resources to the pre-diagnostic frequency band based on the urgency of the fault time and the spatial diffusion trend, including:

[0201] When the urgency of the fault is high and the spatial diffusion trend is low, three discrete dedicated frequency points in the pre-diagnostic frequency band are allocated as exclusive frequency point resources for the corresponding fault area. The exclusive frequency point resources are used for key parameter acquisition.

[0202] When the spatial diffusion trend is high and the fault time urgency is low, continuous frequency band resources with a bandwidth of not less than five times the bandwidth of a single frequency point in the pre-diagnostic frequency band are allocated to the corresponding fault area. The continuous frequency band resources are used for auxiliary monitoring.

[0203] When both the time urgency of the fault and the spatial spread trend are high, exclusive frequency resources and continuous frequency band resources are allocated simultaneously for the corresponding fault area.

[0204] Among these, fault time urgency refers to a quantitative indicator of fault development rate, specifically determined by the maximum rate of change of temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters. Its purpose is to quickly identify fault states requiring urgent handling. Spatial diffusion trend can be understood as a quantitative indicator of the degree to which fault characteristics propagate between adjacent nodes. Specifically, it can be calculated using the weighted sum of the correlation coefficients and gradient values ​​of fault characteristic parameters of adjacent nodes. Its purpose is to assess the spatial expansion risk of the fault. Exclusive frequency resources refer to dedicated discrete frequency resources allocated to critical fault points. In practical applications, three non-interfering discrete frequency points can be used as a redundant configuration to ensure that real-time monitoring of high-urgency faults is not affected by electromagnetic interference. Continuous frequency band resources are specifically broadband frequency band resources allocated for diffuse faults. For example, a continuous frequency band with a bandwidth no less than five times the bandwidth of a single frequency point in the basic inspection frequency band can be used. Its purpose is to support parallel monitoring of multiple nodes to capture the spatial evolution details of the fault.

[0205] Specifically, the solution in this application achieves adaptive configuration of monitoring resources by establishing a precise mapping relationship between fault characteristics and resource allocation. When the system identifies a fault with high time urgency and low spatial diffusion trend, it prioritizes allocating three discrete dedicated frequency points as exclusive resources to ensure that the core fault point obtains interference-free real-time monitoring capabilities and avoids diagnostic interruptions due to frequency point failure. When the spatial diffusion trend is high and the fault time urgency is low, it allocates large-bandwidth continuous frequency band resources to support multi-node parallel communication to track spatial evolution. Its bandwidth design meets the high-density sampling requirements without excessively occupying the spectrum. When both the fault time urgency and spatial diffusion trend are high, exclusive frequency points and continuous frequency band resources are allocated simultaneously. This ensures that the core fault point obtains an exclusive frequency point to guarantee emergency response, while the diffusion area utilizes continuous frequency bands for spatial tracking. The two work together to prevent resource conflicts and form complementary monitoring, thereby effectively addressing the contradiction between limited spectrum resources and diverse fault modes in the reactor's electromagnetic environment.

[0206] As a specific implementation method, the scheme of this application is implemented as follows: When the system detects an inter-turn short circuit at the end node of the reactor winding, and the fault time urgency is determined to be high and the spatial diffusion trend is determined to be low, three discrete dedicated frequency points in the pre-diagnostic frequency band are allocated as exclusive frequency point resources for the fault area to continuously collect key parameters. When a core loosening fault is detected with a high spatial diffusion trend and a low fault time urgency, a continuous frequency band resource with a larger bandwidth is allocated to assist in monitoring the fault diffusion process. When an insulation aging fault simultaneously exhibits high time urgency and a high spatial diffusion trend, exclusive frequency point resources and continuous frequency band resources are allocated simultaneously for emergency diagnosis of the core area and tracking monitoring of the diffusion area, respectively.

[0207] Through the above scheme, this application realizes the dynamic adaptive configuration of monitoring resources according to the fault characteristics, effectively balancing the real-time acquisition of key parameters and the coverage of diffusion area monitoring, avoiding the problem of delayed diagnosis of high-urgency faults due to insufficient frequency resources, and the inability to capture spatial evolution details of high diffusion trend faults due to bandwidth limitations, thereby improving the accuracy of reactor fault diagnosis and the efficiency of spectrum resource utilization under strong electromagnetic interference environment.

