WO2022037172A1 - 一种采样数据异常值修复方法及装置 - Google Patents

一种采样数据异常值修复方法及装置 Download PDF

Info

Publication number
WO2022037172A1
WO2022037172A1 PCT/CN2021/096654 CN2021096654W WO2022037172A1 WO 2022037172 A1 WO2022037172 A1 WO 2022037172A1 CN 2021096654 W CN2021096654 W CN 2021096654W WO 2022037172 A1 WO2022037172 A1 WO 2022037172A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
sampling
sequence
sampled
value
Prior art date
Application number
PCT/CN2021/096654
Other languages
English (en)
French (fr)
Inventor
庞吉耀
Original Assignee
南京磐能电力科技股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 南京磐能电力科技股份有限公司 filed Critical 南京磐能电力科技股份有限公司
Publication of WO2022037172A1 publication Critical patent/WO2022037172A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
    • G06F11/14—Error detection or correction of the data by redundancy in operation
    • G06F11/1479—Generic software techniques for error detection or fault masking

Definitions

  • the invention relates to a method and device for detecting and repairing abnormal values of single-point sampling data, belonging to the technical field of data acquisition and processing.
  • abnormal electrical quantity sampling has become the main cause of protection malfunction. Therefore, it is necessary to judge the sampling value in real time to avoid abnormal data causing malfunction of protection device and system malfunction.
  • Even small abnormal data will affect the measurement accuracy, and the accurate and rapid elimination of abnormal data will further improve the control accuracy and system response time of the industrial control system.
  • the abnormal point detection delay must be very short, but it is necessary to accurately distinguish the flying point sampling data and the normal transient sampling value when the system fails within a short detection data window. , with great difficulty.
  • the system is equipped with two sets of front-end sampling modules. Only when the data of the two sets of sampling modules are normal, the back-end application module will work normally. Otherwise, it will enter the equipment failure response mode. When a single set of sampling modules is abnormal, the application system can ensure that the application system does not send malfunctions.
  • the application module function can be activated only when the sampling data that reaches the threshold in the sampling data window reaches a certain number. Therefore, this solution is suitable for application scenarios that are not sensitive to response time.
  • the above methods can reduce or avoid the impact of abnormal sampling data to a certain extent, but they all have some shortcomings, such as poor pertinence, low sensitivity, difficult threshold setting, long data window requirements, poor real-time performance, and long system response time , can not improve the impact of abnormal points on precision measurement and other issues:
  • the object of the present invention is to overcome the deficiencies in the prior art, and to provide a method and device for repairing abnormal values of sampled data that are simple and easy to implement, have small delay, and have high recovery accuracy, do not need to calculate threshold values, have low computational complexity, and are suitable for hardware acceleration. Fast anomaly detection and outlier repair techniques for sampled data.
  • the present invention adopts the following technical solutions to realize:
  • the present invention provides a method for repairing abnormal values of sampled data, the method comprising the following steps:
  • the method further includes the following steps: continuously sampling the signal at equal intervals to obtain the original sampling value sequence ⁇ x(n) ⁇ .
  • construction method of the data window comprises the following steps:
  • x 5 corresponds to the latest sampling point x(n) at the current moment
  • n is the sequence number of the current latest sampling point in the sampling sequence ⁇ x(n) ⁇
  • x 4 corresponds to The value x(n-1) of the last sampling beat of x(n) in the original sampling sequence
  • x3 corresponds to the value x(n-2) of the last two sampling beats of x(n) in the original sampling sequence
  • x(n-2) 2 corresponds to the repaired sampling data y(m-1) of the previous beat in the repaired output sequence ⁇ y(m) ⁇ at the current moment
  • x 1 corresponds to the sampled data y of the last two sampling beats in the repaired
  • the method for identifying abnormal data includes the following steps:
  • the method for repairing the suspected abnormal data includes the following steps: taking the initial value of the intermediate variable u as x 3 , after finding the suspected abnormal data, if (x 3 -x 4 ) ⁇ (x 4 -x 5 ) is a negative number, Then recalculate the intermediate variable u according to formula (1);
  • the method also includes the steps of: moving the read pointer of the original sampled value sequence and the read/write pointer of the repaired output sequence after the current data window data restoration is completed, and continuing to process the sampled data until the original sampled data is processed.
  • the present invention provides an apparatus for repairing abnormal values of sampled data, the apparatus comprising:
  • Data window module used to select the sampled data in the original sampled value sequence and combine the sampled data in the repaired output sequence to construct a data window;
  • Identification module used to identify abnormal data in the sampled data in the data window
  • Repair module It is used to repair the suspected abnormal data in combination with the sampled data in the data window.
  • sampling module used to sample the signal at equal intervals to generate a sequence of original sampling values.
  • the present invention also provides an apparatus for repairing abnormal values of sampled data, including a processor, a storage medium and an I/O interface device; the storage medium is used for storing instructions; the I/O interface device is used for obtaining The original sampling value sequence collected offline or online; the processor is configured to operate according to the instruction to perform the steps of the above-mentioned method for repairing abnormal values of sampled data.
  • the present invention also provides a computer-readable storage medium above, on which a computer program is stored, and when the program is executed by the processor, the original sampling value sequence is extracted from the I/O interface device, and the above-mentioned sampling data abnormal value restoration method is realized. A step of.
