Transformer on-line monitoring and fault early warning system

Abstract

The invention discloses a transformer on-line monitoring and fault early warning system, and particularly relates to the technical field of power equipment monitoring, and the system comprises a signal collection module, a signal filtering module, a broadband focusing module, a signal complexity analysis module, a threshold self-tuning module, a fault early warning module and a fault positioning module. The signal acquisition module acquires operation data of a transformer in real time through a sensor network, the filtering module removes noise, and the broadband focusing module converts broadband signals into narrow-band signals, so that the resolution is improved; the signal complexity analysis module evaluates the signal complexity, and the threshold value self-tuning module dynamically adjusts the threshold value; the fault early warning module judges whether to trigger early warning based on the adjusted threshold value, and the fault positioning module determines the position of the partial discharge source through a cross positioning algorithm; according to the invention, the problem that the accuracy of early warning is affected due to the fact that a traditional fixed-threshold monitoring method is easy to generate false alarm or missing alarm can be solved, and operation and maintenance personnel can be helped to quickly determine the position of a fault source through cross positioning.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment monitoring, and more specifically, to a transformer online monitoring and fault early warning system. Background Art

[0002] With the rapid development of power systems, transformers are important core equipment in power grids, and their reliability and stability are crucial to the normal operation of power systems. From the perspective of accident analysis, many accidents are sudden accidents caused by the expansion of local faults. However, the current judgment of the operating status of equipment depends on preventive tests. Due to the long preventive test cycle, the current preventive test items are not sensitive to the development of insulation faults. Local defects that develop in a few days to a few weeks are often not discovered in time. Accidents caused by the expansion of such local defects in a few weeks or even less time account for a large proportion of transformer accident statistics. Implementing online monitoring of transformers has become an important part of ensuring the safe and reliable operation of transformers.

[0003] Partial discharge is one of the common insulation faults inside the transformer. The occurrence of partial discharge may lead to the degradation of the insulation material, which in turn causes more serious faults and even damages the transformer. In order to ensure the long-term stable operation of the transformer, the online monitoring and fault warning technology of the transformer has emerged. When identifying partial discharge, the traditional transformer monitoring system usually issues an early warning based on a fixed threshold. However, due to the complexity and diversity of the signal, the fixed threshold monitoring method is prone to false alarms or missed alarms, affecting the accuracy of the early warning. In addition, the existing monitoring system also has certain limitations in locating the fault source, and it is difficult to accurately and quickly lock the location of the partial discharge source. Therefore, it is particularly important to develop a transformer online monitoring and fault warning system that can dynamically adjust the threshold and has high-precision fault location function. Summary of the invention

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A transformer online monitoring and fault warning system, comprising a signal acquisition module, a signal filtering module, a broadband focusing module, a signal complexity analysis module, a threshold self-tuning module, a fault warning module, and a fault location module;

[0006] The signal acquisition module collects all the operating data of the transformer in real time through the sensor network;

[0007] The signal filtering module filters the signals collected by the sensor network to remove high-frequency noise and obtain a processed signal set;

[0008] The broadband focusing module converts the broadband signal in the processed signal set into a narrowband signal through the broadband focusing algorithm, enhances the spatial resolution of the signal, and provides high-quality data for subsequent fault location and signal analysis;

[0009] The signal complexity analysis module uses generalized entropy to analyze the complexity of the signal and evaluate the degree of chaos of signal changes;

[0010] The threshold self-tuning module first determines the amount of threshold adjustment based on the degree of chaos of the signal change, and then dynamically adjusts the threshold according to the amplitude of the signal change to adapt to different signal fluctuations;

[0011] The fault warning module determines whether to trigger the warning mechanism based on the threshold values ​​adjusted by the threshold self-tuning module;

[0012] When the early warning mechanism is triggered, the fault location module uses a cross-location algorithm to determine the two-dimensional or three-dimensional spatial position of the local discharge source based on the signal arrival time difference.

[0013] In a preferred embodiment, the signal filtering module is used to filter the signals collected by the sensor network, and the specific steps are as follows:

[0014] The original signal xi(t) collected by each sensor at time t is multiplied by the preset weighting factor wi(t), so that each sensor signal is processed by the filter and the filtered signal y(t) is calculated according to the following formula:

[0015] β is a preset nonlinear index, λ is a preset filter coefficient, and m is the number of sensors.

[0016] In a preferred embodiment, the broadband focusing algorithm performs weighted focusing on each frequency component of the signal by the following formula:

[0017] N is the total number of frequency components, f n is the frequency component, τ n (x) is the arrival time difference of the signal, which means the arrival time difference of each signal component between different sensors, α is the preset nonlinear control parameter, and w n is the preset proportional coefficient of the nth frequency component, f b (x, t) is the output result after the narrowband signal is processed by the broadband focusing algorithm, that is, the narrowband signal is converted from the broadband signal to the narrowband signal. x represents the coordinate of the signal in space, and t is the time variable, which represents the change of the signal in time.

