Reactor turn-to-turn short circuit monitoring method fusing magnetic field difference method and loss factor method

Abstract

The invention relates to the technical field of electromagnetic data processing, in particular to a reactor turn-to-turn short circuit monitoring method fusing a magnetic field difference method and a loss factor method. According to the method, a plurality of event pairs are constructed by adopting an event correlation method, fundamental decoupling of fault features of thermal response and normal working condition fluctuation is realized by comparing differences between event response features and standard event response features corresponding to working condition data, and dynamic thermal response deviation is obtained. Two-dimensional diagnosis features are mapped into a parameter space by using a parameter space method, so that the position of a mapping point represents the dynamic trend of a fault, tracking is carried out based on the position, a future state can be determined in advance based on the trend before a fault result occurs, and the problem of monitoring time lag is solved. And an effective fusion diagnosis result is obtained for reference of workers.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electromagnetic data processing, in particular to a kind of reactance interturn short circuit monitoring method of fusion magnetic field difference method and loss factor method. BACKGROUND

[0002] Dry-type air-core reactor is the key equipment to ensure power quality and system stability in power system, and its main failure form is winding interturn short circuit.The existing online monitoring technology mainly includes magnetic field difference method and indirect monitoring method based on temperature measurement.

[0003] Magnetic field difference method monitors the symmetry of leakage magnetic field spatial distribution through magnetic field sensor arranged outside the reactor.When interturn short circuit occurs, the short-circuit current formed by fault point will destroy the original distribution of leakage magnetic field.However, the measurement of this method is easily affected by the large fluctuation of load current under normal operating conditions of reactor, so that the weak magnetic field distortion signal generated by early interturn short circuit fault is easily submerged by the magnetic field asymmetry effect generated by normal operating condition change, resulting in low signal-to-noise ratio of monitoring and the possibility of false alarm.

[0004] Loss factor method is an indirect monitoring method based on temperature measurement, and its basis is that interturn short circuit will produce additional joule heat loss, resulting in increased overall energy loss and abnormally high operating temperature of equipment.However, as a large-capacity and large-volume device, the reactor has huge thermal inertia, and its thermal response process has significant delay of minutes to hours.This delay of response makes it impossible to provide timely early warning of fault.

[0005] In order to more effectively monitor the short circuit of reactor, it is necessary to effectively integrate the two monitoring methods, effectively play the advantages of each method and avoid the disadvantages.However, due to the fact that the temperature response of temperature measurement method is much slower than the instantaneous response of electromagnetic field under frequent load change conditions, there is inherent mismatch in time scale between the two measurement means;And the thermal response will be affected by normal operating conditions, so that effective thermal response characteristics cannot be distinguished, and effective integration results cannot be obtained for short circuit monitoring. SUMMARY

[0006] In order to solve the technical problem that the thermal response of temperature measurement monitoring method is affected by normal operating conditions and has time delay with electromagnetic field response, and thus the two monitoring methods cannot be effectively integrated, the purpose of the present application is to provide a kind of reactance interturn short circuit monitoring method of fusion magnetic field difference method and loss factor method, and the technical scheme adopted is as follows: The present application proposes a kind of reactance interturn short circuit monitoring method of fusion magnetic field difference method and loss factor method, the method comprises: Obtain the magnetic field asymmetry between differential magnetic field intensity and main magnetic field intensity; In the thermal power signal, multiple load step events are identified by the signal change rate; the temperature rise acceleration signal of the winding is obtained, and multiple temperature rise acceleration events are identified in the temperature rise acceleration signal using a peak detection algorithm; the load step events and temperature rise acceleration events are correlated and matched according to the time relationship to obtain multiple event pairs; For each event pair, the corresponding operating condition data and event response characteristics are obtained, including the event time difference and the event change rate difference; the difference between the event response characteristics and the standard event response characteristics corresponding to the operating condition data is compared to obtain the dynamic thermal response deviation. For real-time event pairs, the magnetic field asymmetry and the dynamic thermal response deviation within the preset detection time window are used to construct two-dimensional diagnostic features; the short-circuit fault state is determined based on the position of the two-dimensional diagnostic features in the parameter space.