  • the invention constructs a short data window by selecting the sampling data in the original sampling value sequence and combining the sampling data in the repairing output sequence, identifying the abnormal data of the sampling data in the data window according to the extreme value principle, and combining the sampling data in the data window to identify the suspected abnormal data It has the advantages of simple and easy implementation, less data buffering, small delay, high recovery accuracy, no need to calculate the threshold value, low calculation amount, suitable for hardware acceleration, etc., and has good application prospects. .
  • Fig. 1 is the data window structure schematic diagram of the present invention
  • Fig. 2 is the abnormal data detection and repair flow chart of the present invention
  • Figure 3 is sample abnormal data case 1
  • Figure 4 is sample abnormal data case 2
  • Figure 5 is sample abnormal data case 3
  • FIG. 6 is a diagram showing the effect of removing and repairing abnormal data according to the present invention.
  • the purpose of the present invention is to overcome the deficiencies in the prior art, to provide a simple and easy to implement, low delay, high recovery accuracy, no need to calculate the threshold value and low computational complexity, and a method and device for repairing abnormal values of sampled data suitable for hardware acceleration. Fast anomaly detection and outlier repairing technical issues for data.
  • the method for repairing abnormal values of sampled data of the present invention firstly uses continuous three-point extreme value detection to find suspected abnormal sampling data, then further detects adjacent three-point extreme value detection to confirm abnormal sampling points, performs suspected abnormal value identification at most three times, and then selects adjacent three-point extreme value detection to confirm abnormal sampling points. Correctly sample data for data repair.
  • abnormal sampling data means that the sampling value does not match the actual value of the system or the error exceeds a certain range. Locally, it is expressed as the local maximum value (or minimum value) of the sampling sequence, as shown in Figure 3, Appendix. Figure 4 and Figure 5.
  • the extreme point of a function may appear at a discontinuity point, a boundary point or a point where the derivative is 0. The commonly used judgment methods are as follows:
  • f'(x 0 ) 0, and for There are f'(x 0 )>0 (or f'(x 0 ) ⁇ 0) at the same time If f'(x 0 ) ⁇ 0 (or f'(x 0 )>0), then x 0 is the maximum point (or minimum point) of f(x), and the corresponding f(x 0 ) is the function f (x) maximum value (or minimum value).
  • the present invention adopts the difference to replace the differential, and transforms the sampling abnormal point detection into finding the sampling data points that conform to the formula (4) in the sampling sequence ⁇ x(n) ⁇ of the continuous signal:
  • x(n) is the maximum value (or minimum value) among these three points, and is also suspected abnormal sampling data.
  • the values x(t 0 ), x(t 1 ), x(t 2 ) of the known signal x(t) at times t 0 , t 1 , and t 2 correspond to time t 0 ⁇ t ⁇ t 2
  • the sampling value of the signal at time t can be expressed as:
  • the present invention removes the current suspected abnormal sampling data, and at the same time uses the adjacent normal data to estimate the real value of the system at the sampling abnormal time by formula (5), and replaces the abnormal sampling of the abnormal point with this. data to restore data.
  • the implementation case of the present invention is based on a 5-point data window ⁇ x 1 , x 2 , x 3 , , x 4 , x 5 ⁇ , and its construction method is shown in FIG. 1 .
  • x 1 is the earliest
  • x 2 , x 3 , and x 4 are the next
  • x 5 is the latest.
  • x 1 , x 2 come from the repaired output sequence ⁇ y(m) ⁇ , x 3 , x 4 , x 5 correspond to the original sampling sequence ⁇ x(n) ⁇ , where x 5 corresponds to the latest sampling point at the current moment x(n), n is the sequence number of the current latest sampling point in the sampling sequence ⁇ x(n) ⁇ , x 4 corresponds to the value x(n-1), x 3 of the previous sampling beat of x(n) in the original sampling sequence Corresponds to the value x(n-2) of the last two sampling beats of x(n) in the original sampling sequence, and x2 corresponds to the repaired sampling data y(m) of the previous beat in the repaired output sequence ⁇ y(m) ⁇ at the current moment -1), x 1 corresponds to the sampled data y(m-2) of the last two sampling beats in the repair output sequence ⁇ y(m) ⁇ , m is the latest repair to be written to the repair output sequence ⁇ y(
  • Figures 3 to 4 show a case with a sampling abnormal point x3 (at time t), in which Figures 3a and 4a are the abnormal points in the monotonically decreasing curve, and Figures 3b and 4b are monotonous Abnormal points appear in the rising curve, while Figures 5a and 5b correspond to the shifts in Figures 3a and 3b, respectively.
  • the present invention adopts different abnormal point identification and repair methods according to the position of the abnormal point in the data window.
  • Scenario (1) is similar to Figure 3a.
  • the sampled data at time t is suspected to be abnormal, and the corresponding value is x 3 .
  • the curve segment formed by x 2 , x 3 , and x 4 has a minimum value x 3 at t, and the The curve segment formed by x 3 , x 4 , and x 5 has a maximum value of x 4 .
  • the present invention selects data points x 1 , x 2 , x 5 to estimate x 3 .
  • Scenario (2) is similar to Figure 5a.
  • the sampled data at time t is suspected to be abnormal, corresponding to the sampled value x 3 .
  • the curve segment composed of x 2 , x 3 , and x 4 has a maximum value x 3 at t, and at the same time
  • the curve segment consisting of x 1 , x 2 , and x 3 has a minimum value x 2 .