[0018] In a preferred embodiment, the signal complexity analysis module uses generalized entropy to analyze the complexity of the signal, and evaluating the degree of confusion of the signal change refers to:

[0019] Obtain the signal distribution collected by the sensor network, and then calculate the complexity of the signal according to the generalized entropy formula. The specific formula is:

[0020] p i represents the probability distribution of the i-th state in the signal, q is the preset control parameter of the generalized entropy, u is the number of states, and v is the preset fractional-order adjustment parameter used to adjust the calculation sensitivity of the signal complexity. It is the generalized entropy value, which is used to measure the complexity and chaos of the signal.

[0021] In a preferred embodiment, the threshold self-tuning module first determines the amount of threshold adjustment based on the degree of confusion of the signal change, which refers to:

[0022] A detection signal type form is preset, which includes all signal types collected by the sensor network, and is calculated based on the average fluctuation amplitude and fault frequency of each signal type in historical data: F=AXi+BYi+C, Xi and Yi respectively represent the average fluctuation amplitude and fault frequency of the signal type, C is a constant, F is a sorting value, and descending sorting is performed according to the sorting value F to obtain a signal type disordered sorting form, and then the product of the generalized entropy value and the preset transformation proportional coefficient K is calculated, and the obtained product value is rounded up to obtain the number of threshold adjustments, and finally the corresponding number of signal types in the signal type disordered sorting form is selected according to the number of threshold adjustments as the threshold adjustment form.

[0023] In a preferred embodiment, the threshold value self-tuning module dynamically adjusts the threshold value by:

[0024] Get the initial thresholds for the signal types in the threshold adjustment form respectively, and then perform the following calculations:

[0025] represents the reference value of the i-th signal type, S i represents the signal value currently collected by the i-th signal type, δ represents the preset nonlinear control parameter, represents the initial threshold of the i-th signal type, Represents the dynamically adjusted threshold for the ith signal type.

[0026] In a preferred embodiment, the fault warning module determines whether to trigger the warning mechanism based on the threshold values ​​adjusted by the threshold self-tuning module, which means that if the threshold value adjusted by the threshold self-tuning module exceeds the threshold value in the subsequent monitoring process, the warning mechanism is triggered.

[0027] In a preferred embodiment, the fault location module uses a cross location algorithm, which means:

[0028] Combine sensors i and j into a positioning combination. For each positioning combination, obtain the signal arrival time t of sensors i and j. i and t j , and calculate the time difference using the following formula:

[0029] d ij =V s *log e (1+|t i -t j | γ );V s is the propagation speed of the signal, γ is the preset nonlinear adjustment parameter, d ij It represents the difference in distance that the signal propagates from the PD source to sensors i and j.

[0030] In a preferred embodiment, the distance difference d is used ij The two-dimensional or three-dimensional spatial position of the local discharge source corresponding to each positioning combination is solved, and then all the obtained coordinate solutions are averaged to obtain the actual coordinate position of the local discharge source.

[0031] Technical effects and advantages of the present invention:

[0032] The present invention introduces a dynamic threshold self-tuning module, which can perform adaptive threshold adjustment according to the signal changes collected in real time, thus overcoming the limitations of the traditional fixed threshold method in the face of complex signal fluctuations. The traditional system is prone to false alarms and missed alarms due to the setting of fixed thresholds. The system dynamically adjusts the threshold through signal complexity analysis and generalized entropy calculation, which can effectively cope with various complex signal fluctuations and greatly improve the accuracy and sensitivity of fault detection. The system can accurately identify partial discharge anomalies and other situations, provide accurate warnings for operation and maintenance personnel, and ensure the safe operation of the transformer.

[0033] The present invention uses a broadband focusing algorithm to process the signal, converting the broadband signal into a narrowband signal. This conversion greatly enhances the spatial resolution of the signal and reduces the distortion or interference of the signal during the processing. High-quality signal processing directly improves the accuracy of fault analysis and provides a reliable data basis for subsequent fault location and complexity analysis. High-precision signal processing enables the system to extract useful signals from noisy environments. Especially in the face of complex industrial environments or multi-signal interference, the algorithm can maintain the robustness and efficiency of the system.

[0034] The system uses a cross-positioning algorithm and the arrival time difference of signals collected by sensors to quickly calculate the spatial location of the local discharge source. Cross-positioning combines the time difference information of multiple sensors and obtains accurate two-dimensional or three-dimensional position coordinates through geometric calculations, shortening the fault diagnosis time. Compared with traditional methods, this system can perform high-precision spatial positioning at the moment of signal arrival, and further averages the results of multiple sensors to eliminate the interference of noise and error on fault location. Accurate positioning helps operation and maintenance personnel quickly determine the location of the fault source, reducing troubleshooting time and maintenance difficulty.