[0007] Furthermore, the method for obtaining the magnetic field asymmetry includes: The ratio between the differential magnetic field strength and the main magnetic field strength is taken as the magnetic field asymmetry.

[0008] Furthermore, the method for identifying the load step event includes: The first derivative of the thermal power signal is subjected to positive peak detection, and the time corresponding to each peak point and the peak value are taken as the load step event.

[0009] Furthermore, the association matching method includes: For each load step event, the temperature rise acceleration event with the largest peak value within the preset search time window at the time of occurrence is selected as the matching temperature rise acceleration event. If there is no temperature rise acceleration event within the search window, the corresponding composite step event is filtered out.

[0010] Furthermore, the operating condition data includes the average ambient temperature within the time range corresponding to the event pair, and the average main circuit current within the preset operating condition window before the event pair occurs.

[0011] Furthermore, the event change rate difference is the ratio of the peak value corresponding to the temperature rise acceleration event to the peak value corresponding to the load step event.

[0012] Furthermore, the method for obtaining the standard event response features corresponding to the operating condition data includes: A health event time difference data model and a health event change rate difference model are constructed based on the operating condition data of health event pairs of health reactors and their corresponding health event response characteristics, respectively. The input data of the health event time difference data model is the operating condition data, and the output data is the standard event time difference. The input data of the health event change rate difference model is the operating condition data, and the output data is the standard event change rate difference. The operating condition data of the event pairs are input into the health event time difference data model and the health event change rate difference model, respectively, to obtain the standard event response characteristics.

[0013] Furthermore, the method for obtaining the dynamic thermal response deviation includes: The Mahalanobis distance between the event response characteristics and the standard event response characteristics corresponding to the operating condition data is used as the dynamic thermal response deviation.

[0014] Furthermore, the method for obtaining the two-dimensional diagnostic features includes: Before the start time of the real-time event pair, a magnetic field asymmetry sequence within a preset monitoring time window is obtained. The average value of the elements in the magnetic field asymmetry sequence is used as the structural offset feature in the two-dimensional diagnostic features. The dynamic thermal response deviation sequence generated by all event pairs within the preset monitoring time window is statistically analyzed. The dynamic thermal response deviation sequence is processed using the exponential weighted moving average method, and the cumulative value obtained is used as the dynamic instability feature in the two-dimensional diagnostic features.

[0015] Further, determining the short-circuit fault state based on the position of the two-dimensional diagnostic features in the parameter space includes: The two-dimensional diagnostic features are mapped to a parameter space with structural offset features as the horizontal axis and dynamic instability features as the vertical axis. The circuit breaker fault state is determined based on the distance between the mapped point of the two-dimensional diagnostic features and the origin, as well as the angle between the vector from the origin to the mapped point and the horizontal axis.

[0016] The present invention has the following beneficial effects: To eliminate the influence of thermal response under normal operating conditions on short-circuit monitoring, this invention employs an event correlation method to construct multiple event pairs. This ensures that during temperature monitoring, inter-turn short-circuit faults are no longer considered an abnormal amplitude in the signal, but rather identified as an event deviating from their inherent healthy response pattern. Furthermore, by comparing the differences between the event response characteristics and the standard event response characteristics corresponding to the operating condition data, a fundamental decoupling of the fault characteristics of the thermal response from fluctuations under normal operating conditions is achieved, yielding dynamic thermal response deviation. The magnetic field asymmetry obtained by the fusion magnetic field difference method reflects the structural consequences of short-circuit faults, while the dynamic thermal response deviation obtained by the temperature monitoring method reflects the inducing effect of the loss factor generated by short-circuit faults on winding function. Therefore, this invention utilizes the parameter space method to map two-dimensional diagnostic features into a parameter space, allowing the location of the mapping point to characterize the dynamic trend of the fault. Tracking based on location enables the prediction of future states based on trends before the fault occurs, solving the problem of monitoring time lag and thus obtaining effective fusion diagnostic results for staff reference. Attached Figure Description