  • the present invention selects data points x 1 , x 4 , x 5 to estimate x 3 .
  • the curve segment composed of x 2 , x 3 , and x 4 has an extreme value x 3 at t, and at the same time, it is composed of x 1 , x 2 , x 3 .
  • the curve segment and the curve segment formed by x 3 , x 4 , and x 5 do not have extreme values.
  • the present invention selects data points x 1 , x 2 , and x 4 to estimate x 3 .
  • Fig. 1 shows the construction method of the 5-point data window ⁇ x 1 , x 2 , x 3 , , x 4 , x 5 ⁇ of the present invention
  • Fig. 2 shows the detection and repair process based on a single abnormal point.
  • the technical implementation of the invention includes the following steps:
  • Step (1) continuously sample the signal at interval T s to obtain the original sample value sequence ⁇ x(n) ⁇ ;
  • Step (2) take the latest three sampling points of the original signal sample value sequence ⁇ x(n) ⁇ and the latest two sampling points in the repaired output sequence ⁇ y(m) ⁇ to form a data window of 5 sampled data, Denoted as ⁇ x 1 , x 2 , x 3 , x 4 , x 5 ⁇ , where x 5 corresponds to the latest sampling point x(n) at the current moment, and n is the current latest sampling point in the sampling sequence ⁇ x(n) ⁇
  • the serial number of x4 corresponds to the value x(n-1) of the last sampling beat of x(n) in the original sampling sequence, and x3 corresponds to the value x(n) of the last two sampling beats of x(n) in the original sampling sequence -2), and x 2 corresponds to the repair sample data y(m-1) of the previous beat in the repair output sequence ⁇ y(m) ⁇ at the current moment, and x 1 corresponds to the upper two in the repair output sequence ⁇ y(m
  • Step (3) take u as an intermediate variable and set the initial value to x 3 , calculate the sign of the expression [(x 2 -x 3 ) ⁇ (x 3 -x 4 )], if sgn[(x 2 -x 3 ) ⁇ (x 3 -x 4 )] is -1, then go to step (4), otherwise go to step (6), where sgn[x] represents the sign of the parameter x, if the parameter is negative, output -1, If the parameter is regular, output +1, if the parameter is 0, output 0;
  • Step (4) calculate the sign of the expression [(x 3 -x 4 ) ⁇ (x 4 -x 5 )], if sgn[(x 3 -x 4 ) ⁇ (x 4 -x 5 )] is -1 , then firstly update the intermediate variable u according to formula (9) and then go to step (6), otherwise go to step (5) directly:
  • Step (5) calculate the sign of the expression [(x 1 -x 2 ) ⁇ (x 2 -x 3 )], if sgn[(x 1 -x 2 ) ⁇ (x 2 -x 3 )] is -1 , then first update the intermediate variable u according to formula (10) and then go to step (6); otherwise, update the intermediate variable according to formula (11) and then go to step (6);
  • the algorithm will obtain the repaired output sequence ⁇ y(m) ⁇ of the original sampling sequence, and the repaired output sequence ⁇ y(m) ⁇ lags two times in time compared to the original input sequence ⁇ x(n) ⁇ sampling interval.
  • Fig. 6 shows the effect of anomaly detection and repair of the present invention
  • Fig. 6a is a sinusoidal sampling sequence containing abnormal data
  • Fig. 6b is a sinusoidal sampling sequence after the repair processing of the present invention, which further proves the implementation effect of the present invention .
  • the present invention constructs a 5-point data window by combining the original sampling sequence ⁇ x(n) ⁇ and the repaired output sequence ⁇ y(m) ⁇ , and identifies suspected anomalies through the difference sign change between three consecutive points in the data window. point, and reasonably select three adjacent points to complete the repair of suspected abnormal data points through interpolation, and identify suspected abnormal values at most three times.
  • the invention has the advantages of simple and easy implementation, less data buffering, small output delay, high abnormal data recovery accuracy, no need for dynamic calculation threshold, low calculation amount, suitable for hardware acceleration, etc., and has good application prospects.
  • Embodiment 2 is a diagrammatic representation of Embodiment 1:
  • the present invention provides an apparatus for repairing abnormal values of sampled data, which can be used to implement the method steps described in Embodiment 1, and the apparatus includes:
  • Data window module used to select the sampled data in the original sampled value sequence and combine the sampled data in the repaired output sequence to construct a data window;
  • Identification module used to identify abnormal data in the sampled data in the data window
  • Repair module It is used to repair the suspected abnormal data in combination with the sampled data in the data window.