[0208] In some of the embodiments described above in this application, a scanning topology for reconstructing the target node cluster is proposed to optimize the allocation of monitoring resources. However, in its implementation, the aforementioned scheme only generally specifies that faults with high time urgency are monitored using a star topology and faults with high spatial diffusion trends are tracked and monitored using a mesh topology. It does not consider different combinations of fault time urgency and spatial diffusion trends, which makes it impossible to dynamically adapt the optimal topology in actual fault scenarios, affecting the targeting and efficiency of fault monitoring.

[0209] In response, this application further proposes a method for reconstructing the scanning topology of the target node cluster based on the urgency of the failure time and the spatial diffusion trend, including:

[0210] When the urgency of the fault is high and the spatial diffusion trend is low, a star topology centered on the fault location is constructed, with the fault location as the central node and adjacent nodes as subordinate nodes.

[0211] When the spatial diffusion trend is high and the time urgency of the failure is low, a mesh topology covering the failure area is constructed, and all nodes within the failure area are established as peer-to-peer connections.

[0212] When both the time urgency of the fault and the spatial spread trend are high, a hybrid topology is constructed, using a star topology in the fault core area and a mesh topology in the fault spread area.

[0213] Based on the constructed topology, the scanning priority and communication path of each node in the target node cluster are determined.

[0214] Among them, fault time urgency refers to the severity of a fault determined by its development rate. This can be achieved by comparing the maximum rate of change of temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters with a preset urgency threshold, aiming to quickly identify fault states requiring immediate response. Spatial diffusion trend refers to a quantitative parameter characterizing the degree of fault feature propagation between adjacent nodes. This can be achieved by comparing the weighted sum of the correlation coefficients and gradient values ​​of fault feature parameters of adjacent nodes with a preset diffusion threshold, aiming to assess the spatial propagation risk of the fault. Star topology can be understood as a communication structure radiating from a single node to connect surrounding nodes. It can be implemented using a central node polling data acquisition mechanism, aiming to concentrate resources for high-frequency monitoring of key locations. Mesh topology can be understood as a communication structure establishing multi-hop peer-to-peer connections between nodes. It can be implemented using a distributed routing protocol, aiming to cover fault areas and track feature propagation paths. Hybrid topology refers to a composite communication architecture that combines star and mesh structures. It can be implemented using a regional topology partitioning strategy, aiming to balance the dual needs of emergency response and diffusion tracking. Scan priority can be understood as the order in which node data is collected. It can be implemented using a scheduling algorithm based on the comprehensive weight of nodes, with the aim of optimizing resource allocation efficiency. Communication path refers to the physical or logical channel for data transmission between nodes. It can be implemented using a dynamic routing table configuration mechanism, with the aim of ensuring the reliability and real-time performance of data transmission.

[0215] Specifically, the proposed solution dynamically constructs a scanning topology adapted to the fault characteristics by determining both the fault's time urgency and spatial propagation trend. When the system identifies a high fault time urgency and a low spatial propagation trend, it immediately constructs a star topology, designating the fault location as the central node and assigning it the highest scanning priority. This concentrates monitoring resources on the core fault point, avoiding response delays caused by resource dispersion. When the spatial propagation trend is high and the fault time urgency is low, a mesh topology is constructed, establishing peer-to-peer connections between nodes. A multi-hop communication mechanism comprehensively captures the spatial propagation path of the fault, suitable for slowly developing, spreading faults. When both are high, a hybrid topology strategy is adopted. A star topology is deployed in the core fault area to ensure rapid response capabilities, while a mesh topology is deployed in the spreading area to track propagation paths, forming a hierarchical monitoring system. Based on the constructed topology, the system further determines the scanning priority and communication path of each node, transforming the topology logic into specific timing scheduling and routing configurations. This ensures precise matching between monitoring resources and fault characteristics, thereby achieving dynamic adaptive adjustment of the topology in complex fault scenarios.