[0017] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 The flowchart illustrates a reactor inter-turn short-circuit monitoring method that integrates the magnetic field difference method and the loss factor method, as provided in one embodiment of the present invention. Detailed Implementation

[0019] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a reactor inter-turn short-circuit monitoring method integrating the magnetic field differential method and the loss factor method proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] The following description, in conjunction with the accompanying drawings, details a specific scheme for a reactor inter-turn short-circuit monitoring method that integrates the magnetic field difference method and the loss factor method, provided by this invention.

[0022] Please see Figure 1 The diagram illustrates a flowchart of a reactor inter-turn short-circuit monitoring method integrating the magnetic field differential method and the loss factor method, according to an embodiment of the present invention. The method includes: Step S1: Obtain the magnetic field asymmetry between the differential magnetic field strength and the main magnetic field strength.

[0023] Inter-turn short-circuit faults create an abnormal short-circuit circulating current within the reactor windings. The additional magnetic field generated by this circulating current disrupts the symmetry of the original spatial distribution of the reactor's leakage magnetic field. Direct observation of the differential magnetic field strength, used to characterize this asymmetry, is severely affected by changes in the main load current, as the overall increase in load proportionally amplifies the differential magnetic field, making it difficult to distinguish from the distortion caused by the fault. To eliminate this combined effect and thus isolate the magnetic field asymmetry purely caused by changes in the winding's internal structure, this invention compares the differential magnetic field strength with its corresponding main magnetic field strength, which reflects the total ampere-turns. This provides a highly relevant magnetic field asymmetry, which can be considered a dimensionless feature. Its long-term, static shift directly reflects permanent geometric changes that may occur within the reactor, such as insulation aging and carbonization or winding deformation caused by electrodynamic forces. This feature quantifies the static, permanent structural deformation of the reactor windings caused by the fault, reflecting the structural consequences of the fault.

[0024] It should be noted that, since the embodiments of the present invention mainly target the feature data in the loss factor method, the features obtained by the magnetic field difference method can be directly extracted without further analysis.

[0025] This invention utilizes Hall sensor arrays or induction coils and other magnetic field sensors to measure the magnetic field strength of the main magnetic field and the differential magnetic field. The sampling frequency is set to 1 kHz, and the acquired signals can undergo preprocessing such as filtering and noise reduction to ensure signal quality. Specific methods are well-known to those skilled in the art and will not be elaborated upon here.

[0026] Preferably, in this embodiment of the invention, the ratio between the differential magnetic field strength and the main magnetic field strength is used as the magnetic field asymmetry. That is, the greater the differential magnetic field strength relative to the main magnetic field strength, the stronger the magnetic field asymmetry and the more obvious the fault characteristics.

[0027] It should be noted that the obtained magnetic field asymmetry is a dimensionless value. To facilitate subsequent diagnostic analysis, the value can be further normalized to limit its range to between 0 and 1. The normalization method in this embodiment can be range standardization, a technique well-known to those skilled in the art, and will not be elaborated upon here.

[0028] Step S2: In the thermal power signal, identify multiple load step events by the signal change rate; obtain the temperature rise acceleration signal of the winding, and identify multiple temperature rise acceleration events in the temperature rise acceleration signal using the peak detection algorithm; associate and match the load step events and temperature rise acceleration events according to the time relationship to obtain multiple event pairs.

[0029] This invention takes into account that the thermal response characteristics of a healthy reactor are themselves a complex function that dynamically changes with the operation. Therefore, in order to effectively quantify the fault thermal response, this invention abandons direct separation at the signal level. Instead, it extracts features at the pattern level through event correlation and compares them with healthy features. This makes the inter-turn short circuit fault no longer regarded as an abnormal amplitude on the signal, but identified as an event that deviates from its inherent healthy response pattern, thus achieving a fundamental decoupling between fault characteristics and normal operating condition fluctuations.