[0035] Through timely fault warning and precise positioning, the present invention can effectively avoid equipment damage caused by further expansion of faults and reduce significant losses caused by fault escalation. An accurate early warning mechanism can help operation and maintenance personnel discover potential fault hazards at an early stage, so as to take maintenance and repair measures in time to avoid equipment from being damaged on a larger scale due to the deterioration of faults. The number of equipment shutdowns for maintenance due to excessive wear or sudden failures is reduced, thereby reducing maintenance costs and extending the service life of the transformer.

[0036] By dynamically adjusting the signal threshold, the system can reduce false alarms caused by small signal fluctuations. This mechanism avoids unnecessary warning triggering and excessive maintenance operations, and improves the operating efficiency of the monitoring system. The warning module combines the analysis of signal complexity and historical data, allowing the system to issue targeted warnings, ensuring that each warning has a high degree of accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings;

[0038] Figure 1 The schematic diagram of a transformer online monitoring and fault warning system in the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] Embodiment 1: A transformer online monitoring and fault warning system, comprising a signal acquisition module, a signal filtering module, a broadband focusing module, a signal complexity analysis module, a threshold self-tuning module, a fault warning module, and a fault location module; the signal acquisition module, the signal filtering module, the broadband focusing module, the signal complexity analysis module, the threshold self-tuning module, the fault warning module, and the fault location module are communicatively connected with each other.

[0041] The signal acquisition module collects all the operating data of the transformer in real time through the sensor network; the signal acquisition module is the basic part of the system, responsible for collecting all the operating data from multiple sensor networks of the transformer, including electrical signals, vibration, temperature and humidity, etc. These data provide the original input for subsequent analysis, processing and fault location, ensuring that the system can monitor the status of the transformer in real time.

[0042] The signal filtering module filters the signals collected by the sensor network, removes high-frequency noise, and obtains the processed signal set; the signals collected by the sensor network may contain various noises, especially in complex industrial environments. The function of the signal filtering module is to remove high-frequency noise and other interference signals through filtering technology, ensuring that the processed signals are clearer and more accurate, laying the foundation for subsequent broadband focusing and signal analysis.

[0043] The broadband focusing module converts the broadband signal in the processed signal set into a narrowband signal through a broadband focusing algorithm, enhances the spatial resolution of the signal, and provides high-quality data for subsequent fault location and signal analysis; the signal after filtering is still a broadband signal, which makes it difficult to perform high-precision spatial analysis. The broadband focusing module converts the broadband signal into a narrowband signal through a broadband focusing algorithm, enhancing the spatial resolution of the signal. By improving the focus of the signal, the system can more accurately identify the characteristics of the partial discharge signal and provide high-quality data for fault location and subsequent analysis.

[0044] The signal complexity analysis module uses generalized entropy to analyze the complexity of the signal and evaluate the degree of chaos in the signal changes; the signal complexity analysis module uses the generalized entropy algorithm to evaluate the complexity and chaos of the signal. By calculating the complexity of the signal, the system can determine whether the signal contains potential abnormal or fault characteristics. The higher the signal complexity, the greater the chaos and uncertainty in the signal, which may indicate problems in the system. This module provides an important reference for the dynamic adjustment and early warning mechanism of the system.

[0045] The threshold self-tuning module first determines the amount of threshold adjustment based on the degree of chaos of the signal change, and then dynamically adjusts the threshold according to the amplitude of the signal change to adapt to different signal fluctuations; the threshold self-tuning module dynamically adjusts the key threshold in the system by analyzing the amplitude and complexity of the signal change. When the complexity and fluctuation amplitude of the signal are large, the system will adjust the threshold according to these changes, so that the system can flexibly adapt to different signal fluctuation environments. This module ensures that the detection sensitivity of the system can self-adjust according to the actual operating status to avoid frequent false alarms or missed alarms.

[0046] The fault warning module determines whether to trigger the warning mechanism based on the thresholds adjusted by the threshold self-tuning module; the fault warning module is the core warning mechanism of the system. Based on the threshold dynamically adjusted by the threshold self-tuning module, this module monitors in real time whether the signal collected by the sensor exceeds the threshold. Once the preset monitoring parameters corresponding to the signal type, such as the intensity or frequency of the signal, exceed the threshold, the system will trigger a warning to indicate that the transformer may have a fault. This module ensures that the system can issue an alarm in time before a fault occurs or in the early stage of a fault, reducing potential losses caused by the fault.

[0047] When the early warning mechanism is triggered, the fault location module uses a cross-location algorithm to determine the two-dimensional or three-dimensional spatial position of the partial discharge source based on the signal arrival time difference. After the system triggers the early warning mechanism, the fault location module uses a cross-location algorithm to calculate the position of the partial discharge source in two-dimensional or three-dimensional space based on the time difference of the signal reaching each sensor. Through the positioning algorithm, this module can accurately determine the spatial location of the fault, help maintenance personnel quickly find the source of the problem, perform targeted repairs and maintenance, shorten troubleshooting time, and improve operation and maintenance efficiency.