[0216] As a specific implementation method, the scheme of this application is implemented as follows: When the rate of change of temperature characteristic parameters in the winding end region of the reactor exceeds a preset urgency threshold, and the correlation coefficient of fault characteristic parameters between adjacent nodes is less than 0.3, the system determines it as a high time urgency, low spatial propagation trend fault. A star topology centered on the fault location is constructed, the winding end node is set as the central node and configured with the shortest scanning time interval, and adjacent nodes are subordinate nodes using a secondary scanning time interval. When the gradient value of vibration spectrum characteristic parameters in the core region exceeds a preset propagation threshold, but the rate of change of temperature characteristic parameters does not reach the urgency threshold, the system determines it as a high spatial propagation trend, low time urgency fault. A mesh topology covering the core region is constructed, enabling all nodes in the region to establish peer-to-peer connections, and a distributed routing protocol is used to continuously track the fault characteristic propagation path. When both a sudden change in temperature characteristic parameters and an excessive gradient value of adjacent nodes occur simultaneously at the winding end, the system determines it as a double-high fault. A star topology is constructed in the core region of the winding end to ensure real-time monitoring, and a mesh topology is constructed in the surrounding propagation region to track the propagation boundary. A topology partitioning strategy is used to achieve gradient configuration of monitoring resources.

[0217] Through the above technical solutions, the system can dynamically reconstruct the scanning topology based on the combined characteristics of the urgency of the fault time and the spatial diffusion trend. In time-sensitive faults, it can quickly concentrate monitoring resources on core nodes, establish a comprehensive tracking network in spatially diffused faults, and achieve collaborative optimization of core response and boundary tracking in complex faults. This significantly improves the targeting of fault monitoring and the efficiency of resource utilization, and effectively solves the problem of insufficient dynamic adaptation of topology in actual fault scenarios.

[0218] In some of the embodiments described above in this application, a method is proposed to dynamically adjust diagnostic resources by updating the frequency mapping relationship and scanning time series based on the reconstructed scanning topology and allocated monitoring resources. However, in its implementation, there is a lack of specific mechanisms for frequency allocation and time parameter settings under different topologies, resulting in insufficiently refined resource allocation and inability to achieve optimal monitoring scheduling based on fault urgency and propagation trend. For example, in a star topology, the frequency requirements of the central node and subordinate nodes are not distinguished, or in a mesh topology, continuous frequency band resources are not adapted, resulting in insufficient monitoring density in high-urgency fault areas or waste of resources in low-propagation trend areas.

[0219] In response, this application further proposes updating the frequency point mapping relationship of the pre-diagnostic frequency band and the scanning time series of the target node cluster based on the reconstructed scanning topology and allocated monitoring resources, including:

[0220] Based on the role types of nodes in the reconstructed scanning topology, the first set of dedicated frequency points is allocated to the central node of the star topology, and the second set of dedicated frequency points is allocated to the subordinate nodes of the star topology. Continuous frequency band resources are allocated to peer nodes in the mesh topology.

[0221] Based on the type of monitored resources and the structure of the scanning topology, the scanning time parameters for each node in the target node cluster are determined. In a star topology, the central node has a first scanning time interval, and subordinate nodes have a second scanning time interval, with the first scanning time interval being shorter than the second. In a mesh topology, peer nodes have a third scanning time interval.

[0222] Based on the allocated dedicated frequency points and the determined scan time parameters, the frequency point mapping relationship of the pre-diagnostic frequency band and the scan time series of the target node cluster are updated to obtain a new frequency point mapping table and a new scan time series.

[0223] The reconstructed scanning topology refers to the node communication structure dynamically adjusted according to fault characteristics. It can be implemented using star, mesh, or hybrid topologies, aiming to optimize the allocation of monitoring resources in fault areas. Node role type refers to the functional role a node plays in a specific topology, such as a central node or subordinate node in a star topology, or a peer node in a mesh topology. Its purpose is to differentiate resource allocation based on node importance. The first set of dedicated frequency points refers to frequency resources specifically for the central node of the star topology. These can be implemented using discrete frequency points or small-bandwidth frequency bands, aiming to ensure the communication reliability of core monitoring nodes. The second set of dedicated frequency points refers to frequency resources specifically for subordinate nodes of the star topology. These can be implemented using discrete frequency points different from the first set, aiming to achieve resource isolation and avoid interference. Continuous frequency band resources refer to the continuous frequency bandwidth allocated to the mesh topology. These can be implemented using wideband frequency bands, aiming to support multi-node collaborative communication. The first scanning time interval refers to the shorter scanning interval for the central node of the star topology. It can be set to a shorter time interval than that for subordinate nodes, aiming to quickly respond to faults with high time urgency. The second scan interval refers to the longer scan interval for slave nodes in a star topology. It can be set to a longer interval than that for the central node, with the aim of reducing energy consumption while ensuring coverage. The third scan interval refers to the scan interval for peer nodes in a mesh topology. It can be set to a specific value to adapt to spatial diffusion trends, with the aim of tracking fault propagation.