[0030] Since the rapid and significant adjustment of the reactor load is the fundamental reason for the change in its thermal state, this reason will inevitably lead to a delayed response peak in the rate of temperature rise of its windings. Therefore, in order to accurately capture these paired causal relationships, this embodiment of the invention identifies load step events and temperature rise acceleration events, and then performs correlation matching to construct matching relationships for multiple event pairs. This enables subsequent steps to transform continuous data signals into a series of discrete dynamic thermal response event records that encapsulate complete dynamic response information.

[0031] In this embodiment of the invention, the thermal power signal is the square of the reactor main circuit current signal. The reactor main circuit current signal can be acquired by a current sensor, and the acquisition frequency can be set to 1kHz. The load step event is the "cause" of the thermal response, reflecting the preceding event causing a thermal response from the perspective of electrical data. Therefore, multiple load step events on the thermal power signal can be identified based on the signal change rate. It should be noted that the obtained load step event can include two types of data: the time of the event and the peak value of the corresponding rate of change on the thermal power signal.

[0032] Preferably, in this embodiment of the invention, the method for identifying load step events includes: Forward peak detection is performed on the first derivative of the heat power signal, and the time corresponding to each peak point and the peak value are taken as the load step event. It should be noted that the peak detection algorithm is a well-known technique to those skilled in the art. In the forward peak detection process, a detection threshold can be set. This detection threshold can be set according to the rated capacity of the reactor and the common load fluctuation range. In this embodiment of the invention, it is set to 5% of the rated heat power change rate. Only peaks that are greater than the detection threshold are considered as detected peak points.

[0033] In this embodiment of the invention, to address temperature-based monitoring in the loss factor method, temperature sensors, such as fiber optic grating temperature sensors, are installed at key parts of the winding to collect winding temperature signals. Furthermore, other temperature sensors are installed in the environment where the equipment is located to collect ambient temperature signals. The acquisition frequency of both temperature signals is set to 1Hz. That is, the temperature rise in the temperature rise acceleration signal is the temperature rise of the winding relative to the ambient temperature. The temperature rise signal is obtained by subtracting the winding temperature signal from the ambient temperature signal. Considering that the reactor, as a large heat capacity device, has significant inertia and time delay in response to changes in composite current, directly analyzing the first derivative of the temperature rise signal is insufficient to capture weak instantaneous dynamic anomalies caused by faults. Therefore, in this embodiment of the invention, the temperature rise acceleration signal is the second derivative of the temperature rise signal, which can more sensitively reflect the real-time imbalance between heat generation and heat dissipation power. The temperature rise acceleration signal is an effective signal for capturing the inflection point and peak value of the system's dynamic response. Therefore, peak detection is performed on it, and the identified positive peak point is the temperature rise acceleration event. It should be noted that the obtained temperature rise acceleration event can include two types of data: the time of the event and the peak value of the event on the temperature rise acceleration signal.

[0034] By further associating and matching the identified load step events and temperature rise acceleration events according to their time relationship, multiple event pairs can be obtained.

[0035] It should be noted that, since the essential purpose of this embodiment of the invention is to determine the fault status after monitoring real-time event pairs, a monitoring window can be set in terms of timing. In this embodiment of the invention, the time length of the monitoring window is set to 10 minutes, that is, the number of event pairs within the monitoring window range is obtained every ten minutes, and the most recent event pair is regarded as a real-time event pair.