[0048] The signal filtering module is used to filter the signals collected by the sensor network. The specific steps are as follows:

[0049] The original signal xi(t) collected by each sensor at time t is multiplied by the preset weighting factor wi(t), so that each sensor signal is processed by the filter and the filtered signal y(t) is calculated according to the following formula: β is the preset nonlinear index, λ is the preset filter coefficient, and m is the number of sensors. The original signal is processed by the filter to remove high-frequency noise and other interference signals to obtain a relatively pure signal set y(t). By weighting and nonlinear adjustment of different sensor signals, the formula ensures flexibility and accuracy in the filtering process. y(t) is the filtered signal. This signal is the result of filter processing. Compared with the original signal, the noise is effectively suppressed and the signal quality is higher. wi(t) is the weighting factor. It is a parameter weighted according to the importance or contribution of each sensor signal, which determines the weight of each sensor signal in the final filtered signal. The weight can change dynamically over time. xi(t) represents the original signal collected by the i-th sensor at time t. It represents the signal value collected by the sensor without filtering, including the target signal and noise. is a nonlinear adjustment term used to perform nonlinear filtering on the signal. This term exponentially attenuates or amplifies the filtering result according to the signal strength. λ is the filter coefficient, which is used to control the attenuation amplitude of the signal. The larger the coefficient, the faster the signal decays, which helps to suppress strong noise signals. β is a nonlinear exponent that controls the degree of nonlinear attenuation of the signal. By adjusting the β value, the response of the filter to signals of different intensities can be changed. m is the number of sensors. It indicates the number of sensors involved in the filtering calculation in the system, which is usually determined by the scale of the sensor network.

[0050] The broadband focusing algorithm performs weighted focusing on each frequency component of the signal using the following formula:

[0051] N is the total number of frequency components, f n is the frequency component, τ n (x) is the arrival time difference of the signal, which means the arrival time difference of each signal component between different sensors, α is the preset nonlinear control parameter, and w n is the preset proportional coefficient of the nth frequency component, f b (x, t) is the output result of the narrowband signal after being processed by the broadband focusing algorithm, that is, the conversion from a broadband signal to a narrowband signal. x represents the coordinate of the signal in space, and t is the time variable, which represents the change of the signal in time. b(x, t) represents the narrowband signal after being processed by the broadband focusing algorithm, that is, the output result of converting the broadband signal into a narrowband signal. It is used for subsequent signal analysis and fault location, and represents the clarity and focus of the signal after processing. x represents the coordinates of the signal in space, which refers to the position of the signal source in two-dimensional or three-dimensional space. t is a time variable, which represents the change of the signal over time and is used to reflect the propagation of the signal on the time axis. N is the number of total frequency components, which indicates the number of all frequency components involved in the focusing process. The larger N is, the more frequency components the signal contains. f n is the nth frequency component, which represents the frequency component of the signal in the frequency domain. Broadband signals contain multiple frequency components, and different frequency components have different effects on the spatial resolution and focusing effect of the signal. n (x) is the arrival time difference of the signal, which indicates the time difference of the nth frequency component of the signal between different sensors. The time difference is used to calculate the relative propagation path and distance of the signal, thereby determining the location of the signal source. α is a preset nonlinear control parameter used to control the nonlinear response of the frequency component to the signal. It determines the nonlinear focusing effect of the signal under different frequency components. A larger α value will enhance the response to a specific frequency, while a smaller α value will make the response smoother. w n is the weighting factor of the nth frequency component, which indicates the weight of different frequency components in the focusing process. The weighting factor can be adjusted according to the importance of the frequency or the signal characteristics to optimize the focusing effect of the signal.

[0052] The core purpose of the broadband focusing algorithm is to convert broadband signals into narrowband signals. The focusing process greatly improves the spatial resolution of the signal, so that the position of the signal source can be more accurately identified and located, which is especially important in fault location scenarios. By weighting and nonlinearly adjusting each frequency component, the influence of noise from different frequencies can be effectively suppressed, thereby improving the quality of the signal and ensuring the accuracy of subsequent analysis. n By combining the (x) and frequency components, the system can more accurately calculate the signal propagation path and perform accurate positioning calculations based on these data. This is critical for scenarios that require high-precision positioning, such as local discharge source positioning in fault detection systems.

