Power supply telecontrol detection system capable of accurately positioning faults
By combining initial current acquisition, feature extraction, deep learning analysis and intelligent frequency control modules in the power supply remote detection system, the current acquisition frequency is dynamically adjusted, which solves the problem that existing systems cannot detect dynamic changes in the transformer winding in a timely manner, and improves fault detection The accuracy and real-time nature of the company prevent potential failures.
Patent Information
- Application Number
- CN202411097457.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-11-01
Smart Images

Figure CN118884126A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power supply telecontrol detection, and in particular to a power supply telecontrol detection system for accurately locating faults. Background Art
[0002] The power supply remote detection system with accurate fault location is an efficient power monitoring and management system, mainly used for fault monitoring and location of power systems. The system can quickly detect abnormal conditions in the power system by monitoring the operating status of power equipment in real time, collecting and analyzing parameters such as current, voltage, and power. When a fault occurs, the system can accurately locate the fault point and transmit the fault information to the monitoring center through remote communication technology, so that the staff can handle and maintain the fault in time, thereby reducing power outage time and improving power supply reliability.
[0003] In addition, the system usually integrates advanced intelligent algorithms and data analysis technologies, which can intelligently diagnose and predict faults and discover potential hidden dangers in advance. This preventive maintenance capability can effectively reduce the failure rate of equipment and extend the service life of equipment. The power supply telecontrol detection system can also be interconnected with other power management systems to achieve global optimization and coordinated control, further improving the operating efficiency and safety of the power system. This not only improves the operating efficiency of power companies, but also provides users with a more stable and reliable power supply.
[0004] The power supply telecontrol detection system monitors the current of the transformer windings. Its main function is to detect the current status of the windings in real time, and promptly detect and warn of problems such as winding short circuits, local overheating, and current imbalance. This monitoring can help the power system quickly locate the fault point, avoid further damage to the transformer, reduce power outage time and maintenance costs, and improve the reliability and safety of power supply. Through accurate monitoring, the system can detect potential problems at an early stage and implement preventive maintenance, thereby extending the service life of the transformer.
[0005] When the power supply remote detection system monitors the current of the transformer winding, the current acquisition frequency is usually pre-set and cannot be dynamically adjusted. If the current acquisition frequency of the existing technology cannot meet the monitoring requirements of the dynamic current changes of the transformer winding, it may cause the transformer winding short circuit to not be discovered in time. Transformer winding short circuit is one of the common faults of transformers. Failure to discover it in time may cause winding overheating and insulation damage, and eventually cause the transformer to catch fire or even explode, resulting in large-scale power outages and equipment damage.
[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention
[0007] The purpose of the present invention is to provide a power supply remote detection system for accurately locating faults. The current information of the transformer winding is obtained by pre-setting the initial current acquisition frequency, and the current information is extracted and deeply analyzed. These simplified features are input into a pre-trained deep learning model to accurately identify complex patterns in the current characteristics. In this way, the system can monitor the dynamic changes of the current of the transformer winding in real time, and divide it into stable dynamic change type and rapid dynamic change type. For the rapid dynamic change type, the system will intelligently adjust the actual current acquisition frequency to capture the details of the current change at a higher frequency. This combination of dynamic adjustment and intelligent analysis greatly improves the accuracy and real-time performance of fault detection, ensuring that any abnormalities in the transformer winding can be discovered in time, thereby preventing potential faults and accidents, so as to solve the problems in the above-mentioned background technology.