[0036] Preferably, in this embodiment of the invention, since the rapid and significant adjustment of the reactor load is the fundamental reason for the change in its thermal state, and this reason inevitably leads to a delayed response peak in the winding temperature rise rate, i.e., the temperature rise acceleration event should be the result of a load step event, the correlation matching method includes: For each load step event, the temperature rise acceleration event with the largest peak value within a preset search time window at the time of occurrence is selected as the matching temperature rise acceleration event. If no temperature rise acceleration event exists within the search window, the corresponding composite step event is filtered out. In this embodiment of the invention, the time range of the search window is set to... ,in Let be the time when the load step event occurs in the i-th event pair.

[0037] It should be noted that, for the peak characteristics corresponding to each event, in order to facilitate comparison in subsequent steps, they can be normalized to eliminate the influence of units and range. The specific method will not be elaborated here.

[0038] Step S3: For each event pair, obtain the corresponding operating condition data and event response characteristics, including the event time difference and the event change rate difference; compare the difference between the event response characteristics and the standard event response characteristics corresponding to the operating condition data to obtain the dynamic thermal response deviation.

[0039] For each event pair, this invention uses the time difference and rate of change between the two events as event response features, characterizing the response delay and response difference between the two events. Further, the operating condition data corresponding to the event pair is determined. This operating condition data represents the actual operating conditions of the reactor turns at the corresponding time. Based on this operating condition data, the standard event response features corresponding to a healthy reactor can be found. Since an inter-turn short-circuit fault will form an abnormal energy dissipation loop inside the winding, whose thermal capacity and thermal resistance characteristics differ from those of the main winding, and the thermal behavior of this loop is not governed by the health law described by the aforementioned benchmark thermal response model, comparing it with the obtained event response features yields the dynamic thermal response deviation after removing the influence of normal operating conditions.

[0040] Preferably, in this embodiment of the invention, the operating condition data includes the average ambient temperature within the time range corresponding to the event pair, and the average main circuit current within a preset operating condition window before the event pair occurs. The time range is from the start of the load step event to the start of the temperature rise acceleration event. The preset operating condition window length is set to 60 seconds, meaning the time range corresponding to the preset operating condition window before the event pair occurs is... The average main loop current characterizes the initial thermal state of the system before it undergoes a dynamic response.

[0041] Preferably, in this embodiment of the invention, the difference in event change rate is the ratio of the peak value corresponding to the temperature rise acceleration event to the peak value corresponding to the load step event.

[0042] Preferably, in this embodiment of the invention, since the heat dissipation efficiency (mainly natural air convection) of a dry-type air-core reactor is nonlinearly affected by ambient temperature and the temperature rise of the winding itself, the standard event response characteristics exhibited by the dynamic thermal response of a healthy reactor are not fixed values, but rather change smoothly and continuously with the changes in time-domain operating conditions. Therefore, this embodiment of the invention employs a method of constructing a classification model, by learning from historical health data, to construct a benchmark thermal response model that can accurately describe this multivariate, nonlinear functional relationship. That is, the method for obtaining the standard event response characteristics corresponding to the operating condition data includes: A health event time difference data model and a health event rate of change difference model are constructed based on the operating condition data of health event pairs of health reactors and their corresponding health event response characteristics. The input data for the health event time difference data model is the operating condition data, and the output data is the standard event time difference. The input data for the health event rate of change difference model is the operating condition data, and the output data is the standard event rate of change difference. It should be noted that both models in this embodiment of the invention use the Gaussian process regression method for model construction, and a radial basis function (RBF) kernel function with automatic correlation determination (ARD) is selected. This is a well-known configuration of GPR, which can adapt to different scales and importance of input variables.

[0043] The operating condition data of the event pairs are input into the health event time difference data model and the health event change rate difference model, respectively, to obtain the standard event response characteristics.

[0044] Preferably, in this embodiment of the invention, the method for obtaining the dynamic thermal response deviation includes: The Mahalanobis distance between the event response characteristics and the standard event response characteristics corresponding to the operating condition data is used as the dynamic thermal response deviation. The method for obtaining the Mahalanobis distance is well-known to those skilled in the art, and it represents the standard deviation multiple by which the dynamic behavior of the current event pair deviates from its inherent health law.