[0053] The signal complexity analysis module uses generalized entropy to analyze the complexity of the signal. The evaluation of the degree of chaos of the signal change refers to:

[0054] The system obtains partial discharge signals from the sensor network. The signal contains multiple states (for example, signal strength or frequency distribution at different time points), which constitute the probability distribution of the signal. The signal distribution collected by the sensor network is obtained, and then the complexity of the signal is calculated according to the generalized entropy formula. The specific formula is:

[0055] p i represents the probability distribution of the i-th state in the signal, q is the preset control parameter of the generalized entropy, u is the number of states, and v is the preset fractional-order adjustment parameter used to adjust the calculation sensitivity of the signal complexity. It is the generalized entropy value, which is used to measure the complexity and disorder of the signal. i is the probability distribution of the signal, which indicates the probability of occurrence of different states of the signal. This parameter is obtained by statistical analysis of the collected signal. For example, in a time series signal, the signal strength at each moment can be regarded as a state, and the frequency of occurrence of the signal strength can be used to calculate its probability distribution p i . q is the control parameter of generalized entropy. In classical entropy, when q=1, entropy represents the average uncertainty of the system. By adjusting the value of q, generalized entropy can flexibly respond to different signal complexities. When q is greater than 1, generalized entropy is more sensitive to high-probability events; when q is less than 1, entropy is more sensitive to low-probability events. When q=1, the formula degenerates into classical entropy. n is the number of signal states, which indicates the number of different states in the signal distribution. This parameter is determined by the complexity of the signal and the sampling time interval. The more states there are, the richer the details of the signal changes. ν is a fractional order adjustment parameter that controls the calculation method of generalized entropy. In some cases, the complexity of the signal has fractional dimensional characteristics (such as fractals or chaotic phenomena). By using the ν parameter, the system can capture the complex patterns in the signal more finely, especially for the analysis of nonlinear systems.

[0056] The generalized entropy formula is used to quantify the complexity of a signal and can help the system identify potential faults or anomalies in the signal. When the complexity of a signal increases, it means that the fluctuation and chaos of the signal are increasing, which may indicate that there is a problem with certain parameters or equipment in the system. By controlling the parameter q and the fractional-order adjustment parameter ν, the formula can adapt to different types of signals, including linear and nonlinear signals, and is particularly suitable for processing signals with complex dynamic characteristics. This flexibility allows the system to be more widely used in different monitoring scenarios. The introduction of the parameter ν increases the sensitivity of the analysis of complex signals, especially when processing signals with fractal characteristics or chaotic phenomena. The system can better capture subtle changes in the signal by adjusting ν.

[0057] The threshold self-tuning module first determines the amount of threshold adjustment based on the degree of chaos of the signal change, which refers to:

[0058] A detection signal type form is preset, which includes all signal types collected by the sensor network, and is calculated based on the average fluctuation amplitude and fault frequency of each signal type in historical data: F=AXi+BYi+C, Xi and Yi represent the average fluctuation amplitude and fault frequency of the signal type respectively, C is a constant, F is a sorting value, A and B are the preset influence coefficients of Xi and Yi respectively, and descending sorting is performed according to the sorting value F to obtain a signal type disorder sorting form, and then the product of the generalized entropy value and the preset transformation ratio coefficient K is calculated, and the obtained product value is rounded up to obtain the number of threshold adjustments, and finally the corresponding number of signal types in the signal type disorder sorting form is selected according to the number of threshold adjustments as the threshold adjustment form.

[0059] F represents the ranking value of the signal type. It is calculated based on the average fluctuation amplitude Xi and fault frequency Yi of the signal type and is used to measure the importance of the signal type. The larger the ranking value F of the signal type, the more likely the signal type is to cause a fault or anomaly in the system, and the threshold needs to be adjusted first. A and B: weighted coefficients of the signal fluctuation amplitude and fault frequency, respectively. These two coefficients are used to balance the relative importance of the fluctuation amplitude and fault frequency in calculating F. The larger the coefficient A, the greater the impact of the fluctuation amplitude on the ranking value; the larger the coefficient B, the greater the impact of the fault frequency on the ranking value. Xi represents the average fluctuation amplitude of signal type i. A large fluctuation amplitude may mean that the signal changes frequently during operation and is prone to abnormal or unstable states. Yi represents the fault frequency of signal type i. This parameter indicates the frequency of faults of this signal type in historical data. A high fault frequency means that this signal type is more likely to have problems. C is a constant, which represents the benchmark value when calculating the ranking value to ensure the comparability of the ranking results. K is the transformation proportional coefficient of generalized entropy. This coefficient is used to adjust the impact of signal complexity on the number of threshold adjustments. When the signal complexity increases, its impact can be amplified by multiplying by K, prompting the system to adjust more thresholds. The number of threshold adjustments is obtained by rounding up the product of the generalized entropy value and the proportionality coefficient K, and represents the number of signal types that require threshold adjustment.