[0224] Specifically, the proposed solution first allocates frequency resources based on the role type of nodes in the reconstructed scanning topology. A first set of dedicated frequencies is allocated to the central node in a star topology to ensure the communication reliability of the core monitoring node. A second set of dedicated frequencies is allocated to subordinate nodes to achieve resource isolation and avoid interference. Continuous frequency band resources are allocated to peer nodes in a mesh topology to support multi-node collaborative communication. Then, scanning time parameters are determined based on the type of monitoring resources and the structure of the scanning topology. A shorter first scanning interval is set for the central node to quickly respond to high-urgency faults. A longer second scanning interval is set for subordinate nodes to reduce energy consumption while ensuring coverage. A third scanning interval is set for peer nodes in a mesh topology to adapt to the tracking requirements of spatial diffusion trends. Finally, the frequency mapping relationship and scanning time sequence are updated based on the allocated dedicated frequencies and determined scanning time parameters. This transforms the association between topology roles and resource characteristics into an executable frequency mapping table and scanning time sequence, thereby achieving refined and dynamic optimization of diagnostic resource scheduling. This ensures sufficient monitoring density in high-urgency fault areas while avoiding resource waste in low-diffusion-trend areas.

[0225] As a preferred embodiment, the solution of this application is implemented as follows: When a high-time-urgency fault is detected, the scanning topology is reconstructed into a star topology, with the central node using a high-performance microcontroller as the processing unit and allocated a first set of dedicated frequency points. Slave nodes use low-power wireless communication modules and are allocated a second set of dedicated frequency points. The scanning time parameter is set to a shorter scanning interval for the central node and a longer scanning interval for the slave nodes. When a fault with a high spatial diffusion trend is detected, the topology is reconstructed into a mesh topology, with peer nodes using wideband transceivers and allocated continuous frequency band resources. The scanning time interval is set to a moderate value to track fault diffusion. When both the fault's time urgency and spatial diffusion trend are high, a hybrid topology is constructed, using a star topology in the fault core area and a mesh topology in the diffusion area, with corresponding frequency point resources and scanning time parameters configured respectively.

[0226] Through the above scheme, this application achieves refined resource scheduling for different fault characteristics, avoiding the problems of insufficient monitoring density in high-urgency fault areas and resource waste in low-diffusion-trend areas, and improving the real-time response capability and resource adaptation accuracy of fault monitoring.

[0227] All of the above-mentioned optional technical solutions can be combined in any way to form the optional embodiments of this application, and will not be described in detail here.

[0228] Figure 7 This is a schematic diagram of a wireless temperature measurement device for a reactor based on frequency conversion technology, provided in an embodiment of this application. See also... Figure 7 The device includes:

[0229] The execution module 701 is used to perform dynamic frequency conversion scanning. Based on the electromagnetic environment of the reactor and the sensor topology, the working frequency band is divided into a basic inspection frequency band and a pre-diagnosis frequency band. In the basic inspection frequency band, the reactor sensor nodes are cyclically frequency-converted to obtain monitoring data. When abnormal spectral characteristics are identified, high-density sampling is triggered in the pre-diagnosis frequency band to obtain a composite frequency signal modulated by temperature, vibration and magnetic field parameters.

[0230] The decoupling module 702 is used to perform feature decoupling on the composite frequency signal selected based on the abnormal spectral features to obtain temperature feature parameters, vibration spectrum feature parameters and magnetic field harmonic feature parameters.

[0231] The generation module 703 is used to generate inter-turn short circuit identification results, core loosening judgment results, and insulation aging assessment results based on the temperature characteristic parameter, the vibration spectrum characteristic parameter, and the magnetic field harmonic characteristic parameter.

[0232] The adjustment module 704 is used to feed the result back to the dynamic frequency conversion scan, and dynamically adjust the allocation of the pre-diagnostic frequency band and the scan sequence of the target node cluster based on the fault type.

[0233] It should be noted that the wireless temperature measurement device for reactors based on frequency conversion technology provided in the above embodiments is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the wireless temperature measurement device for reactors based on frequency conversion technology provided in the above embodiments and the wireless temperature measurement method for reactors based on frequency conversion technology belong to the same concept. The specific implementation process is detailed in the method embodiments and will not be repeated here.