[0045] It should be noted that the healthy reactor data defined in this embodiment refers to the initial period of reactor commissioning, specifically the first three months. During this phase, the system is in pure data acquisition mode, and the equipment is assumed to be healthy. The collected data is stored for training the aforementioned model. After the initial model is built, an update process can be introduced during the monitoring phase. For each newly generated event pair, if the dynamic thermal response deviation is lower than a preset confidence threshold, it can be determined to be in a healthy state, and the corresponding data is a valid healthy sample and stored in the database. The system will periodically retrain the healthy event time difference data model and the healthy event change rate difference model, either after the database has been expanded to a certain extent, to achieve fine-tuning and iteration of the baseline model, enabling it to adapt to the slow characteristic drift caused by normal aging of the equipment. In this embodiment, the confidence threshold is set to 2, the update cycle can be set to 24 hours, and the database update condition can be set to after 100 new samples are added.

[0046] Step S4: For real-time event pairs, construct two-dimensional diagnostic features from the magnetic field asymmetry and dynamic thermal response deviation within the preset detection time window; determine the short-circuit fault state based on the position of the two-dimensional diagnostic features in the parameter space.

[0047] After feature extraction through the above steps, dynamic thermal response deviations and magnetic field asymmetry can be obtained for real-time event pairs, eliminating the influence of normal operating condition fluctuations. The magnetic field asymmetry obtained by the fusion magnetic field difference method reflects the structural consequences of short-circuit faults, while the dynamic thermal response deviation obtained by the temperature monitoring method reflects the influence of short-circuit fault loss factors on winding function. Therefore, this embodiment of the invention utilizes the parameter space method to map two-dimensional diagnostic features into a parameter space. The two coordinate axes in the parameter space correspond to two independent diagnostic features, allowing the physical process of fault evolution to be presented and tracked along a clear trajectory within this space. The location of the mapped point characterizes the dynamic trend of the fault. Tracking based on location allows for the determination of future states based on trends before the fault occurs, solving the problem of monitoring time lag.

[0048] Preferably, in this embodiment of the invention, considering that the real-time acquired magnetic field asymmetry is high-frequency data, and that a single data point may be affected by short-term noise, and that dynamic thermal response deviation occurs in the form of discrete events, in order to construct a diagnostic space for the long-term health status of a stable reaction device, this embodiment of the invention performs temporal smoothing and trend accumulation processing on the two features in time series. That is, the method for acquiring two-dimensional diagnostic features includes: Before the start of the real-time event pair, a magnetic field asymmetry sequence within a preset monitoring time window is obtained. The average value of the elements in the magnetic field asymmetry sequence is used as the structural offset feature in the two-dimensional diagnostic features. By smoothing the average value, short-term noise and glitches caused by instantaneous operating condition fluctuations or electromagnetic interference can be filtered out, and the slowly accumulated static baseline offset caused by permanent structural changes such as winding deformation or insulation carbonization can be extracted.

[0049] Since dynamic thermal response deviations are generated as discrete events, to quantify the frequency and severity of "dynamic behavior mismatch" occurring in the equipment in the near future, this embodiment of the invention statistically analyzes the dynamic thermal response deviation sequence generated by all event pairs within a preset monitoring time window, processes the dynamic thermal response deviation sequence using an exponentially weighted moving average method, and uses the obtained cumulative value as the dynamic instability feature in the two-dimensional diagnostic features. That is, the exponentially weighted moving average algorithm can give higher weight to recently occurring events, which better meets the needs of assessing the current dynamic stability of the equipment. In this embodiment of the invention, the smoothing coefficient in the exponentially weighted moving average algorithm is set to 0.2. The specific algorithm is a well-known technique and will not be described in detail here.

[0050] In this embodiment of the invention, the monitoring time window is 24 hours, using a longer time window to eliminate the influence of short-term noise.