  • sampling module used to sample the signal at equal intervals to generate a sequence of original sampling values
  • the present invention also provides a sampled data abnormal value restoration device, comprising a processor, a storage medium and an I/O interface device; the storage medium is used for storing instructions; the I/O interface device is used to obtain offline collected data. Or a sequence of raw sample values collected online; the processor is configured to operate according to the instructions to perform steps according to the following method:
  • Step (1) obtain the original sampling value sequence ⁇ x(n) ⁇ collected offline through the I/O interface device, or continuously sample the signal online at an interval T s to obtain the original sampling value sequence ⁇ x(n) ⁇ ;
  • Step (2) take the latest three sampling points of the original signal sampling value sequence ⁇ x(n) ⁇ and the latest two sampling points in the repaired output sequence ⁇ y(m) ⁇ to form a data window of 5 sampling data, Denoted as ⁇ x 1 , x 2 , x 3 , x 4 , x 5 ⁇ , where x 5 corresponds to the latest sampling point x(n) at the current moment, and n is the current latest sampling point in the sampling sequence ⁇ x(n) ⁇
  • the serial number of x4 corresponds to the value x(n-1) of the last sampling beat of x(n) in the original sampling sequence, and x3 corresponds to the value x(n) of the last two sampling beats of x(n) in the original sampling sequence -2), and x 2 corresponds to the repair sample data y(m-1) of the previous beat in the repair output sequence ⁇ y(m) ⁇ at the current moment, and x 1 corresponds to the upper two in the repair output sequence ⁇ y(m)
  • Step (3) take u as an intermediate variable and set the initial value to x 3 , calculate the sign of the expression [(x 2 -x 3 ) ⁇ (x 3 -x 4 )], if sgn[(x 2 -x 3 ) ⁇ (x 3 -x 4 )] is -1, then go to step (4), otherwise go directly to step (6), where sgn[x] represents the sign of the parameter x;
  • Step (4) calculate the sign of the expression [(x 3 -x 4 ) ⁇ (x 4 -x 5 )], if sgn[(x 3 -x 4 ) ⁇ (x 4 -x 5 )] is -1 , then first update the intermediate variable u according to formula (12) and then go to step (6), otherwise go to step (5) directly:
  • Step (5) calculate the sign of the expression [(x 1 -x 2 ) ⁇ (x 2 -x 3 )], if sgn[(x 1 -x 2 ) ⁇ (x 2 -x 3 )] is -1 , then first update the intermediate variable u according to formula (13) and then go to step (6), otherwise, update the intermediate variable according to formula (14) and then go to step (6);
  • the algorithm will obtain the repaired output sequence ⁇ y(m) ⁇ of the original sampling sequence, and the repaired output sequence ⁇ y(m) ⁇ lags two times in time compared to the original input sequence ⁇ x(n) ⁇ sampling interval.
  • Embodiment 4 is a diagrammatic representation of Embodiment 4:
  • An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the following method are implemented:
  • Step (1) extract the original sample value sequence ⁇ x(n) ⁇ from the I/O interface device
  • Step (2) take the latest three sampling points of the original signal sample value sequence ⁇ x(n) ⁇ and the latest two sampling points in the repaired output sequence ⁇ y(m) ⁇ to form a data window of 5 sampled data, Denoted as ⁇ x 1 , x 2 , x 3 , x 4 , x 5 ⁇ , where x 5 corresponds to the latest sampling point x(n) at the current moment, and n is the current latest sampling point in the sampling sequence ⁇ x(n) ⁇
  • the serial number of x4 corresponds to the value x(n-1) of the last sampling beat of x(n) in the original sampling sequence, and x3 corresponds to the value x(n) of the last two sampling beats of x(n) in the original sampling sequence -2), and x 2 corresponds to the repair sample data y(m-1) of the previous beat in the repair output sequence ⁇ y(m) ⁇ at the current moment, and x 1 corresponds to the upper two in the repair output sequence ⁇ y(m
  • Step (3) take u as an intermediate variable and set the initial value to x 3 , calculate the sign of the expression [(x 2 -x 3 ) ⁇ (x 3 -x 4 )], if sgn[(x 2 -x 3 ) ⁇ (x 3 -x 4 )] is -1, then go to step (4), otherwise go directly to step (6), where sgn[x] represents the sign of the parameter x;
  • Step (4) calculate the sign of the expression [(x 3 -x 4 ) ⁇ (x 4 -x 5 )], if sgn[(x 3 -x 4 ) ⁇ (x 4 -x 5 )] is -1 , then update the intermediate variable u according to formula (15) and then go to step (6), otherwise go to step (5) directly:
  • Step (5) calculate the sign of the expression [(x 1 -x 2 ) ⁇ (x 2 -x 3 )], if sgn[(x 1 -x 2 ) ⁇ (x 2 -x 3 )] is -1 , then first update the intermediate variable u according to formula (16) and then go to step (6); otherwise, update the intermediate variable according to formula (17) and then go to step (6);
  • the algorithm will obtain the repaired output sequence ⁇ y(m) ⁇ of the original sampling sequence, and the repaired output sequence ⁇ y(m) ⁇ will be temporally compared to the original input sequence ⁇ x(n) ⁇ lag by two sample intervals.
  • the embodiments of the present application may be provided as a method, a system, or a computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
  • computer-usable storage media including, but not limited to, disk storage, CD-ROM, optical storage, etc.