[0234] The above are merely optional embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A wireless temperature measurement method for reactors based on frequency conversion technology, characterized in that, The method includes: An electromagnetic environment map of the reactor's surroundings is generated by performing a full-band spectrum scan of the operating frequency band and recording the background noise level and interference source distribution characteristics at each frequency point. Based on the electromagnetic environment map and sensor topology, the operating frequency band is divided into a basic inspection frequency band and a pre-diagnosis frequency band. The basic inspection frequency band selects a continuous frequency band with optimal electromagnetic environment, while the pre-diagnosis frequency band allocates discrete diagnostic-specific frequency points according to the number of sensor nodes. Based on the spatial position and electrical connection relationships of each node in the sensor topology, corresponding scanning priorities and frequency point allocation strategies are configured for the basic inspection frequency band and the pre-diagnosis frequency band, respectively. In the basic inspection frequency band, cyclic frequency conversion addressing is performed on the reactor sensor node to obtain monitoring data. When abnormal spectral characteristics are identified, high-density sampling is triggered on the target node cluster in the pre-diagnosis frequency band to obtain a composite frequency signal modulated by temperature, vibration and magnetic field parameters. The composite frequency signal selected based on the abnormal spectral features is subjected to feature decoupling to obtain temperature feature parameters, vibration spectral feature parameters, and magnetic field harmonic feature parameters; Based on the temperature characteristic parameters, the vibration spectrum characteristic parameters, and the magnetic field harmonic characteristic parameters, the inter-turn short circuit identification result, the core loosening judgment result, and the insulation aging assessment result are generated. The diagnostic results are fed back to dynamic frequency conversion scanning. Based on the inter-turn short circuit identification results, the core loosening judgment results, and the insulation aging assessment results, the fault time urgency and spatial diffusion trend are determined. The fault time urgency is determined according to the fault development rate, and the spatial diffusion trend is determined according to the degree of propagation of fault characteristics in adjacent nodes. Based on the fault time urgency and the spatial diffusion trend, corresponding monitoring resources are allocated to the pre-diagnostic frequency band. Faults with high time urgency are allocated exclusive frequency resources, and faults with high spatial diffusion trends are allocated continuous frequency band resources. Based on the fault time urgency and the spatial diffusion trend, the scanning topology of the target node cluster is reconstructed. Faults with high time urgency are monitored using a star topology, and faults with high spatial diffusion trends are monitored using a mesh topology. Based on the reconstructed scanning topology and the allocated monitoring resources, the frequency point mapping relationship of the pre-diagnostic frequency band and the scanning time series of the target node cluster are updated.

2. The method according to claim 1, characterized in that, Based on the spatial and electrical relationships of the nodes in the sensor topology, the method configures corresponding scanning priorities and frequency allocation strategies for the basic inspection frequency band and the pre-diagnostic frequency band, including: Based on the spatial positional relationship of each node in the sensor topology, the regional importance weight is determined. Among them, the regional importance weight of the nodes located at the end of the reactor winding, the iron core yoke and the heat dissipation duct area is greater than that of the nodes located in the reactor shell and the middle. Based on the electrical connection relationship of each node in the sensor topology, the electrical correlation weight is determined, wherein the electrical correlation weight of the node directly connected to the winding input terminal, undertaking the task of inter-turn voltage monitoring, and located on the main magnetic flux path is greater than that of the node used only for environmental monitoring. Based on the regional importance weight and the electrical correlation weight, the comprehensive scanning priority of each node is determined, and the nodes with the highest comprehensive scanning priority are allocated frequency points with lower background noise levels in the pre-diagnostic frequency band. A scan sequence lookup table is generated based on the comprehensive scan priority, and frequency point allocation mapping tables are established for the basic inspection frequency band and the pre-diagnosis frequency band, respectively, to complete the configuration of the scan priority and frequency point allocation strategy.