[0008] In order to achieve the above-mentioned object, the present invention provides the following technical solutions: a power supply remote detection system for accurately locating faults, comprising an initial current acquisition module, a feature extraction module, a deep learning analysis module, a dynamic change classification module, a stable dynamic monitoring module and an intelligent frequency control module; The initial current acquisition module obtains the current information of the transformer winding through the current sensor at a preset initial current acquisition frequency to ensure that the current changes can be captured under normal circumstances; The feature extraction module extracts features from the current information of the transformer winding after obtaining it, and further analyzes the extracted current features to convert the complex current data into simplified features that can be used for subsequent analysis. The deep learning analysis module inputs the analyzed current characteristics into a pre-trained deep learning model to accurately identify the complex patterns in the current characteristics, thereby performing intelligent analysis of the dynamic changes of the current; The dynamic change classification module divides the dynamic change of transformer winding current into stable dynamic change type and rapid dynamic change type based on the output of the deep learning model; The stable dynamic monitoring module continues to acquire current information at the preset initial current acquisition frequency for transformer winding current information classified as stable dynamic change type, efficiently capturing current changes while saving resource consumption; The intelligent frequency control module intelligently controls the actual current acquisition frequency based on the preset initial current acquisition frequency for the fast and dynamically changing transformer winding current information. By dynamically adjusting the acquisition frequency, the system can capture the current change details at a higher frequency and detect potential faults and abnormalities in a timely manner.
[0009] Preferably, after obtaining the current information of the transformer winding, the extracted current information features include instantaneous peaks in the current signal and spectral energy changes in the current signal. Within the monitoring window, the instantaneous peaks in the current signal and the spectral energy changes in the current signal are further analyzed to generate a peak current index and a spectral energy change index, thereby converting complex current data into simplified features that can be used for subsequent analysis.
[0010] Preferably, the analyzed peak current index and spectrum energy variation index are input into a pre-trained deep learning model to generate a dynamic current variation coefficient, and the complex patterns in the current characteristics are accurately identified through the dynamic current variation coefficient.
[0011] Preferably, the dynamic current change coefficient output by the deep learning model is compared and analyzed with a preset dynamic current change coefficient reference threshold, and the current dynamic change of the transformer winding is divided, and the division result is as follows: If the dynamic current variation coefficient is greater than or equal to the dynamic current variation coefficient reference threshold, the current dynamic variation of the transformer winding under the monitoring window is classified as a fast dynamic variation type; If the dynamic current variation coefficient is less than the dynamic current variation coefficient reference threshold, the current dynamic variation of the transformer winding in the monitoring window is classified as a stable dynamic variation type.
[0012] Preferably, the specific steps of intelligently regulating the actual current acquisition frequency based on the preset initial current acquisition frequency are as follows: Set the initial current acquisition frequency to , the reference threshold of dynamic current change coefficient is ; For the dynamic current change coefficient greater than the dynamic current change coefficient reference threshold generated by the deep learning model under the monitoring window , based on the initial current acquisition frequency , Dynamic current change coefficient reference threshold and a dynamic current change coefficient greater than a dynamic current change coefficient reference threshold The actual current acquisition frequency is intelligently controlled, and the control formula is: , where Indicates the actual current acquisition frequency after intelligent control. Represents the adjustment coefficient, which controls the amplitude of frequency adjustment. Its value range is [0, 1]. The actual current acquisition frequency calculated Applied to the current information collection of transformer windings, ensuring high-frequency acquisition can capture details of fast dynamic changes.
[0013] Preferably, within the monitoring window, after further analyzing the instantaneous peak in the current signal, the steps of generating the peak current index are as follows: In the monitoring window, the acquired current signal is used To represent the current signal Perform preprocessing to remove noise and interference and obtain the denoised current signal ; The spike is detected by calculating the instantaneous rate of change of the current signal. First, the first-order derivative of the current signal is calculated. and the second derivative , the calculation expression is: , , and secondly, the first-order derivative and the second derivative The first-order derivative reference threshold is set and the second derivative reference threshold Perform comparative analysis to detect peak points , if satisfied and , then mark the point as a peak point; For each detected peak point, extract the peak amplitude and spike duration , the peak amplitude is the value of the current signal at the peak point, and the peak duration is the time interval of the peak duration, which is calculated by the zero crossing points before and after the peak. The calculation expression is: ,in, and are the zero crossing points before and after the spike, respectively; Calculate the energy of each spike , the calculation expression is: ; Calculate the number of spike points per unit time to represent the spike density. The calculation expression of the spike density is: , where represents the spike density, Indicates the length of the monitoring window, represents the number of spikes, which indicates the total number of spikes detected within the monitoring window; The spike energy and spike density are combined to generate the spike current index, and the generated expression is: , where represents the peak current index, represents the energy of the i-th peak.
