Leakage current signal preprocessing method for power utilization protection device

Abstract

The invention relates to the technical field of leakage current signal preprocessing, in particular to a leakage current signal preprocessing method for an electric protection device, and the method comprises the steps: obtaining a historical leakage current signal in a transformer substation, and inputting the historical leakage current signal into an LMS adaptive filter in a segmented manner; determining an initial step length factor of each section of input signal based on an error change condition between the adjacent section of input signal and the expected signal of the corresponding section; correcting the initial step length factor by combining all wave crests and all local error distribution in each section of input signal to obtain a final step length factor of each section of input signal, and de-noising each section of input signal; and calculating the de-noised signal of the current input signal by using the multiple sections of input signals with similar characteristics with the current input signal. The invention aims to improve the algorithm accuracy and accelerate the algorithm processing speed by analyzing the noise in the leakage current signal.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of leakage current signal preprocessing, in particular to a leakage current signal preprocessing method for an electric power protection device. BACKGROUND

[0002] The substation engineering construction operation is wide, the electric power safety protection is poor, and the personal electric shock risk of the construction personnel is relatively high when operating. In view of this situation, the electric power industry is actively taking measures to improve the construction safety. On the one hand, the risk of accidental injury is reduced through formulating and implementing strict safety operation procedures, providing necessary safety training, and equipping with personal protection equipment and the like. On the other hand, the industry also strengthens the safety management personnel training, improves the safety consciousness and legal consciousness of the safety management personnel, so as to cope with the safety hidden danger in the construction. Meanwhile, the risk management of the construction site is also emphasized, including optimizing the construction scheme, regularly checking and maintaining the equipment, strictly implementing the safety construction management system and the like.

[0003] On the basis of the above, in order to better protect the safety of the construction personnel, the safety of the substation also needs to be guaranteed by using the safety anti-electric shock protection device. However, in the substation, the device fault signal usually contains a large amount of noise when the anti-electric shock protection device acquires the leakage current signal, and the noise will interfere with the accuracy of the signal detection, and affect the judgment of the equipment state and the fault diagnosis. SUMMARY

[0004] In order to solve the above technical problems, the application provides a leakage current signal preprocessing method for an electric power protection device to solve the existing problems.

[0005] The leakage current signal preprocessing method for an electric power protection device provided by the application adopts the following technical scheme: One embodiment of the application provides a leakage current signal preprocessing method for an electric power protection device, which comprises the following steps: Step one: acquiring the historical leakage current signal in the substation and inputting the historical leakage current signal into the LMS adaptive filter in segments; Step two: adjusting the step length of the LMS adaptive filter in real time according to each segment of the input signal, using the adjusted filter to carry out the noise removal processing on the input signal, and calculating the signal after the noise removal of the current input signal; specifically: A1, acquiring the leakage current signal under the ideal state and segmenting, and respectively recording as the expected signal corresponding to the segment input signal; A2, determining the initial step length factor of each segment of the input signal based on the overall error change difference and the local error fluctuation amplitude between the adjacent segments of the input signal and the expected signal corresponding to the segment, and the average amplitude of each segment of the input signal; A3, correcting the initial step length factor by combining the average value of the half-peak width of all the peaks in each input signal segment and the average level of the variance of all the local errors, to obtain a final step length factor of each input signal segment, and denoising each input signal segment; A4, clustering all the input signal segments in the historical leakage current signal by using the feature difference between the input signals; selecting a cluster with the greatest similarity to the current input signal from the clustering results; and weighting and summing the signals after denoising all the input signals in the cluster by using the feature difference between the current input signal and all the input signals in the cluster, to obtain the signal after denoising the current input signal.

[0006] Preferably, the adjacent input signal segment is further determined as the input signal of the current input signal segment and the left adjacent input signal segment.

[0007] Preferably, the fluctuation change range of the local error is further determined as follows: calculating the fluctuation degree of the local error between each input signal segment and the expected signal corresponding to the segment; and taking the difference between the fluctuation degrees of adjacent segments as the fluctuation change range of the local error.

[0008] Preferably, the determination method of the initial step length factor is as follows: positively fusing the change difference of the overall error and the fluctuation change range of the local error; taking the ratio of the average amplitude to the positively fused result as the initial step length factor.