  • These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions
  • the apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Complex Calculations (AREA)
  • Emergency Protection Circuit Devices (AREA)

Abstract

本发明公开了一种采样数据异常值修复方法，通过选取原始采样值序列中采样数据并结合修复输出序列中的采样数据构建数据窗，根据函数极值原理对数据窗内采样数据进行异常数据识别，并结合数据窗内采样数据对疑似异常的数据进行修复，实现了对采样异常数据的剔除。本发明具有简单易实现，延迟小，修复精度高，无需计算门槛值和运算量低等优点，具有良好的应用前景。

Description

一种采样数据异常值修复方法及装置 技术领域
本发明涉及一种对单点采样数据异常值的检测和修复方法与装置，属于数据采集和处理技术领域。
背景技术
随着电子技术的进步，在工业过程控制、精密测量、电力自动化等领域，数字采样系统的应用越来越广泛。对模拟采样的系统而言，由于采样元件的偶发故障或者外部的电气骚扰，会使采样数据出错，对基于数字传输的远程采样系统来说，即使前端采样结果正确，也会因为通信系统的偶发性误码未被系统检出导致错误的数据。根据工程实践，对于雷击浪涌等强干扰导致的异常数据以及由通信系统误码引发的异常数据大多时候偏离正常数据集较远，比较容易识别，而对于磁场骚扰等，采样系统多表现为个别采样点轻微异常，往往会导致识别不出来，另外一方面，工业过程量和电力系统的电气量受负载扰动、系统控制或设备故障等原因往往会伴随着暂态过程，引发测量信号的暂态波动进一步增加异常数据识别难度。
对继电保护来说，电气量采样异常已经成为造成保护误动的主要原因，因此必须对采样值进行实时判断，避免异常数据引起保护装置误动、系统误操作等情况发生。而对测量系统来说，即使较小的异常数据也会影响测量精度，准确快速地剔除异常数据还将进一步提高工业控制系统的控制精度和系统响应时间。对于工业过程控制和电力系统继电保护等系统，要求异常点检测延时必须很短，但是要实现在较短的检测数据窗内准确区分飞点采样数据和系统故障时 的正常暂态采样值，有很大的难度。
目前，对于采样异常点的影响已有了一些解决方法：
(1)、滤波法
在采样前端增加模拟滤波器，并降低滤波器带宽和提升阻带衰减，或者将采样数据序列经过数字滤波器，可以根据需要增加滤波器的冲激响应长度，利用数字滤波器来减少或消弱异常点的影响。
(2)、冗余采样方案
系统配置两套前端采样模块，只有两套采样模块数据正常后端应用模块才正常工作，否则进入设备故障应对模式，在单套采样模块异常时可以保证应用系统不发送误动作。
(3)、基于门槛的采样点判决
由于电磁骚扰或其偶发性采样数据异常所持续的时间一般很短，通过设置固定的门槛或动态计算的门槛，只有在采样数据窗内达到门槛的采样数据达到一定数量才可以启动应用模块功能。所以此方案适用于对响应时间不敏感的应用场景。
以上几种方法能在一定程度上减少或避免异常采样数据的影响，均存在一些不足，如针对性不强、灵敏度不高、门槛不易整定、数据窗要求较长、实时性差、系统响应时间长，无法改善异常点对精密测量的影响等问题：
(1)、无论模拟滤波器还是数字滤波器，其实质是将单个异常点的影响分摊到后续的多个采样点中，并没有针对性地剔除异常点，增加传统滤波器的滤波效果是以延长系统响应时间为代价的。此外对于一些精密测量系统来说，滤波器的影响并不能改善异常点对测量精度的影响；
(2)、对于外部电磁骚扰引发的采样异常，即使采样前端冗余也无法解决，因为此时冗余系统采样结果是一致的，而且若冗余系统中的一套系统出现偶发性采样异常必将影响系统整体响应时间。
(3)、基于门槛值的多次判别实际上延长了系统响应时间，严重影响对时间敏感的应用实时性能，如基于采样值策略超快速继电保护、基于采样值控制的电力电子系统等。
综上所述，采样数据的快速异常检测和异常值修复是当前首要解决的问题。
发明内容
本发明的目的在于克服现有技术中的不足，提供一种简单易实现，延迟小，恢复精度高，无需计算门槛值和运算量低且适合硬件加速的采样数据异常值修复方法与装置，解决采样数据的快速异常检测和异常值修复技术问题。
为达到上述目的，本发明是采用下述技术方案实现的：
第一方面，本发明提供了一种采样数据异常值修复方法，所述方法包括以下步骤：
选取原始采样值序列中采样数据并结合修复输出序列中的采样数据构建数据窗；根据函数极值原理对数据窗内采样数据进行异常数据识别；结合数据窗内采样数据对疑似异常的数据进行修复。
进一步的，所述方法还包括以下步骤：对信号进行等间隔连续采样，获得原始采样值序列{x(n)}。
进一步的，所述数据窗的构建方法包括以下步骤：
取原始采样值序列{x(n)}的最新三个采样点以及修复输出序列{y(m)}中的最新两个采样点，构成5个采样数据的数据窗，记为{x 1,x 2,x 3,x 4,x 5}，其中，x 5对应 当前时刻的最新采样点x(n)，n为采样序列{x(n)}中当前最新采样点的序号，x 4对应原始采样序列中x(n)的上一采样节拍的值x(n-1)，x 3对应原始采样序列中x(n)的上两个采样节拍的值x(n-2)，而x 2对应当前时刻修复输出序列{y(m)}中上一节拍的修复采样数据y(m-1)，x 1对应修复输出序列{y(m)}中上两个采样节拍的采样数据y(m-2)，m为拟写入到修复输出序列{y(m)}的最新修复输出值序号。