[0060] Data-driven threshold adjustment: Calculating the ranking value F of each signal type by the fluctuation amplitude and fault frequency in the historical data can help the system dynamically identify those signal types that are more likely to cause faults during actual operation. Prioritizing the thresholds of these signals can better prevent potential faults. Through the weighting coefficients A and B, the system can flexibly adjust the impact of each signal type on the threshold adjustment according to the relative importance of different signal characteristics (fluctuation amplitude and fault frequency). This ensures that the system pays attention to both signal changes and the frequency of historical faults, thereby more comprehensively assessing signal risks. The ranking value F of the signal type is used to determine which signal types need to have their thresholds adjusted first. By sorting F, the system can allocate resources more effectively and prioritize those signals that are most likely to cause faults, rather than distributing them equally to all signal types. This method improves the efficiency of the system. Dynamically adjust the threshold according to complexity: By multiplying the generalized entropy value with the transformation scale coefficient K, the system can flexibly respond to changes in signal complexity. The higher the complexity, the more chaotic and unpredictable the signal is, so the system will choose to adjust more signal type thresholds to adapt to different operating conditions. The larger F is, the higher the average fluctuation amplitude Xi and fault frequency Yi of the signal type. Signal types with larger F values ​​are more likely to fail or be abnormal, so the threshold of this signal type needs to be adjusted first to reduce the risk of system failure. When the F value is large, it means that this signal type shows high fluctuations or high failure frequencies in historical data, and the system should focus on these signals to prevent potential failure hazards.

[0061] The threshold self-tuning module dynamically adjusts the threshold value, which means:

[0062] Get the initial thresholds for the signal types in the threshold adjustment form respectively, and then perform the following calculations:

[0063] represents the reference value of the ith signal type, Si represents the signal value currently collected by the ith signal type, δ represents the preset nonlinear control parameter, represents the initial threshold of the i-th signal type, Represents the dynamically adjusted threshold for the ith signal type. Indicates the new threshold value after dynamic adjustment of the i-th signal type. After dynamic adjustment, this threshold value will be used to determine whether the signal triggers an early warning. Represents the initial threshold of the ith signal type. That is, the warning threshold set before dynamic adjustment. Si represents the signal value currently collected by the ith signal type. This signal value is the data collected by the sensor in real time. Represents the reference value of the ith signal type. This value is usually set based on historical data or system standards and is used to measure the normal range of the signal. δ is a preset nonlinear control parameter that controls the amplitude and sensitivity of the threshold adjustment. A higher δ value makes the threshold more sensitive to signal changes, while a lower δ value makes the adjustment smoother. The hyperbolic tangent function is used to smooth the effect of signal changes on threshold adjustment. Its value is between -1 and 1, which can effectively avoid the extreme effects of excessive or small signal fluctuations on threshold adjustment.

[0064] By analyzing the difference between the currently acquired signal value and the reference value of the signal type, the system can dynamically adjust the threshold of each signal type. This ensures that the system's warning threshold can adapt to the fluctuation of the signal in real time, preventing false alarms or missed alarms caused by setting the fixed threshold too high or too low. The nonlinear adjustment method can respond more sensitively to large signal changes, while the adjustment of small changes is smoother. By using the hyperbolic tangent function tanh, the extreme impact of excessive or small signal fluctuations on the threshold adjustment can be suppressed to ensure the stability of the threshold adjustment. By adjusting the nonlinear control parameter δ, the system can flexibly respond to the fluctuation characteristics of different signal types. For example, a larger δ is suitable for frequently fluctuating signals and can quickly adjust the threshold; while a smaller δ is suitable for stable signals to avoid over-adjusting the threshold. The formula introduces the difference between the current signal value and the reference value into the adjustment mechanism, allowing the system to adaptively adjust the threshold. This adaptive method enables the system to cope with complex and changing operating environments and avoids warning failures or false triggering caused by overly rigid threshold settings. The use of nonlinear control parameters δ and hyperbolic tangent function tanh enables the system to respond appropriately to signal changes of different amplitudes. The use of the hyperbolic tangent function can avoid excessive or insufficient threshold adjustments due to extreme changes, ensuring that the system remains stable when facing drastic signal changes. By adjusting the δ parameter, the sensitivity of the threshold adjustment can be flexibly controlled. When the δ value is large, the system is more sensitive to signal changes and is suitable for scenes with frequent fluctuations; when the δ value is small, the system adjusts smoothly and is suitable for relatively stable signal environments. This adjustable mechanism provides a more reasonable threshold adjustment strategy for different types of signals.