[0014] Preferably, within the monitoring window, after further analyzing the spectrum energy variation in the current signal, the steps of generating the spectrum energy variation index are as follows: The current signal is preprocessed to remove noise and DC offset to make the signal smoother. The preprocessed current signal is segmented according to the monitoring window, each segment is N in length, and the current signal segment in the kth monitoring window is marked as , n is the index of the discrete time point, indicating the specific sampling point position in each monitoring window (from 0 to N−1); Perform Fourier transform on the current signal segment of each monitoring window to obtain the frequency domain signal. The expression of Fourier transform is: , where represents the frequency domain signal of the kth monitoring window, f represents the frequency, and j represents the imaginary unit; Calculate the energy of each frequency component and add them up to get the total energy. The calculation expression is: , represents the spectrum amplitude of the current signal at frequency f in the kth monitoring window, represents the total spectrum energy of the kth monitoring window; Calculate the spectrum energy change rate of adjacent monitoring windows. The calculation expression is: , where represents the rate of change of spectrum energy in the kth monitoring window; The spectrum energy change rate is smoothed, and the smoothing expression is: , represents the rate of change of spectrum energy after smoothing, Represents the smoothing coefficient, and its value range is between [0, 1]. It represents the total spectrum energy of the previous monitoring window after smoothing; Calculate the spectrum energy variation index. The calculation expression is: , where Represents the spectrum energy change index of the kth monitoring window.
[0015] In the above technical solution, the technical effects and advantages provided by the present invention are: The present invention obtains the current information of the transformer winding by pre-setting the initial current acquisition frequency, extracts features and conducts in-depth analysis on the current information, and inputs these simplified features into a pre-trained deep learning model, thereby accurately identifying complex patterns in the current characteristics. In this way, the system can monitor the dynamic changes of the current of the transformer winding in real time, and divide it into stable dynamic change type and rapid dynamic change type. For the rapid dynamic change type, the system will intelligently adjust the actual current acquisition frequency to capture the details of the current change at a higher frequency. This combination of dynamic adjustment and intelligent analysis greatly improves the accuracy and real-time performance of fault detection, ensuring that any abnormalities in the transformer winding can be discovered in time, thereby preventing potential faults and accidents. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.
[0017] Figure 1 The module diagram of a power supply remote detection system for accurately locating faults according to the present invention is shown in FIG. DETAILED DESCRIPTION
[0018] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.
[0019] The present invention provides Figure 1 A power supply remote detection system for accurately locating faults shown in the figure includes an initial current acquisition module, a feature extraction module, a deep learning analysis module, a dynamic change classification module, a stable dynamic monitoring module and an intelligent frequency control module; The initial current acquisition module obtains the current information of the transformer winding through the current sensor at a preset initial current acquisition frequency to ensure that the current changes can be captured under normal circumstances; The power supply remote detection system uses current sensors to collect current data of transformer windings at a preset frequency, so that the current changes can be effectively monitored under normal operating conditions to ensure the stable and safe operation of the power system.
[0020] The initial current acquisition frequency is a fixed sampling frequency pre-set according to the operating characteristics of the transformer and the requirements of the power system. This frequency is usually determined based on the design specifications of the transformer, historical operating data, and possible fault modes. Its purpose is to be able to collect current data at a sufficient frequency under normal circumstances to capture the details of current changes. For example, for a transformer operating at 50Hz, the acquisition frequency may be set between several hundred Hz and several kHz to ensure that the current change trend and transient events can be accurately captured.