[0051] Preferably, in this embodiment of the invention, considering that the evolution of inter-turn short circuits has clear physical stages: early faults may only manifest as weak energy conversion anomalies without causing measurable macroscopic structural deformation; as the fault develops, continuous local overheating leads to insulation carbonization and winding deformation, ultimately causing significant distortion of the static magnetic field. Therefore, the two-dimensional diagnostic features are mapped to a parameter space with structural offset features as the horizontal axis and dynamic instability features as the vertical axis. The region near the origin of the parameter space is the healthy region. When the mapped point deviates from the healthy region, its distance and direction of movement contain rich diagnostic information. For example, if the mapped point mainly moves along the vertical axis, it indicates that the dynamic instability features have increased significantly, while the structural offset features have not changed much, indicating that the fault is still in its initial nascent stage. Its physical meaning is: the fault has begun to trigger abnormal dynamic thermal response, but has not yet caused measurable permanent damage to the macroscopic magnetic field structure of the reactor; if the mapped point moves to the upper right in space, it indicates that both the vertical and horizontal axes have increased significantly, indicating that the fault has entered the development stage. The physical meaning is that the fault has continued to develop, not only causing the dynamic behavior to completely deviate from the healthy pattern, but also causing permanent physical damage such as insulation carbonization or winding deformation. Therefore, this embodiment of the invention can further determine the open circuit fault state based on the distance between the two-dimensional diagnostic feature mapping point and the origin, and the angle between the vector from the origin to the mapping point and the horizontal axis. That is, the larger the distance, the farther the reactor deviates from the healthy center, and the more serious the fault; when the angle is close to 90 degrees, it indicates that the fault is mainly characterized by abnormal dynamic behavior and is in a very early stage, indicating that attention and monitoring should be strengthened; when the angle is close to 45 degrees, it indicates that static structural damage has begun to appear and is comparable to the severity of dynamic abnormality, the fault has been confirmed to have entered the development stage, and a maintenance plan should be considered; the rate of change of the angle with the mapping point can be used to judge the evolution speed of the fault nature. It should be noted that the specific quantification and definition methods of the fault state can be set by those skilled in the art according to the reactor properties. By setting the threshold of the distance, the threshold of the angle, and the threshold of the rate of change of the angle, the corresponding fault state can be obtained. This embodiment of the invention does not elaborate or limit these methods.

[0052] In summary, this invention employs an event association method to construct multiple event pairs. By comparing the differences between the event response characteristics and the standard event response characteristics corresponding to the operating condition data, it achieves a fundamental decoupling between the fault characteristics of thermal response and normal operating condition fluctuations, thus obtaining dynamic thermal response deviation. Utilizing the parameter space method, two-dimensional diagnostic features are mapped into a parameter space, allowing the location of the mapped points to characterize the dynamic trend of the fault. Location-based tracking enables the prediction of future states based on trends before the fault occurs, solving the problem of monitoring time lag and obtaining effective fusion diagnostic results for staff reference.

[0053] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0054] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for monitoring inter-turn short circuits in reactors that integrates the magnetic field difference method and the loss factor method, characterized in that, The method includes: To obtain the magnetic field asymmetry between the differential magnetic field strength and the main magnetic field strength; In the thermal power signal, multiple load step events are identified by the signal change rate; the temperature rise acceleration signal of the winding is obtained, and multiple temperature rise acceleration events are identified in the temperature rise acceleration signal using a peak detection algorithm; the load step events and temperature rise acceleration events are correlated and matched according to the time relationship to obtain multiple event pairs; For each event pair, the corresponding operating condition data and event response characteristics are obtained, including the event time difference and the event change rate difference; the difference between the event response characteristics and the standard event response characteristics corresponding to the operating condition data is compared to obtain the dynamic thermal response deviation. For real-time event pairs, the magnetic field asymmetry and the dynamic thermal response deviation within the preset detection time window are used to construct two-dimensional diagnostic features; the short-circuit fault state is determined based on the position of the two-dimensional diagnostic features in the parameter space.