[0009] Preferably, the acquisition method of the final step length factor is as follows: the final step length factor of the nth input signal segment is denoted as , ; In the formula, the final step length factor of the nth input signal segment is denoted as the initial step length factor of the nth input signal segment is denoted as the noise intensity of the (n-1)th input signal segment is denoted as the average value of the half-peak width of all the peaks in the nth input signal segment is denoted as norm, and norm represents a normalization function.

[0010] Preferably, the noise intensity is determined by the average level of the square of the local error at all the collection time points in the input signal segment.

[0011] Preferably, the clustering distance in the clustering process is set as follows: the signal mean, signal skewness, signal kurtosis, noise intensity and frequency domain characteristic parameters of the input signal constitute the characteristic vector of the input signal; and the difference in the characteristic vectors between the input signals is used as the clustering distance between the input signals during clustering.

[0012] Preferably, the similarity is calculated by: the overall difference between the feature vectors of all segment input signals within the cluster and the current input signal.

[0013] Preferably, the method for obtaining the signal after denoising the current input signal is: Calculating and normalizing the feature difference between the current input signal and any segment of the input signal in the cluster, and weighting the denoised signal of any segment of the input signal; The weighted results of all the input signal segments in the cluster are summed up, and the weighted summed result is used as the denoised signal of the current input signal.

[0014] Preferably, after denoising the current input signal, the method further includes analyzing abnormal frequencies in the frequency domain of the denoised signal of the current input signal, and determining whether to perform power outage on the substation based on the analysis result, specifically: Perform Fourier transform on the denoised signal of the current input signal to obtain a frequency domain signal; If the frequency in the frequency domain signal that is greater than a preset amplitude threshold exceeds a preset percentage, it is determined that the current input signal is abnormal and the substation is powered off.

[0015] This application has at least the following beneficial effects: In response to the above problems, the present application combines LMS adaptive filtering to analyze and process the acquired historical leakage current signal to remove useless noise in the leakage current signal; and in order to meet the high requirements for demand response speed in substations, the initial step length factor is determined according to the difference between the input signal after segmentation of the historical leakage current signal and the expected signal, which is used to preliminarily determine the step length in the filtering process to prevent system oscillation or divergence, and at the same time accelerate convergence; then, the final step length factor is determined by using the peaks in the input signal and the distribution of local errors, and each segment of the input signal is denoised to ensure the stability of the filter; finally, by using the denoised signals of multiple segments of input signals similar to the current input signal in the historical leakage current signal, the denoised signal of the current input signal is calculated, which improves the accuracy of the algorithm while speeding up the algorithm processing speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the accompanying drawings required by the embodiments or the prior art description will be briefly introduced. Obviously, the accompanying drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0017] Figure 1 A flow chart of a leakage current signal preprocessing method for a power utilization protection device is provided in the present application. Figure 2 A flow chart of a calculation process of a signal after denoising of a current input signal is provided in an embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined purpose, the specific embodiments, structures, features and effects of the leakage current signal preprocessing method for a power utilization protection device according to the present application are described in detail below in combination with the accompanying drawings and preferred embodiments. Different "one embodiment" or "another embodiment" in the following description do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0020] The specific scheme of the leakage current signal preprocessing method for a power utilization protection device provided by the present application is specifically described below in combination with the accompanying drawings.

[0021] The leakage current signal preprocessing method for a power utilization protection device provided in an embodiment of the present application.

[0022] The substation engineering construction operation is wide, the power utilization safety protection is poor, and the risk of personal electric shock of the construction personnel during operation is higher. In order to protect the personal safety of the construction personnel, a safety anti-electric shock protection device can be used to handle the power supply in time when the leakage current danger occurs.

[0023] The main purpose of the present application is to improve the response speed as much as possible when necessary when the leakage current signal collected by the leakage current sensor in the substation is processed by using the LMS adaptive filter to improve the data quality, so that the signal segment with abnormality is paid attention to in time, and the stable operation of the overall system can be maintained in a long-term range.

[0024] Specifically, a leakage current signal preprocessing method for a power utilization protection device is provided as follows, please refer to Figure 1The method comprises the following steps: Step one: obtaining the historical leakage current signal in the substation and inputting it into the LMS adaptive filter in segments.

[0025] The leakage current sensor is used to monitor the leakage current signal in the substation in real time. Generally, when leakage current occurs, the leakage current sensor converts the leakage current signal into a voltage signal, which is proportional to the size of the leakage current.