3. The method according to claim 1, characterized in that, Based on the spatial and electrical relationships of the nodes in the sensor topology, the method configures corresponding scanning priorities and frequency allocation strategies for the basic inspection frequency band and the pre-diagnostic frequency band, including: Based on the spatial positional relationship of each node in the sensor topology, the regional importance weight is determined. Among them, the regional importance weight of the nodes located at the end of the reactor winding, the iron core yoke and the heat dissipation duct area is greater than that of the nodes located in the reactor shell and the middle. Based on the electrical connection relationship of each node in the sensor topology, the electrical correlation weight is determined, wherein the electrical correlation weight of the node directly connected to the winding input terminal, undertaking the task of inter-turn voltage monitoring, and located on the main magnetic flux path is greater than that of the node used only for environmental monitoring. Based on the regional importance weight and the electrical correlation weight, the comprehensive scanning priority of each node is determined, and the nodes with the highest comprehensive scanning priority are allocated frequency points with lower background noise levels in the pre-diagnostic frequency band. A scan sequence lookup table is generated based on the comprehensive scan priority, and frequency point allocation mapping tables are established for the basic inspection frequency band and the pre-diagnosis frequency band, respectively, to complete the configuration of the scan priority and frequency point allocation strategy.

4. The method according to claim 1, characterized in that, The process of performing feature decoupling on the composite frequency signal selected based on the abnormal spectral features to obtain temperature feature parameters, vibration spectral feature parameters, and magnetic field harmonic feature parameters includes: The composite frequency signal is decomposed into three signal components: temperature gradient component, vibration modulation component, and magnetic field harmonic component. The temperature gradient component, the vibration modulation component, and the magnetic field harmonic component are purified by using an adaptive Wiener filter bank to obtain purified temperature signal, purified vibration signal, and purified magnetic field signal with suppressed cross-interference. Time-domain feature extraction is performed on the purified temperature signal, the purified vibration signal, and the purified magnetic field signal to obtain the temperature feature parameter, the vibration spectrum feature parameter, and the magnetic field harmonic feature parameter.

5. The method according to claim 4, characterized in that, The step of using an adaptive Wiener filter bank to perform frequency band purification on the temperature gradient component, the vibration modulation component, and the magnetic field harmonic component to obtain purified temperature signal, purified vibration signal, and purified magnetic field signal with suppressed cross-interference includes: The dominant frequency bands of the temperature gradient component, the vibration modulation component, and the magnetic field harmonic component are determined, wherein the dominant frequency band of the temperature gradient component is the lowest frequency band, the dominant frequency band of the vibration modulation component is the mid-frequency band, and the dominant frequency band of the magnetic field harmonic component is the discrete frequency point where the power frequency and its harmonics are located. Based on the dominant frequency bands of the temperature gradient component, the vibration modulation component, and the magnetic field harmonic component, the cross-correlation function between each pair of the three signal components is calculated, and a covariance matrix characterizing the interference intensity between components is constructed based on the peak value of the cross-correlation function. With minimizing the off-diagonal elements in the covariance matrix as the optimization objective, the coefficients of the Wiener filter bank are iteratively updated using an adaptive algorithm. Inverse filtering is then applied to the temperature gradient component, the vibration modulation component, and the magnetic field harmonic component to obtain the purified temperature signal, the purified vibration signal, and the purified magnetic field signal with suppressed cross-interference.

6. The method according to claim 1, characterized in that, The generation of inter-turn short circuit identification results, core loosening judgment results, and insulation aging assessment results based on the temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters includes: Based on the correlation characteristics of the vibration spectrum characteristic parameters and the magnetic field harmonic characteristic parameters, the inter-turn short circuit identification result is generated. The correlation characteristics include the correlation coefficient and covariance trend between the vibration spectrum characteristic parameters and the magnetic field harmonic characteristic parameters. Based on the difference between the vibration spectrum characteristic parameter and the temperature characteristic parameter, the core loosening determination result is generated. The difference includes the degree of difference between the rate of change of the vibration spectrum characteristic parameter and the rate of change of the temperature characteristic parameter. The insulation aging assessment results are generated based on the distribution pattern of the temperature characteristic parameters of the reactor at multiple nodes. The distribution pattern includes the standard deviation and gradient distribution characteristics of the temperature characteristic parameters of each node.

7. The method according to claim 6, characterized in that, The generation of the inter-turn short-circuit identification result based on the correlation features of the vibration spectrum characteristic parameters and the magnetic field harmonic characteristic parameters includes: Determine the correlation coefficient between the amplitude of the vibration spectrum characteristic parameter and the amplitudes of the third and fifth harmonics in the magnetic field harmonic characteristic parameter; Determine the covariant relationship in the time domain between the changing trends of the vibration spectrum characteristic parameters and the changing trends of the magnetic field harmonic characteristic parameters; Based on the comparison results of the correlation coefficient and the first preset threshold, and the comparison results of the covariance relationship and the second preset threshold, the inter-turn short circuit identification result containing the inter-turn short circuit status identifier and confidence level is generated.