进一步的，识别异常数据的方法包括以下步骤：
若(x 2-x 3)×(x 3-x 4)为非负数，则认为该所述数据窗的采样值正常，否则判别该所述数据窗的采样值为疑似异常。
进一步的，疑似异常数据的修复的方法包括如下步骤：取中间变量u的初值为x 3，发现疑似异常数据后，若(x 3-x 4)×(x 4-x 5)为负数，则按公式(1)重新计算中间变量u；
Figure PCTCN2021096654-appb-000001
若(x 3-x 4)×(x 4-x 5)不为负数，则计算表达式[(x 1-x 2)×(x 2-x 3)]的符号，若(x 1-x 2)×(x 2-x 3)为负数，则按公式(2)更新中间变量u，否则按公式(3)更新中间变量u；
Figure PCTCN2021096654-appb-000002
Figure PCTCN2021096654-appb-000003
更新中间变量u到输出变量y(m)，即y(m)＝u，将y(m)作为修复输出值更新至修复输出序列{y(m)}。
进一步的，所述方法还包括以下步骤：当前数据窗数据修复完成后移动原始采样值序列的读取指针和修复输出序列的读写指针，继续对采样数据进行处 理直至原始采样数据被处理完。
第二方面，本发明提供了一种采样数据异常值修复装置，所述装置包括：
数据窗模块：用于选取原始采样值序列中采样数据并结合修复输出序列中的采样数据构建数据窗；
识别模块：用于对数据窗内采样数据进行异常数据识别；
修复模块：用于结合数据窗内采样数据对疑似异常的数据进行修复。
进一步的，还包括采样模块：用于对信号进行等间隔采样，生成原始采样值序列。
第三方面，本发明还提供了一种采样数据异常值修复装置，包括处理器及存储介质和I/O接口设备；所述存储介质用于存储指令；所述I/O接口设备用于获取离线采集的或者在线采集的原始采样值序列；所述处理器用于根据所述指令进行操作以执行根据上述采样数据异常值修复方法的步骤。
第四方面，本发明上述还提供了计算机可读存储介质，其上存储有计算机程序，该程序被处理器执行时从I/O接口设备提取原始采样值序列，实现上述采样数据异常值修复方法的步骤。
与现有技术相比，本发明所达到的有益效果：
本发明通过选取原始采样值序列中采样数据并结合修复输出序列中的采样数据构建短数据窗，根据极值原理对数据窗内采样数据进行异常数据识别，并结合数据窗内采样数据对疑似异常的数据进行修复，实现了对采样异常数据的剔除，具有简单易实现，数据缓冲少、延迟小，恢复精度高，无需计算门槛值和运算量低、适合硬件加速等优点，具有良好的应用前景。
附图说明
图1是本发明的数据窗构造示意图；
图2是本发明的异常数据检测和修复流程图；
图3是采样异常数据案例1；
图4是采样异常数据案例2；
图5是采样异常数据案例3；
图6是本发明异常数据剔除和修复效果图。
具体实施方式
下面结合附图对本发明作进一步描述。以下实施例仅用于更加清楚地说明本发明的技术方案，而不能以此来限制本发明的保护范围。
本发明的目的在于克服现有技术中的不足，提供一种简单易实现，延迟小，恢复精度高，无需计算门槛值和运算量低适合硬件加速的采样数据异常值修复方法与装置，解决采样数据的快速异常检测和异常值修复技术问题。
实施例一：
本发明的采样数据异常值修复方法，首先利用连续的三点极值检测发现疑似异常采样数据，随后进一步检测邻近三点极值检测确认异常采样点，最多三次进行疑似异常值识别，再选取邻近正确采样数据进行数据修复。
下面将结合说明书附图，从技术原理和工程实施两个方面对本发明作进一步的说明。
1.技术基础
1.1.异常检测
工程上，异常采样数据是指采样值和系统实际值不符或误差超过一定的范围表现，就局部来说，表现为采样序列的局部极大值(或极小值)，如附图3、 附图4和附图5所示。而在数学上，函数极值点可能出现在间断点、边界点或导数为0点，常用判别方法如下：
若函数f(x)可导，f'(x 0)＝0，且
Figure PCTCN2021096654-appb-000004
对于
Figure PCTCN2021096654-appb-000005
有f'(x 0)＞0(或f'(x 0)＜0)同时
Figure PCTCN2021096654-appb-000006
有f'(x 0)＜0(或f'(x 0)＞0)，则x 0是f(x)的极大点(或极小点)，对应的f(x 0)为函数f(x)的极大值(或极小值)。
本发明根据上述数学原理，采用差分替代微分，将采样异常点检测转变为在连续信号的采样序列{x(n)}中寻找符合公式(4)采样数据点：
[x(n-1)-x(n)][x(n)-x(n+1)]＜0            (4)
进一步地，认为x(n)为这3点之中的极大值(或极小值)，同时也是疑似异常采样数据。
1.2.数据修复
根据二阶拉格朗日插值原理，已知信号x(t)在t 0,t 1,t 2时刻的值x(t 0),x(t 1),x(t 2)，对应时刻t 0≤t≤t 2，则信号在时刻t的采样值可以表示为：
x(t)＝k ax(t 0)+k bx(t 1)+k cx(t 2)              (5)
其中：
Figure PCTCN2021096654-appb-000007