[0021] The setting of the initial current acquisition frequency also needs to consider the system resources and data processing capabilities. Although a too high acquisition frequency can provide more detailed data, it will increase the burden of data processing and storage requirements. On the other hand, a too low acquisition frequency may result in the inability to timely discover the details of current changes and potential faults. Therefore, the setting of the initial acquisition frequency needs to find a balance between data accuracy and system resources to ensure the effectiveness of monitoring without placing too much burden on the system.
[0022] The feature extraction module extracts features from the current information of the transformer winding after obtaining it, and further analyzes the extracted current features to convert the complex current data into simplified features that can be used for subsequent analysis. After obtaining the current information of the transformer winding, the extracted current information features include instantaneous peaks in the current signal and spectral energy changes in the current signal. Within the monitoring window, the instantaneous peaks in the current signal and the spectral energy changes in the current signal are further analyzed to generate a peak current index and a spectral energy change index, thereby converting complex current data into simplified features that can be used for subsequent analysis.
[0023] If there are many spike currents in the current signal of the transformer winding, it indicates that there are complex dynamic current changes in the transformer winding. Spike currents are usually caused by instantaneous load changes, switching operations or fault events. Their frequent occurrence indicates that there are frequent and violent fluctuations in the current signal. This fluctuation not only reflects the rapid changes in current in a short period of time, but also may reveal that the transformer winding is affected by unstable operating conditions. These conditions may include sudden load fluctuations in the power system, partial discharge inside the transformer or interference from other power equipment. Therefore, the frequent occurrence of spike currents indicates that there are high nonlinearities and irregularities in the current signal.
[0024] After further analysis of the instantaneous spikes in the current signal within the monitoring window, the steps to generate the spike current index are as follows: In the monitoring window, the acquired current signal is used To represent the current signal Perform preprocessing to remove noise and interference and obtain the denoised current signal ; Wavelet transform can be used for denoising.
[0025] The spike is detected by calculating the instantaneous rate of change of the current signal. First, the first-order derivative of the current signal is calculated. and the second derivative , the calculation expression is: , , and secondly, the first-order derivative and the second-order derivative The first-order derivative reference threshold is set and the second derivative reference threshold Perform comparative analysis to detect peak points , if satisfied and , then mark the point as a peak point; For each detected peak point, extract the peak amplitude and spike duration , the peak amplitude is the value of the current signal at the peak point, and the peak duration is the time interval of the peak duration, which is calculated by the zero crossing points before and after the peak. The calculation expression is: ,in, and are the zero crossing points before and after the spike, respectively; The zero crossing point refers to the point where the current signal changes from a positive value to a negative value or from a negative value to a positive value. In the time series of the current signal, the zero crossing point indicates the moment when the current signal passes through the zero value (that is, the current is zero). In other words, the zero crossing point is the point where the current signal curve intersects the time axis.
[0026] Calculate the energy of each spike , the calculation expression is: ; Calculate the number of spike points per unit time to represent the spike density. The calculation expression of the spike density is: , where represents the spike density, Indicates the length of the monitoring window, represents the number of spikes, which indicates the total number of spikes detected within the monitoring window; The spike energy and spike density are combined to generate the spike current index, and the generated expression is: , where represents the peak current index, represents the energy of the i-th peak.
[0027] It can be seen from the peak current index that within the monitoring window, the greater the performance value of the peak current index generated after further analysis of the instantaneous spikes in the current signal, the more complex the dynamic changes of the current in the transformer winding are. This is because the peak current index reflects the nonlinearity and irregularity of the current signal by comprehensively measuring the frequency, amplitude and energy characteristics of the instantaneous spikes in the current signal. When the peak current index is large, it means that there are more spike events in the current signal. These spike events are usually caused by rapid load changes, faults or other transient phenomena, indicating that the transformer winding is in a complex dynamic operating state. On the contrary, if the peak current index is small, it means that the current signal is relatively stable, there are fewer instantaneous spike events, the dynamic changes of the current in the transformer winding are less complex, and the operating state is relatively stable.