2. The method for monitoring inter-turn short circuits in reactors that integrates the magnetic field difference method and the loss factor method according to claim 1, characterized in that, The method for obtaining the magnetic field asymmetry includes: The ratio between the differential magnetic field strength and the main magnetic field strength is taken as the magnetic field asymmetry.

3. The method for monitoring inter-turn short circuits in reactors that integrates the magnetic field difference method and the loss factor method according to claim 1, characterized in that, The method for identifying the load step event includes: The first derivative of the thermal power signal is subjected to positive peak detection, and the time corresponding to each peak point and the peak value are taken as the load step event.

4. The method for monitoring inter-turn short circuits in reactors that integrates the magnetic field difference method and the loss factor method according to claim 1, characterized in that, The association matching method includes: For each load step event, the temperature rise acceleration event with the largest peak value within the preset search time window at the time of occurrence is selected as the matching temperature rise acceleration event. If there is no temperature rise acceleration event within the search window, the corresponding composite step event is filtered out.

5. The method for monitoring inter-turn short circuits in a reactor that integrates the magnetic field difference method and the loss factor method according to claim 1, characterized in that, The operating condition data includes the average ambient temperature within the time range corresponding to the event pair, and the average main circuit current within the preset operating condition window before the event pair occurs.

6. The method for monitoring inter-turn short circuits in a reactor that integrates the magnetic field difference method and the loss factor method according to claim 3, characterized in that, The event change rate difference is the ratio of the peak value corresponding to the temperature rise acceleration event to the peak value corresponding to the load step event.

7. The method for monitoring inter-turn short circuits in a reactor that integrates the magnetic field difference method and the loss factor method according to claim 1, characterized in that, The method for obtaining the standard event response characteristics corresponding to the operating condition data includes: A health event time difference data model and a health event change rate difference model are constructed based on the operating condition data of health event pairs of health reactors and their corresponding health event response characteristics, respectively. The input data of the health event time difference data model is the operating condition data, and the output data is the standard event time difference. The input data of the health event change rate difference model is the operating condition data, and the output data is the standard event change rate difference. The operating condition data of the event pairs are input into the health event time difference data model and the health event change rate difference model, respectively, to obtain the standard event response characteristics.

8. The method for monitoring inter-turn short circuits in a reactor that integrates the magnetic field difference method and the loss factor method according to claim 1, characterized in that, The method for obtaining the dynamic thermal response deviation includes: The Mahalanobis distance between the event response characteristics and the standard event response characteristics corresponding to the operating condition data is used as the dynamic thermal response deviation.

9. The method for monitoring inter-turn short circuits in a reactor that integrates the magnetic field difference method and the loss factor method according to claim 1, characterized in that, The method for obtaining the two-dimensional diagnostic features includes: Before the start time of the real-time event pair, a magnetic field asymmetry sequence within a preset monitoring time window is obtained. The average value of the elements in the magnetic field asymmetry sequence is used as the structural offset feature in the two-dimensional diagnostic features. The dynamic thermal response deviation sequence generated by all event pairs within the preset monitoring time window is statistically analyzed. The dynamic thermal response deviation sequence is processed using the exponential weighted moving average method, and the cumulative value obtained is used as the dynamic instability feature in the two-dimensional diagnostic features.

10. A method for monitoring inter-turn short circuits in a reactor that integrates the magnetic field difference method and the loss factor method according to claim 9, characterized in that, The determination of short-circuit fault status based on the position of the two-dimensional diagnostic features in the parameter space includes: The two-dimensional diagnostic features are mapped to a parameter space with structural offset features as the horizontal axis and dynamic instability features as the vertical axis. The circuit breaker fault state is determined based on the distance between the mapped point of the two-dimensional diagnostic features and the origin, as well as the angle between the vector from the origin to the mapped point and the horizontal axis.

Images