[0065] The fault warning module determines whether to trigger the warning mechanism based on the thresholds adjusted by the threshold self-tuning module. This means that if the threshold adjusted by the threshold self-tuning module exceeds the corresponding signal data in the subsequent monitoring process, the warning mechanism is triggered. By combining with the threshold self-tuning module, the system can dynamically adjust the threshold according to the real-time changes of the signal. This means that the warning mechanism can adapt to the signal fluctuations of the system in different states, whether it is due to signal changes in normal operation of the equipment or abnormal fluctuations caused by faults. This can avoid false alarms or missed alarms caused by the rigidity of the fixed threshold. After adjusting the threshold, the system can perform real-time monitoring based on the new threshold. If the signal exceeds the dynamically adjusted threshold, it means that the signal may exceed the normal operating range, so the warning mechanism will be triggered immediately. This design ensures that the system can issue a warning in time when a real abnormal situation is detected, reducing the false alarm rate and improving the accuracy of the warning. The monitoring process after the threshold is adjusted is more sensitive and can respond dynamically according to the degree of signal fluctuation. For example, in some environments, the signal has large noise or interference, and the adjusted threshold can increase the tolerance, while in fine monitoring scenarios, the threshold can be lowered to enhance sensitivity. The system can quickly respond to potential failures, preventing equipment from continuing to operate undetected in the early stages of a failure. Timely early warning triggers can respond before or at the early stages of a failure, reducing the possibility of serious damage to the equipment. Through real-time early warning and early intervention, the system can significantly reduce maintenance costs and downtime risks, and improve equipment reliability and service life.

[0066] The fault location module uses the cross-location algorithm, which means:

[0067] Combine sensors i and j into a positioning combination. For each positioning combination, obtain the signal arrival times ti and tj of sensors i and j. Different sensors receive signals at different times, resulting in a time difference. The time difference is calculated using the following formula:

[0068] dij=Vs*loge(1+|ti-tj|γ); Vs is the propagation speed of the signal, such as the speed of sound in the air or the speed of other signal propagation media, which is used to convert the time difference into the distance difference. γ is a preset nonlinear adjustment parameter used to control the nonlinear correction effect of the time difference. A larger γ value can amplify the effect of the time difference on positioning, while a smaller γ value can make the effect of the time difference smoother. dij represents the distance difference between the signal propagating from the local discharge source to sensors i and j, reflecting the conversion of the signal propagation time difference into a spatial distance difference, which is used for subsequent fault source location calculations. loge(1+|ti-tj|γ) is a nonlinear processing term used to correct the effect of the time difference. This term can improve the accuracy of positioning through nonlinear processing and avoid errors caused by noise or other interference.

[0069] The distance difference dij is used to solve the two-dimensional or three-dimensional spatial position of the local discharge source corresponding to each positioning combination, and then the average of all the obtained coordinate solutions is calculated to obtain the actual coordinate position of the local discharge source, which is specifically:

[0070] In two-dimensional space, let the coordinates of the local discharge source be (x, y), and the positioning combinations be (xi, yi) and (xj, yj). Then the distance difference dij is numerically equal to the following calculation result:

[0071]

[0072] In three-dimensional space, let the coordinates of the local discharge source be (x, y, z), and the positioning combinations be (xi, yi, zi) and (xj, yj, zj). Then the distance difference dij is numerically equal to the following calculation result:

[0073]

[0074] Solve the coordinates of the local discharge source corresponding to each positioning combination, then take the average of all the coordinate solutions to get the actual coordinate position of the local discharge source, and finally get the two-dimensional spatial position (xavg, yavg) or three-dimensional spatial position (xavg, yavg, zavg) of the local discharge source. The cross-positioning method can effectively utilize the sensor time difference information and accurately calculate the spatial position of the local discharge source.

[0075] By calculating the time difference between signals received by multiple sensors and converting these time differences into spatial distance differences, the system can determine the three-dimensional or two-dimensional spatial position of the partial discharge source through triangulation or cross-positioning methods. The nonlinear processing term in the formula can correct the time difference and reduce the error caused by signal fluctuations or noise interference, thereby improving the accuracy of fault source positioning. In practical applications, the signal propagation process may be subject to various interferences, such as reflection, scattering, etc., resulting in the signal propagation time not being completely linear. By introducing nonlinear adjustment parameters and performing nonlinear correction on the time difference, the system can better adapt to the complex signal propagation environment and ensure the stability and accuracy of positioning. The algorithm calculates the distance difference of the partial discharge source through the combination of multiple pairs of sensors (each pair of sensors i and j), and finally solves the spatial position of the partial discharge source based on the results of the combination of multiple sensors. By averaging the positioning results of multiple sensors, the errors of individual sensors can be effectively eliminated, and the robustness and overall accuracy of positioning can be improved. This method is particularly suitable for the detection of partial discharge faults. It can quickly calculate the location of the fault source after signal acquisition, helping operation and maintenance personnel to locate and repair problems in a timely manner to avoid further damage to the equipment.

[0076] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0077] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0078] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0079] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0080] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A transformer online monitoring and fault warning system, characterized in that: It includes signal acquisition module, signal filtering module, broadband focusing module, signal complexity analysis module, threshold self-tuning module, fault warning module and fault location module; The signal acquisition module collects all the operating data of the transformer in real time through the sensor network; The signal filtering module filters the signals collected by the sensor network to remove high-frequency noise and obtain a processed signal set; The broadband focusing module converts the broadband signal in the processed signal set into a narrowband signal through the broadband focusing algorithm, enhances the spatial resolution of the signal, and provides high-quality data for subsequent fault location and signal analysis; The signal complexity analysis module uses generalized entropy to analyze the complexity of the signal and evaluate the degree of chaos of signal changes; The threshold self-tuning module first determines the amount of threshold adjustment based on the degree of chaos of the signal change, and then dynamically adjusts the threshold according to the amplitude of the signal change to adapt to different signal fluctuations; The fault warning module determines whether to trigger the warning mechanism based on the threshold values ​​adjusted by the threshold self-tuning module; When the early warning mechanism is triggered, the fault location module uses a cross-location algorithm to determine the two-dimensional or three-dimensional spatial position of the local discharge source based on the signal arrival time difference.