Figure PCTCN2021096654-appb-000008
Figure PCTCN2021096654-appb-000009
进一步地，本发明在检测出异常采样数据后，剔除当前疑似异常采用数据，同时利用邻近的正常数据，通过公式(5)估计采样异常时刻的系统真实值，并以此替代异常点的异常采样数据，实现数据修复。
2.工程实施
2.1.数据窗构造
本发明实施案例基于5点的数据窗{x 1,x 2,x 3,,x 4,x 5}，其构造方法如附图1所示。从时间上来看，x 1最早，x 2,x 3,x 4次之，x 5最新。从数据来源看x 1,x 2来自修复输出序列{y(m)}，x 3,x 4,x 5对应原始采样序列{x(n)}，其中，x 5对应当前时刻的最新采样点x(n)，n为采样序列{x(n)}中当前最新采样点的序号，x 4对应原始采样序列中x(n)的上一采样节拍的值x(n-1)，x 3对应原始采样序列中x(n)的上两个采样节拍的值x(n-2)，而x 2对应当前时刻修复输出序列{y(m)}中上一节拍的修复采样数据y(m-1)，x 1对应修复输出序列{y(m)}中上两个采样节拍的采样数据y(m-2)，m为拟写入到修复输出序列{y(m)}的最新修复输出值序号。
2.2.异常检测和修复
设系统采样间隔为Ts，附图3～4给出具有一个采样异常点x3(位于时刻t)案例，其中图3a和图4a为单调下降的曲线中出现异常点，图3b和图4b为单调上升的曲线中出现异常点，而附图5a附图5b则分别对应图3a和图3b的移位。本发明根据异常点在数据窗的位置采取不同的异常点识别和修复方法。
场景(1)类似于图3a，在时刻t采样数据出现疑似异常，对应值为x 3，此时x 2,x 3,x 4构成的曲线段在t处出现极小值x 3，而以x 3,x 4,x 5构成的曲线段出现极大值为x 4。对于此种情况，本发明选取数据点x 1,x 2,x 5来估计x 3。
进一步地，参照公式5和附图3，记t 0＝0、t 1＝T s、t 2＝4T s，取t＝2T s，则计算k a、k b、k c得
Figure PCTCN2021096654-appb-000010
则疑似异常x 3的估算值
Figure PCTCN2021096654-appb-000011
可用下式计算：
Figure PCTCN2021096654-appb-000012
场景(2)类似于图5a，在时刻t采样数据出现疑似异常，对应采样值x 3,此时由x 2,x 3,x 4构成的曲线段在t处出现极大值x 3，同时由x 1,x 2,x 3构成的曲线段出现极小值x 2。对于此种情况，本发明选取数据点x 1,x 4,x 5来估计x 3。
进一步地参照公式(5)和附图5，记t 0＝0、t 1＝3T s、t 2＝4T s，取t＝2T s，则计算k a、k b、k c的
Figure PCTCN2021096654-appb-000013
则疑似异常x 3的估算值
Figure PCTCN2021096654-appb-000014
可用下式计算：
Figure PCTCN2021096654-appb-000015
场景(3)对于在时刻t采样数据出现疑似异常，此时由x 2,x 3,x 4构成的曲线段在t处出现极值x 3，同时由x 1,x 2,x 3构成的曲线段和x 3,x 4,x 5构成的曲线段均未出现极值，对于此种情况，本发明选取数据点x 1,x 2,x 4来估计x 3。
进一步地参照公式5和附图5，记t 0＝0、t 1＝T s、t 2＝3T s其t＝2T s，则计算k a、k b、k c的
Figure PCTCN2021096654-appb-000016
则疑似异常x 3的估算值
Figure PCTCN2021096654-appb-000017
可用下式计算：
Figure PCTCN2021096654-appb-000018
2.3.技术实现
附图1展示了本发明5点的数据窗{x 1,x 2,x 3,,x 4,x 5}的构造方法，附图2则展示了基于单个异常点的检测与修复流程，本发明的技术实施包括以下步骤：
步骤(1)，以间隔T s对信号进行连续采样，获得原始采样值序列{x(n)}；
步骤(2)，取信号原始采样值序列{x(n)}的最新三个采样点以及修复输出序列{y(m)}中的最新两个采样点，构成5个采样数据的数据窗，记为{x 1,x 2,x 3,x 4,x 5}，其中，x 5对应当前时刻的最新采样点x(n)，n为采样序列{x(n)}中当前最新采样点的序号，x 4对应原始采样序列中x(n)的上一采样节拍的值x(n-1)，x 3对应原始采 样序列中x(n)的上两个采样节拍的值x(n-2)，而x 2对应当前时刻修复输出序列{y(m)}中上一节拍的修复采样数据y(m-1)，x 1对应修复输出序列{y(m)}中上两个采样节拍的采样数据y(m-2)，m为拟写入到修复输出序列{y(m)}的最新修复输出值序号；
步骤(3),取u为中间变量并设初值为x 3，计算表达式[(x 2-x 3)×(x 3-x 4)]的符号，若sgn[(x 2-x 3)×(x 3-x 4)]为-1则转步骤(4),否则直接转步骤(6)，其中sgn[x]表示参数x的正负号，若参数为负则输出-1，若参数为正则输出+1，若参数为0则输出0；