[0026] However, the signal output by the leakage current sensor may be affected by noise during transmission, including electromagnetic interference, ripple of high-frequency current, etc. These noises may mask the true signal of the leakage current, resulting in inaccurate monitoring results of the leakage current in the substation by the electric protection device.

[0027] Therefore, in order to improve the effect of using LMS adaptive filter to process the collected leakage current signal, the signal needs to be preliminarily processed after being collected. Since the leakage current is a low-frequency signal, a low-pass filter is used to preliminarily process it. When designing the low-pass filter, in order to ensure that the main signal components can be retained and various noises generated by different reasons can be effectively attenuated, the cutoff frequency of the low-pass filter is set to 70Hz, and the initial signal is processed using the filter.

[0028] In order to make the processed result directly input into the LMS adaptive filter, the voltage signal output by the leakage current sensor needs to be restored to a current signal, the signal needs to be amplified, and the processed result needs to be input into the LMS adaptive filter in segments.

[0029] At this point, the leakage current signal in the substation can be obtained and input into the LMS adaptive filter in segments through the above steps.

[0030] Step two: adjusting the step size of the LMS adaptive filter in real time according to each segment of the input signal, using the adjusted filter to process the input signal, and calculating the signal after denoising of the current input signal.

[0031] In this embodiment, the calculation process flow chart of the signal after denoising of the current input signal is shown in the accompanying Figure 2 , and specifically: A1, obtaining the leakage current signal in the ideal state and segmenting it, and respectively recording it as the expected signal of the corresponding segment input signal.

[0032] The specific process is as follows: 1. Electrical model: First, collect the electrical characteristic parameters of various devices in the substation, including rated voltage, rated current, insulation resistance, leakage current threshold, etc. These parameters can be obtained through device manuals or field tests. According to the electrical characteristics of the device, the corresponding equivalent circuit model is established.

[0033] 2. Load model: Analyze the types of different loads in the substation (such as linear load, nonlinear load, etc.), and establish the corresponding load model according to its working state (such as normal operation, overload, short circuit, etc.). Simulate the dynamic changes of the load through simulation software, and observe its influence on the leakage current. For example, different load change conditions can be introduced in the power grid model to simulate the leakage current signal under different load conditions.

[0034] 3. Operation simulation: Use electromagnetic transient simulation software to establish a complete power grid model, including equivalent circuit model and load model, input the collected device parameters and load conditions in the simulation software, set the initial conditions and simulation time. Run the simulation to observe the leakage current signal under ideal conditions, and obtain multiple leakage current signals by setting different simulation scenarios.

[0035] 4. Signal amplification: Amplify the leakage current signal under ideal conditions according to the same signal amplification parameters in step one.

[0036] At this point, the leakage current signal under ideal conditions can be obtained, and the leakage current signal under ideal conditions can be segmented and recorded as the expected signal of the corresponding segment input signal.

[0037] A2, based on the overall error between the adjacent segment input signal and its corresponding segment expected signal, and the fluctuation amplitude of the local error, as well as the average amplitude of each segment input signal, determine the initial step length factor of each segment input signal.

[0038] Leakage current signals are usually collected in complex power environments, and various noises may be superimposed in the signals, which will affect the accuracy of the leakage current signal and make the detection of the leakage current difficult. Moreover, the composition of these noises is complex, and simple low-pass filtering processing cannot meet the requirements of noise removal in most cases. At this time, it is necessary to adjust the noise removal method according to the characteristics of the input signal in order to minimize the noise and improve the quality of the signal. In this case, the LMS adaptive filter can effectively suppress noise and preserve the true characteristics of the leakage current by adjusting the parameters in real time.

[0039] In the leakage current monitoring of the substation, accuracy is crucial, especially when comparing the difference between the leakage current and the normal current, errors may lead to false judgments. Therefore, the step of the LMS adaptive filter needs to be adjusted in real time in combination with the input signal, and the real-time nature of the step adjustment can more finely control the convergence process of the filter and reduce the estimation error caused by excessive adjustment.

[0040] First, the ideal state leakage current signal obtained in step A1 is selected according to the specific scene, and the corresponding segment is input as the expected signal into the LMS adaptive filter, and the difference between the input signal and the expected signal, i.e. the error, is monitored.

[0041] If the error continues to decrease, it means that the algorithm update is proceeding in the correct direction, at which time the step size can be appropriately increased to speed up convergence; if the error increases, it means that the step size is too large at this time, which can cause the update direction to deviate from the target, at which time the step size needs to be reduced to avoid system instability and prevent oscillation or divergence.