[0028] If the spectrum energy in the current signal of the transformer winding changes rapidly, it indicates that there are complex dynamic current changes in the transformer winding. This is because the spectrum energy reflects the distribution of the current signal on different frequency components, and the rapidly changing spectrum energy means that the frequency components of the current signal have changed significantly in a short period of time. Such changes are usually caused by a variety of factors, such as sharp fluctuations in load, electromagnetic interference, and the sudden increase or disappearance of harmonics. These factors work together to cause the current signal to present complex dynamic characteristics. Rapid changes in spectrum energy not only reflect the fluctuations in current amplitude, but also reveal potential high-frequency harmonics and transient phenomena in the current signal. These phenomena indicate that the winding current is not a stable and steady state, but contains complex and rapidly changing dynamic processes. Therefore, the rapid change of spectrum energy is an important feature of complex dynamic current changes in the winding.
[0029] Within the monitoring window, after further analyzing the spectrum energy variation in the current signal, the steps to generate the spectrum energy variation index are as follows: The current signal is preprocessed to remove noise and DC offset to make the signal smoother. The preprocessed current signal is segmented according to the monitoring window, each segment is N in length, and the current signal segment in the kth monitoring window is marked as , n is the index of the discrete time point, indicating the specific sampling point position in each monitoring window (from 0 to N−1); Perform Fourier transform on the current signal segment of each monitoring window to obtain the frequency domain signal. The expression of Fourier transform is: , where represents the frequency domain signal of the kth monitoring window, f represents the frequency, and j represents the imaginary unit; Calculate the energy of each frequency component and add them up to get the total energy. The calculation expression is: , represents the spectrum amplitude of the current signal at frequency f in the kth monitoring window, represents the total spectrum energy of the kth monitoring window; Calculate the spectrum energy change rate of adjacent monitoring windows. The calculation expression is: , where represents the rate of change of spectrum energy in the kth monitoring window; The spectrum energy change rate is smoothed (to reduce the impact of high-frequency noise). The smoothing expression is: , represents the rate of change of spectrum energy after smoothing, Represents the smoothing coefficient, and its value range is between [0, 1]. It represents the total spectrum energy of the previous monitoring window after smoothing; Calculate the spectrum energy variation index. The calculation expression is: , where Represents the spectrum energy change index of the kth monitoring window.
[0030] It can be seen from the spectrum energy variation index that within the monitoring window, the larger the performance value of the spectrum energy variation index generated after further analysis of the spectrum energy variation in the current signal, the more complex the dynamic changes of the current in the transformer winding are. Specifically, the spectrum energy variation index reveals the rapid changes of the frequency components in the current signal by measuring the rate of change of the spectrum energy between different monitoring windows. When the spectrum energy variation index is large, it means that the spectrum energy has changed significantly in a short period of time, reflecting the complex dynamic changes in the current signal, including high-frequency harmonics, transient phenomena, and sharp load fluctuations. On the contrary, when the spectrum energy variation index value is small, it means that the current signal is relatively stable and the changes in the frequency components are not obvious, indicating that the dynamic changes of the current in the transformer winding are relatively simple and stable.
[0031] The deep learning analysis module inputs the analyzed current characteristics into a pre-trained deep learning model to accurately identify the complex patterns in the current characteristics, thereby performing intelligent analysis of the dynamic changes of the current; The peak current index after analysis and spectral energy variation index Input into the pre-trained deep learning model to generate dynamic current change coefficients , through the dynamic current variation coefficient Accurately identify complex patterns in current signatures; The deep learning model is not specifically limited here, and can achieve the peak current index and spectral energy variation index Perform comprehensive analysis to generate dynamic current change coefficient In order to realize the technical solution of the present invention, the present invention provides a specific implementation method; Dynamic current variation coefficient The resulting calculation formula is: , where , Peak Current Index and spectral energy variation index The preset scaling factor of , Both are greater than 0.