8. The method according to claim 1, characterized in that, The determination of fault urgency and spatial spread trend based on the inter-turn short circuit identification results, the core loosening determination results, and the insulation aging assessment results includes: Based on the time series data in the inter-turn short circuit identification results, the core loosening judgment results, and the insulation aging assessment results, the rate of change of the temperature characteristic parameter, the vibration spectrum characteristic parameter, and the magnetic field harmonic characteristic parameter is calculated, and the maximum rate of change is taken as the fault development rate. Based on the spatial distribution of the multi-node temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters of the reactor, the correlation coefficient and gradient value of the fault characteristic parameters between adjacent nodes are determined, and the weighted sum of the average correlation coefficient and gradient value is used as the propagation index. The fault development rate is compared with a preset urgency threshold to determine the fault time urgency, wherein a fault development rate exceeding the preset urgency threshold is defined as high time urgency. The propagation index is compared with a preset diffusion threshold to determine the spatial diffusion trend, wherein a propagation index exceeding the preset diffusion threshold is defined as a high spatial diffusion trend.

9. The method according to claim 1, characterized in that, The process of reconstructing the scanning topology of the target node cluster based on the fault time urgency and the spatial diffusion trend includes: When the time urgency of the fault is high and the spatial diffusion trend is low, a star topology centered on the fault location is constructed, with the fault location as the central node and adjacent nodes as subordinate nodes. When the spatial diffusion trend is high and the fault time urgency is low, a mesh topology covering the fault area is constructed, and all nodes in the fault area are established as peer-to-peer connections. When both the fault urgency and the spatial spread trend are high, a hybrid topology is constructed, using a star topology in the fault core region and a mesh topology in the fault spread region. Based on the constructed topology, the scanning priority and communication path of each node in the target node cluster are determined.

10. A wireless temperature measurement device for a reactor based on frequency conversion technology, characterized in that, The device includes: The execution module generates an electromagnetic environment map of the reactor's surroundings. It performs a full-band spectrum scan of the operating frequency band and records the background noise level and interference source distribution characteristics at each frequency point. Based on the electromagnetic environment map and sensor topology, the operating frequency band is divided into a basic inspection band and a pre-diagnosis band. The basic inspection band selects a continuous frequency band with optimal electromagnetic environment, while the pre-diagnosis band allocates discrete diagnostic frequency points based on the number of sensor nodes. Based on the spatial and electrical relationships of the nodes in the sensor topology, corresponding scanning priorities and frequency point allocation strategies are configured for the basic inspection band and the pre-diagnosis band, respectively. In the basic inspection band, cyclic frequency conversion addressing is performed on the reactor sensor nodes to acquire monitoring data. When abnormal spectral characteristics are identified, high-density sampling is triggered on the target node cluster in the pre-diagnosis band to acquire a composite frequency signal modulated by temperature, vibration, and magnetic field parameters. The decoupling module is used to perform feature decoupling on the composite frequency signal selected based on abnormal spectral features to obtain temperature feature parameters, vibration spectral feature parameters, and magnetic field harmonic feature parameters. The generation module is used to generate inter-turn short circuit identification results, core loosening judgment results, and insulation aging assessment results based on temperature characteristic parameters, vibration spectrum characteristic parameters, and magnetic field harmonic characteristic parameters. An adjustment module is used to feed back diagnostic results to dynamic frequency conversion scanning. Based on the inter-turn short circuit identification results, the core loosening judgment results, and the insulation aging assessment results, it determines the fault time urgency and spatial diffusion trend. The fault time urgency is determined according to the fault development rate, and the spatial diffusion trend is determined according to the degree of propagation of fault characteristics in adjacent nodes. Based on the fault time urgency and the spatial diffusion trend, corresponding monitoring resources are allocated to the pre-diagnostic frequency band. Faults with high time urgency are allocated exclusive frequency resources, and faults with high spatial diffusion trends are allocated continuous frequency band resources. Based on the fault time urgency and the spatial diffusion trend, the scanning topology of the target node cluster is reconstructed. Faults with high time urgency are monitored using a star topology, and faults with high spatial diffusion trends are monitored using a mesh topology. Based on the reconstructed scanning topology and the allocated monitoring resources, the frequency point mapping relationship of the pre-diagnostic frequency band and the scanning time sequence of the target node cluster are updated.

Images