[0042] Therefore, the error change trend needs to be analyzed, the error between the input signal and the expected signal is calculated, and the error change trend in a short time is analyzed. When the error change trend of adjacent segments gradually increases, it is considered that the system has a divergence trend; when the error change trend of adjacent segments fluctuates greatly, it is considered that the system has an oscillation trend.

[0043] Preferably, in the present embodiment, the overall error between each input signal and its corresponding expected signal is calculated; the divergence trend of each input signal is determined according to the change difference of the overall error between adjacent segments. The fluctuation degree of the local error between each input signal and its corresponding expected signal is calculated; the difference between the fluctuation degrees of adjacent segments is taken as the fluctuation change amplitude of the local error, and the oscillation trend of each input signal is determined according to the fluctuation change amplitude of the local error between adjacent segments.

[0044] As an implementation, the divergence trend of the kth input signal is calculated as follows: For example, the calculation formula is: ; in the formula, the divergence trend of the kth input signal is represented; the mean of all the shortest distance values in the DTW distance matrix of the kth input signal and its expected signal is represented; the mean of all the shortest distance values in the DTW distance matrix of the k-1th input signal and its expected signal is represented; norm represents a normalization function. The calculation process of the DTW distance matrix is a known technology and will not be described again.

[0045] It should be understood that the DTW distance matrix between the input signal and its corresponding expected signal is dynamically calculated by the DTW algorithm, and the overall error change is analyzed, i.e. the change trend of the shortest path value in the DTW distance matrix is used to reflect the error change. When the overall error change trend in the DTW distance matrix gradually increases, it is considered that the system has a divergence trend; when the overall error change trend in the distance matrix fluctuates greatly, it is considered that the system has an oscillation trend.

[0046] At the same time, according to the above logic, we can get the oscillation trend. It represents the oscillation trend of the k-th input signal, which is obtained by the change in the variance of all distance values ​​in the DTW distance matrix between the k-th input signal and its expected signal. The calculation formula is: , where represents the variance of all distance values ​​in the DTW distance matrix between the kth segment input signal and its expected signal, Represents the variance of all distance values ​​in the DTW distance matrix between the k-1th input signal and its expected signal.

[0047] In substations, the strength of leakage current signals is often affected by factors such as power equipment failures and grounding issues, and the amplitude of the leakage current signal may fluctuate over time. When the leakage current signal amplitude is large, the step size can be appropriately increased to accelerate filter convergence; when the signal amplitude is small, the step size should be appropriately reduced to avoid residual noise caused by over-adjustment.

[0048] The divergence trend obtained from the change difference of the overall error analyzed above and the oscillation trend obtained from the fluctuation amplitude of the local error are combined with the average amplitude of each input signal to determine the initial step length factor of each input signal.

[0049] Preferably, in this embodiment, the method for determining the initial step length factor is: forwardly fusing the change difference of the overall error and the fluctuation change amplitude of the local error; and taking the ratio of the average amplitude to the forward fusion result as the initial step length factor.

[0050] It can be understood that forward fusion is a fusion method such as addition and multiplication between data. The specific forward fusion method is determined by the implementer according to the actual situation, and this application does not impose any special restrictions.

[0051] As an implementation method, the initial step length factor of the nth segment input signal is For example, the calculation formula is: Where, Indicates the initial step length factor of the nth segment input signal; Indicates the divergent trend of the n-1th segment input signal; Indicates the oscillation trend of the n-1th segment input signal; Indicates the average amplitude of the nth segment input signal.

[0052] It should be understood that Together they represent the convergence trend of the system represented by the nth segment input signal. When both the oscillation trend and the divergence trend are smaller, the corresponding step size should be larger, and the initial step length factor should be set larger.

[0053] A3, the initial step length factor is modified by combining the average value of the half-peak width of all the peaks in each input signal and the average level of the variance of all the local errors, to obtain the final step length factor of each input signal, and each input signal is denoised.

[0054] Background noise often exists in the leakage current signal, especially noise caused by equipment operation, external electromagnetic interference and other factors. If the noise is strong, the step length can be reduced to ensure the stability of the filter, and when the noise level is low, the step length can be increased to accelerate the convergence of the filter.

[0055] Since the noise signal is uncertain and cannot be sampled, the square of the signal error is used to estimate the strength of the noise, which is used to represent the noise level of the input signal in this segment.