[0032] It can be seen from the dynamic current variation coefficient that, under the monitoring window, the larger the expression value of the peak current index generated after further analysis of the instantaneous peak in the current signal, the larger the expression value of the spectrum energy variation index generated after further analysis of the spectrum energy variation in the current signal, that is, the larger the expression value of the dynamic current variation coefficient generated under the monitoring window, the more complex the dynamic current variation in the transformer winding is, and vice versa, the less complex the dynamic current variation in the transformer winding is.
[0033] The dynamic change classification module divides the dynamic change of the transformer winding current into stable dynamic change type and rapid dynamic change type based on the output results of the deep learning model; The dynamic current change coefficient output by the deep learning model is compared and analyzed with the preset dynamic current change coefficient reference threshold, and the dynamic change of the current of the transformer winding is divided. The results of the division are as follows: If the dynamic current variation coefficient is greater than or equal to the dynamic current variation coefficient reference threshold, the current dynamic variation of the transformer winding under the monitoring window is classified as a fast dynamic variation type; If the dynamic current variation coefficient is less than the dynamic current variation coefficient reference threshold, the current dynamic variation of the transformer winding in the monitoring window is classified as a stable dynamic variation type.
[0034] The stable dynamic monitoring module continues to acquire current information at the preset initial current acquisition frequency for transformer winding current information classified as stable dynamic change type, efficiently capturing current changes while saving resource consumption; During the monitoring process, if the deep learning model determines that the current change of the transformer winding is stable, that is, the current signal has no significant dynamic changes or complex frequency component fluctuations, then the system will continue to use the initially set acquisition frequency to obtain current data. This acquisition frequency is pre-set according to the normal operating status and monitoring requirements of the transformer, and can effectively capture the normal changes in current.
[0035] The benefit of maintaining the initial acquisition frequency is that it ensures that the system can continuously monitor the current without missing any subtle changes, while avoiding the data processing and storage burden caused by excessively high frequencies. Although high-frequency acquisition can capture more signal details, it will generate a large amount of data, increasing the system's processing pressure and storage requirements. For stable current changes, this high-frequency acquisition is not necessary. Therefore, by maintaining the initial frequency, while ensuring the monitoring effect, the use of system resources can be optimized, and the consumption of data processing and storage can be reduced, thereby achieving efficient use of resources and long-term stable operation of the system.
[0036] Intelligent frequency control module, for fast and dynamic changing transformer winding current information, intelligently controls the actual current acquisition frequency based on the preset initial current acquisition frequency. By dynamically adjusting the acquisition frequency, the system can capture the current change details at a higher frequency and detect potential faults and abnormalities in a timely manner; The specific steps for intelligently controlling the actual current acquisition frequency based on the preset initial current acquisition frequency are as follows: Set the initial current acquisition frequency to , the reference threshold of dynamic current change coefficient is ; For the dynamic current change coefficient greater than the dynamic current change coefficient reference threshold generated by the deep learning model under the monitoring window , based on the initial current acquisition frequency , Dynamic current change coefficient reference threshold and a dynamic current change coefficient greater than a dynamic current change coefficient reference threshold The actual current acquisition frequency is intelligently controlled, and the control formula is: , where Indicates the actual current acquisition frequency after intelligent control. It represents the adjustment coefficient, which controls the amplitude of frequency adjustment. Its value range is between [0, 1] and can be determined according to actual needs. No specific limitation is given here. The actual current acquisition frequency calculated Applied to the current information collection of transformer windings, ensuring that high-frequency acquisition can capture the details of fast dynamic changes; It should be noted that in order to prevent the actual current acquisition frequency after intelligent control from being too high, which will cause the system to be overloaded, an upper limit of the actual acquisition frequency is set. .