2. A transformer online monitoring and fault warning system according to claim 1, characterized in that: The signal filtering module is used to filter the signals collected by the sensor network. The specific steps are as follows: The original signal xi(t) collected by each sensor at time t is multiplied by the preset weighting factor wi(t), so that each sensor signal is processed by the filter and the filtered signal y(t) is calculated according to the following formula: β is a preset nonlinear index, λ is a preset filter coefficient, and m is the number of sensors.

3. A transformer online monitoring and fault early warning system according to claim 2, characterized in that: The broadband focusing algorithm performs weighted focusing on each frequency component of the signal using the following formula: N is the total number of frequency components, f n is the frequency component, τ n (x) is the arrival time difference of the signal, which means the arrival time difference of each signal component between different sensors, α is the preset nonlinear control parameter, and w n is the preset proportional coefficient of the nth frequency component, f b (x, t) is the output result after the narrowband signal is processed by the broadband focusing algorithm, that is, the narrowband signal is converted from the broadband signal to the narrowband signal. x represents the coordinate of the signal in space, and t is the time variable, which represents the change of the signal in time.

4. A transformer online monitoring and fault warning system according to claim 3, characterized in that: The signal complexity analysis module uses generalized entropy to analyze the complexity of the signal. The evaluation of the degree of chaos of the signal change refers to: Obtain the signal distribution collected by the sensor network, and then calculate the complexity of the signal according to the generalized entropy formula. The specific formula is: pi represents the probability distribution of the i-th state in the signal, q is the preset control parameter of the generalized entropy, u is the number of states, and v is the preset fractional-order adjustment parameter used to adjust the calculation sensitivity of the signal complexity. It is the generalized entropy value, which is used to measure the complexity and chaos of the signal.

5. A transformer online monitoring and fault warning system according to claim 4, characterized in that: The threshold self-tuning module first determines the amount of threshold adjustment based on the degree of chaos of the signal change, which refers to: A detection signal type form is preset, which includes all signal types collected by the sensor network, and is calculated based on the average fluctuation amplitude and fault frequency of each signal type in historical data: F=AXi+BYi+C, Xi and Yi represent the average fluctuation amplitude and fault frequency of the signal type respectively, C is a constant, F is a sorting value, A and B are the preset influence coefficients of Xi and Yi respectively, and descending sorting is performed according to the sorting value F to obtain a signal type disorder sorting form, and then the product of the generalized entropy value and the preset transformation ratio coefficient K is calculated, and the obtained product value is rounded up to obtain the number of threshold adjustments, and finally the corresponding number of signal types in the signal type disorder sorting form is selected according to the number of threshold adjustments as the threshold adjustment form.

6. A transformer online monitoring and fault early warning system according to claim 5, characterized in that: The threshold self-tuning module dynamically adjusts the threshold value, which means: Get the initial thresholds for the signal types in the threshold adjustment form respectively, and then perform the following calculations: represents the reference value of the i-th signal type, S i represents the signal value currently collected by the i-th signal type, δ represents the preset nonlinear control parameter, represents the initial threshold of the i-th signal type, Represents the dynamically adjusted threshold for the ith signal type.

7. A transformer online monitoring and fault early warning system according to claim 6, characterized in that: The fault warning module determines whether to trigger the warning mechanism based on the threshold values ​​adjusted by the threshold self-tuning module: if the threshold value adjusted by the threshold self-tuning module exceeds the threshold value in the subsequent monitoring process, the warning mechanism is triggered.

8. A transformer online monitoring and fault early warning system according to claim 7, characterized in that: The fault location module uses the cross-location algorithm, which means: Combine sensors i and j into a positioning combination. For each positioning combination, obtain the signal arrival time t of sensors i and j. i and t j , and calculate the time difference using the following formula: d ij =V s *log e (1+|t i -t j | γ );V s is the propagation speed of the signal, γ is the preset nonlinear adjustment parameter, d ij It represents the difference in distance that the signal propagates from the PD source to sensors i and j.

9. A transformer online monitoring and fault early warning system according to claim 8, characterized in that: Use the distance difference d ij The two-dimensional or three-dimensional spatial position of the local discharge source corresponding to each positioning combination is solved, and then all the obtained coordinate solutions are averaged to obtain the actual coordinate position of the local discharge source.

Images