[0056] Preferably, in this embodiment, the average level of the square of the local error at each collection time in each input signal is taken as the noise strength of the input signal in this segment.

[0057] As an implementation, the noise strength of the nth input signal is taken as an example, and the calculation formula is: ; In the formula, represents the noise strength of the nth input signal; represents the length of the nth input signal, that is, the number of collection times contained; represents the local error at the tth time in the nth input signal, that is, the shortest path value in the DTW distance matrix of the nth input signal and the expected signal at the tth time of the input signal.

[0058] At the same time, in the substation, the leakage current signal may be affected by the change of equipment state (such as switch operation, load fluctuation, etc.), causing the signal to become unstable in a short time. If the statistical characteristics of the signal change, it will directly affect the effect of the LMS algorithm, so it is necessary to use the method of sliding window to estimate the stationarity of the signal in a period of time, and if it is detected that the statistical characteristics of the signal change significantly, the step length should be reasonably adjusted. When the signal is stable, the step length can be increased to improve the processing efficiency; and when the signal is unstable, the step length should be reduced to improve the stability.

[0059] Preferably, in this embodiment, the final step length factor of the nth input signal is taken as an example, and the calculation formula is:

[0060] In the formula, represents the final step length factor of the nth input signal; ​​​an initial step length factor of the nth segment of the input signal; a noise intensity of the (n-1)th segment of the input signal; a half-peak width average of all peaks in the nth segment of the input signal; norm represents a normalization function.

[0061] It should be understood that when the half-peak width of the peak is larger, it means that the signal change is more moderate and the signal is more stable, so the step length can be increased to accelerate the convergence of the filter; when the noise intensity is stronger, the step length can be reduced to ensure the stability of the filter. In other embodiments of the present application, when calculating the final step length factor of the nth segment of the input signal , the variance of all signal values in the nth segment of the input signal can also be further considered . When the variance of the signal values is larger, it means that the signal values change more and the signal is less stable. After considering the variance , the calculation formula is: .

[0062] The final step length factor of each segment of the input signal is used as the step length when filtering the segment of the input signal using the LMS adaptive filter, and each segment of the input signal is denoised to obtain the output signal after denoising each segment of the input signal.

[0063] A4, using the feature difference between the input signals to cluster all segments of the input signals in the historical leakage current signal; selecting a cluster with the highest similarity to the current input signal from the clustering results; and using the feature difference between all input signals in the cluster and the current input signal to weight all input signals in the cluster, and using the weighted result as the signal after denoising the current input signal.

[0064] However, the process of obtaining the denoised signal by filtering is relatively complex and the amount of calculation is large, which may cause the algorithm to fail to respond quickly in some cases, so on the basis of using the LMS adaptive filter, in combination with the known input signal and the output signal, when the similarity between the current input signal and a certain segment of the input signal in the obtained historical leakage current signal is high enough, the output signal obtained before can be directly used as the signal after denoising the current input signal without filtering operation, which not only can improve the accuracy of the algorithm, but also can monitor in real time.

[0065] A41, using the feature difference between the input signals to cluster all segments of the input signals in the historical leakage current signal.

[0066] All the segment input signals in the historical leakage current signal are clustered to obtain a plurality of clusters, each of which contains a plurality of segment input signals. This embodiment uses the K-means clustering algorithm for clustering, and other embodiments may also use the DBSCAN clustering algorithm. The K-means clustering algorithm is a well-known technique and will not be described in detail.

[0067] Preferably, in this embodiment, the clustering distance between the input signals when clustering is performed is the characteristic difference between the input signals. The specific method for determining the clustering distance is: the signal mean, signal skewness, signal kurtosis, noise intensity and frequency domain characteristic parameters of the input signal are combined to form the characteristic vector of the input signal; the difference in the characteristic vectors between the input signals is used as the clustering distance between the input signals when clustering.

[0068] Among them, the signal mean, signal skewness, signal kurtosis, noise intensity and frequency domain characteristic parameters are all common knowledge and will not be described in detail in this embodiment.

[0069] In other embodiments, the signal difference between the input signals may be directly used as the clustering distance between the input signals.

[0070] A42 selects the cluster with the greatest similarity to the current input signal from the clustering results.