[0037] The present invention obtains the current information of the transformer winding by pre-setting the initial current acquisition frequency, extracts features and conducts in-depth analysis on the current information, and inputs these simplified features into a pre-trained deep learning model, thereby accurately identifying complex patterns in the current characteristics. In this way, the system can monitor the dynamic changes of the current of the transformer winding in real time, and divide it into stable dynamic change type and rapid dynamic change type. For the rapid dynamic change type, the system will intelligently adjust the actual current acquisition frequency to capture the details of the current change at a higher frequency. This combination of dynamic adjustment and intelligent analysis greatly improves the accuracy and real-time performance of fault detection, ensuring that any abnormalities in the transformer winding can be discovered in time, thereby preventing potential faults and accidents.
[0038] 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.
[0039] 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.
[0040] The above only describes some exemplary embodiments of the present invention by way of illustration. It is undoubted that, for those skilled in the art, the described embodiments can be modified in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.
Claims
1. A power supply remote detection system for accurately locating faults, It is characterized in that It includes initial current acquisition module, feature extraction module, deep learning analysis module, dynamic change classification module, stable dynamic monitoring module and intelligent frequency control module; The initial current acquisition module obtains the current information of the transformer winding through the current sensor at a preset initial current acquisition frequency to ensure that the current changes can be captured under normal circumstances; The feature extraction module extracts features from the current information of the transformer winding after obtaining it, and further analyzes the extracted current features to convert the complex current data into simplified features that can be used for subsequent analysis. The deep learning analysis module inputs the analyzed current characteristics into a pre-trained deep learning model to accurately identify the complex patterns in the current characteristics, thereby performing intelligent analysis of the dynamic changes of the current; The dynamic change classification module divides the dynamic change of transformer winding current into stable dynamic change type and rapid dynamic change type based on the output of the deep learning model; The stable dynamic monitoring module continues to acquire current information at the preset initial current acquisition frequency for transformer winding current information classified as stable dynamic change type, efficiently capturing current changes while saving resource consumption; The intelligent frequency control module intelligently controls the actual current acquisition frequency based on the preset initial current acquisition frequency for the fast and dynamically changing transformer winding current information. By dynamically adjusting the acquisition frequency, the system can capture the current change details at a higher frequency and detect potential faults and abnormalities in a timely manner.
2. A power supply remote detection system for accurately locating faults according to claim 1, It is characterized in that After obtaining the current information of the transformer winding, the extracted current information features include instantaneous peaks in the current signal and spectral energy changes in the current signal. Within the monitoring window, the instantaneous peaks in the current signal and the spectral energy changes in the current signal are further analyzed to generate a peak current index and a spectral energy change index, thereby converting complex current data into simplified features that can be used for subsequent analysis.
3. A power supply remote detection system for accurately locating faults according to claim 2, It is characterized in that The analyzed peak current index and spectrum energy variation index are input into a pre-trained deep learning model to generate a dynamic current variation coefficient, which is used to accurately identify complex patterns in current characteristics.
4. A power supply remote detection system for accurately locating faults according to claim 3, It is characterized in that The dynamic current change coefficient output by the deep learning model is compared and analyzed with the preset dynamic current change coefficient reference threshold, and the dynamic change of the current of the transformer winding is divided. The results of the division are as follows: If the dynamic current variation coefficient is greater than or equal to the dynamic current variation coefficient reference threshold, the current dynamic variation of the transformer winding under the monitoring window is classified as a fast dynamic variation type; If the dynamic current variation coefficient is less than the dynamic current variation coefficient reference threshold, the current dynamic variation of the transformer winding in the monitoring window is classified as a stable dynamic variation type.
5. A power supply remote detection system for accurately locating faults according to claim 4, It is characterized in that The specific steps for intelligently controlling the actual current acquisition frequency based on the preset initial current acquisition frequency are as follows: Set the initial current acquisition frequency to , the reference threshold of dynamic current change coefficient is ; For the dynamic current change coefficient greater than the dynamic current change coefficient reference threshold generated by the deep learning model under the monitoring window , based on the initial current acquisition frequency , Dynamic current change coefficient reference threshold and a dynamic current change coefficient greater than a dynamic current change coefficient reference threshold The actual current acquisition frequency is intelligently controlled, and the control formula is: , where Indicates the actual current acquisition frequency after intelligent control. Represents the adjustment coefficient, which controls the amplitude of frequency adjustment. Its value range is [0, 1]. The actual current acquisition frequency calculated Applied to the current information collection of transformer windings, ensuring high-frequency acquisition can capture details of fast dynamic changes.
6. A power supply remote detection system for accurately locating faults according to claim 2, It is characterized in that After further analysis of the instantaneous spikes in the current signal within the monitoring window, the steps to generate the spike current index are as follows: In the monitoring window, the acquired current signal is used To represent the current signal Perform preprocessing to remove noise and interference and obtain the denoised current signal ; The spike is detected by calculating the instantaneous rate of change of the current signal. First, the first-order derivative of the current signal is calculated. and the second derivative , the calculation expression is: , , and secondly, the first-order derivative and the second derivative The first-order derivative reference threshold is set and the second derivative reference threshold Perform comparative analysis to detect peak points , if satisfied and , then mark the point as a peak point; For each detected peak point, extract the peak amplitude and spike duration , the peak amplitude is the value of the current signal at the peak point, and the peak duration is the time interval of the peak duration, which is calculated by the zero crossing points before and after the peak. The calculation expression is: ,in, and are the zero crossing points before and after the spike, respectively; Calculate the energy of each spike , the calculation expression is: ; Calculate the number of spike points per unit time to represent the spike density. The calculation expression of the spike density is: , where represents the spike density, Indicates the length of the monitoring window, represents the number of spikes, which indicates the total number of spikes detected within the monitoring window; The spike energy and spike density are combined to generate the spike current index, and the generated expression is: , where represents the peak current index, represents the energy of the i-th peak.
7. A power supply remote detection system for accurately locating faults according to claim 2, It is characterized in that Within the monitoring window, after further analyzing the spectrum energy variation in the current signal, the steps to generate the spectrum energy variation index are as follows: The current signal is preprocessed to remove noise and DC offset to make the signal smoother. The preprocessed current signal is segmented according to the monitoring window, each segment is N in length, and the current signal segment in the kth monitoring window is marked as , n is the index of the discrete time point, indicating the specific sampling point position in each monitoring window (from 0 to N−1); Perform Fourier transform on the current signal segment of each monitoring window to obtain the frequency domain signal. The expression of Fourier transform is: , where represents the frequency domain signal of the kth monitoring window, f represents the frequency, and j represents the imaginary unit; Calculate the energy of each frequency component and add them up to get the total energy. The calculation expression is: , represents the spectrum amplitude of the current signal at frequency f in the kth monitoring window, represents the total spectrum energy of the kth monitoring window; Calculate the spectrum energy change rate of adjacent monitoring windows. The calculation expression is: , where represents the rate of change of spectrum energy in the kth monitoring window; The spectrum energy change rate is smoothed, and the smoothing expression is: , represents the rate of change of spectrum energy after smoothing, Represents the smoothing coefficient, and its value range is between [0, 1]. It represents the total spectrum energy of the previous monitoring window after smoothing; Calculate the spectrum energy variation index. The calculation expression is: , where Represents the spectrum energy change index of the kth monitoring window.
Citation Information
Cited By
Terminal security intelligent monitoring system and method based on Internet of Things
CN119355426A
An Internet of Things-based terminal security intelligent monitoring system and method
CN119355426B
Circuit fault feature active enhancement detection method for distribution box
CN119988841A