[0071] When the similarity of the input signals within a cluster obtained after clustering the current input signal and the input signals in the historical leakage current signal is high enough, the weighted average of the input signals within the cluster can be used as the denoised signal of the current input signal, saving calculation time.

[0072] Preferably, in this embodiment, the similarity is calculated by: the overall difference between the feature vectors of all the segment input signals in the cluster and the current input signal.

[0073] As an implementation method, the similarity between the current input signal and the mth cluster is used For example, the calculation formula is: Where, Indicates the current input signal and the The similarity between clusters; Represents the feature vector of the current input signal; Indicates the The characteristic vector of the input signal in the i-th segment of the cluster; norm represents the normalization function, Represents the number of input signal segments in the mth cluster. Recorded as the feature difference between the current input signal and the i-th input signal in the m-th cluster .

[0074] The cluster with the greatest similarity to the current input signal is screened out, which can reflect the relatively greater similarity between the current input signal and the input signals in the cluster.

[0075] A43, and the feature difference between all the input signals in the cluster and the current input signal is used to weight all the input signals in the cluster, and the weighted result is taken as the signal after denoising of the current input signal.

[0076] Preferably, in the embodiment, the method for obtaining the signal after denoising of the current input signal is: calculating the feature difference between the current input signal and any segment of input signal in the cluster and normalizing, weighting the signal after denoising of the any segment of input signal; summing the weighted results of all segments of input signal in the cluster, and taking the weighted sum result as the signal after denoising of the current input signal.

[0077] As an implementation, taking the mth cluster as an example of the cluster with the greatest similarity to the current input signal, the signal after denoising of the current input signal is calculated , and the calculation formula is: ; in the formula, represents the signal after denoising of the current signal; represents the feature difference between the current input signal and the ith segment of input signal in the mth cluster; represents the signal after denoising of the ith segment of input signal in the mth cluster; and norm represents a normalization function. The normalization function is intended to normalize all the feature differences in the mth cluster, that is, to ensure that the sum of the normalized results of all the feature differences in the mth cluster is 1, which is used as a weight to weight the signal after denoising of the corresponding segment of input signal.

[0078] When there is no cluster satisfying the similarity condition, an LMS adaptive filter needs to be used to calculate the output value, and when there are multiple clusters satisfying the similarity condition, the cluster with the greatest similarity is selected. In addition, without using the LMS adaptive filter to process the data, the step size parameter and the weight parameter are still obtained according to the related steps, so as not to affect the subsequent calculation.

[0079] Step three: analyzing the abnormal frequency in the frequency domain of the signal after denoising of the current input signal, and judging whether to perform power-off processing on the transformer substation according to the analysis result.

[0080] After denoising the data, the current input signal needs to be further analyzed to determine whether there is a leakage problem by analyzing the fluctuation of the signal after denoising of the current input signal.

[0081] The Fourier transform is performed on the signal after the current input signal is denoised to obtain a frequency domain signal, and it is checked whether there is abnormal fluctuation of certain specific frequency in the frequency domain signal. This method can identify periodic faults or noise sources in the operation of the device, evaluate the signal strength at different frequencies, and find out the abnormal frequency band. The Fourier transform is a well-known technology and will not be described again.

[0082] A magnitude threshold is preset, and the size of the magnitude threshold is set as the average amplitude in the normal working state of the power utilization protection device. The specific value is determined by the implementer according to the mean value of the frequency domain signal in the normal working state period marked by artificial.

[0083] If the frequency greater than the magnitude threshold in the frequency domain signal exceeds the preset percentage, it is determined that the current input signal is abnormal. In this embodiment, the preset percentage is 5%, which can be set by the implementer. Then the power utilization protection device is controlled to cut off the power supply of the transformer substation, and the transformer substation is powered off to avoid more serious electrical accidents caused by leakage current.

[0084] Each embodiment in the present application is described in a progressive manner, and the same and similar parts between each embodiment can be referred to each other. Each embodiment focuses on the difference from other embodiments.

[0085] It should be noted that unless otherwise specified and limited, terms such as "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the circuit structure, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such article or device. Without more limitations, the element limited by the statement "including a" does not exclude the presence of another identical element in the article or device including the element. In addition, the term "and / or" used herein includes any and all combinations of one or more related listed items.

[0086] Other embodiments of the present application will be readily apparent to those skilled in the art upon considering the description herein with the benefit of the present disclosure. The present application is intended to cover any variations, uses, or adaptations of the application following the general principles thereof and including such departures from the present disclosure as come within known or customary practice within the art to which the application pertains.

[0087] It should be understood that the present application is not limited to the precise structures described and shown herein, and that various modifications and changes can be made without departing from its scope.

Claims

1. A leakage current signal preprocessing method for an electrical protection device, characterized in that: The method comprises the following steps: Step 1: Obtain the historical leakage current signal in the substation and input it into the LMS adaptive filter in sections; Step 2: Adjust the step size of the LMS adaptive filter in real time according to the input signal of each segment, use the adjusted filter to denoise the input signal, and calculate the denoised signal of the current input signal; specifically: A1, obtain the leakage current signal under ideal conditions and divide it into segments, each of which is recorded as the expected signal of the corresponding segment input signal; A2, based on the difference in overall error between adjacent input signal segments and their corresponding expected signals and the fluctuation amplitude of local errors, as well as the average amplitude of each input signal segment, determines the initial step length factor of each input signal segment; A3, combining the average half-width of all peaks in each input signal segment and the average level of the variance of all local errors, correcting the initial step length factor to obtain a final step length factor for each input signal segment, and denoising each input signal segment; A4, clustering all segment input signals in the historical leakage current signal using the feature differences between the input signals; screening out the cluster with the greatest similarity to the current input signal from the clustering results; and using the feature differences between all input signals in the cluster and the current input signal, performing weighted summation on the denoised signals of all input signals in the cluster, and using the result of the weighted summation as the denoised signal of the current input signal.

2. A leakage current signal preprocessing method for an electric protection device according to claim 1, characterized in that: The adjacent segment input signals are further determined to be the current segment input signal and the input signal of the left adjacent segment.

3. A leakage current signal preprocessing method for an electric protection device according to claim 2, characterized in that: The fluctuation range of the local error is further determined by calculating the fluctuation degree of the local error between each input signal segment and the expected signal of the corresponding segment; and taking the difference in the fluctuation degree between adjacent segments as the fluctuation range of the local error.

4. A leakage current signal preprocessing method for an electric protection device according to claim 1, characterized in that: The method for determining the initial step length factor is: Forward fusion of the change difference of the overall error and the fluctuation change amplitude of the local error; The ratio of the average amplitude to the forward fusion result is used as the initial step length factor.

5. A leakage current signal preprocessing method for an electric protection device according to claim 4, characterized in that: The method for obtaining the final step length factor is: The final step length factor of the nth segment input signal is recorded as , ; Where, represents the final step length factor of the nth segment input signal; Indicates the initial step length factor of the nth segment input signal; Indicates the noise intensity of the n-1th segment input signal; It represents the average half-width of all peaks in the nth segment of the input signal; norm represents the normalization function.

6. A leakage current signal preprocessing method for an electric protection device according to claim 5, characterized in that: The noise intensity is determined by the average level of the square of the local errors in the input signal at all acquisition moments.

7. The leakage current signal preprocessing method for an electric protection device according to claim 1, characterized in that: The clustering distance in the clustering process is set as follows: the signal mean, signal skewness, signal kurtosis, noise intensity and frequency domain characteristic parameters of the input signal are combined into a feature vector of the input signal; and the difference in the feature vectors between the input signals is used as the clustering distance between the input signals during clustering.

8. The leakage current signal preprocessing method for an electric protection device according to claim 1, characterized in that: The similarity is calculated by: the overall difference between the feature vectors of all the segment input signals in the cluster and the current input signal.

9. The leakage current signal preprocessing method for an electric protection device according to claim 1, characterized in that: The method for obtaining the signal after the current input signal is denoised is: Calculating and normalizing the feature difference between the current input signal and any segment of the input signal in the cluster, and weighting the denoised signal of any segment of the input signal; The weighted results of all the input signal segments in the cluster are summed up, and the weighted summed result is used as the denoised signal of the current input signal.

10. The leakage current signal preprocessing method for an electric protection device according to claim 1, characterized in that: After the current input signal is de-noised, the method further includes analyzing the abnormal frequencies in the frequency domain of the de-noised signal and determining whether to cut off the power supply to the substation based on the analysis results. Specifically, the method includes: Perform Fourier transform on the denoised signal of the current input signal to obtain a frequency domain signal; If the frequency in the frequency domain signal that is greater than a preset amplitude threshold exceeds a preset percentage, it is determined that the current input signal is abnormal and the substation is powered off.

Images