步骤(4),计算表达式[(x 3-x 4)×(x 4-x 5)]的符号，若sgn[(x 3-x 4)×(x 4-x 5)]为-1，则先按公式(9)更新中间变量u后再转步骤(6)，否则直接转步骤(5)：
Figure PCTCN2021096654-appb-000019
步骤(5),计算表达式[(x 1-x 2)×(x 2-x 3)]的符号，若sgn[(x 1-x 2)×(x 2-x 3)]为-1，则先按公式(10)更新中间变量u后再转步骤(6)，否则按公式(11)更新中间变量后转步骤(6)；
Figure PCTCN2021096654-appb-000020
Figure PCTCN2021096654-appb-000021
步骤(6)，更新中间变量u到输出变量y(m)，即y(m)＝u，将y(m)作为修复输出值更新至修复输出序列{y(m)}，并转步骤(2)继续检查新的采样数据。
随着采样的进行，算法将获得原始采样序列的修复输出序列{y(m)},且修复后输出序列{y(m)}在时间上相比原始输入序列{x(n)}滞后两个采样间隔。
附图6展示了本发明一个异常检测与修复的效果，图6a为一个包含异常数 据的正弦采样序列，图6b为经过本发明的修复处理之后的正弦采样序列，进一步证明了本发明的实施效果。
综上所述，本发明结合原始采样序列{x(n)}和修复输出序列{y(m)}构建了5点数据窗，通过数据窗内连续3点间的差分符号变化来识别疑似异常点，并合理选择邻近的3个点通过插值完成疑似异常数据点修复，最多三次进行疑似异常值识别。本发明具有简单易实现，同时具有数据缓冲少，输出延迟小，异常数据恢复精度高，无需动态计算门槛，运算量低，适合硬件加速等优点，具有良好的应用前景。
实施例二：
本发明提供了一种采样数据异常值修复装置，能够用于实现实施例一所述的方法步骤，所述装置包括：
数据窗模块：用于选取原始采样值序列中采样数据并结合修复输出序列中的采样数据构建数据窗；
识别模块：用于对数据窗内采样数据进行异常数据识别；
修复模块：用于结合数据窗内采样数据对疑似异常的数据进行修复。
还包括采样模块：用于对信号进行等间隔采样，生成原始采样值序列；
实施例三：
本发明还提供了一种采样数据异常值修复装置，包括处理器、存储介质和和I/O接口设备；所述存储介质用于存储指令；所述I/O接口设备用于获取离线采集的或者在线采集的原始采样值序列；所述处理器用于根据所述指令进行操作以执行根据下述方法的步骤：
步骤(1)，通过I/O接口设备获取离线采集的原始采样值序列{x(n)}，或者以间隔T s在线对信号进行连续采样，获得原始采样值序列{x(n)}；
步骤(2)，取信号原始采样值序列{x(n)}的最新三个采样点以及修复输出序列{y(m)}中的最新两个采样点，构成5个采样数据的数据窗，记为{x 1,x 2,x 3,x 4,x 5}，其中，x 5对应当前时刻的最新采样点x(n)，n为采样序列{x(n)}中当前最新采样点的序号，x 4对应原始采样序列中x(n)的上一采样节拍的值x(n-1)，x 3对应原始采样序列中x(n)的上两个采样节拍的值x(n-2)，而x 2对应当前时刻修复输出序列{y(m)}中上一节拍的修复采样数据y(m-1)，x 1对应修复输出序列{y(m)}中上两个个采样节拍的采样数据y(m-2)，m为拟写入到修复输出序列{y(m)}的最新修复输出值序号；
步骤(3),取u为中间变量并设初值为x 3，计算表达式[(x 2-x 3)×(x 3-x 4)]的符号，若sgn[(x 2-x 3)×(x 3-x 4)]为-1则转步骤(4),否则直接转步骤(6)，其中sgn[x]表示参数x的正负号；
步骤(4),计算表达式[(x 3-x 4)×(x 4-x 5)]的符号，若sgn[(x 3-x 4)×(x 4-x 5)]为-1，则先按公式(12)更新中间变量u后再转步骤(6)，否则直接转步骤(5)：
Figure PCTCN2021096654-appb-000022
步骤(5),计算表达式[(x 1-x 2)×(x 2-x 3)]的符号，若sgn[(x 1-x 2)×(x 2-x 3)]为-1，则先按公式(13)更新中间变量u后再转步骤(6)，否则按公式(14)更新中间变量后转步骤(6)；
Figure PCTCN2021096654-appb-000023
Figure PCTCN2021096654-appb-000024
步骤(6)，更新中间变量u到输出变量y(m)，即y(m)＝u，将y(m)作为修复输出值更新至修复输出序列{y(m)}，并转步骤(2)继续检查新的采样数据。
随着采样的进行，算法将获得原始采样序列的修复输出序列{y(m)},且修复后输出序列{y(m)}在时间上相比原始输入序列{x(n)}滞后两个采样间隔。
实施例四：
本发明实施例还提供了一种计算机可读存储介质，其上存储有计算机程序，该程序被处理器执行时实现下述方法的步骤：
步骤(1)，从I/O接口设备提取原始采样值序列{x(n)}；
步骤(2)，取信号原始采样值序列{x(n)}的最新三个采样点以及修复输出序列{y(m)}中的最新两个采样点，构成5个采样数据的数据窗，记为{x 1,x 2,x 3,x 4,x 5}，其中，x 5对应当前时刻的最新采样点x(n)，n为采样序列{x(n)}中当前最新采样点的序号，x 4对应原始采样序列中x(n)的上一采样节拍的值x(n-1)，x 3对应原始采样序列中x(n)的上两个采样节拍的值x(n-2)，而x 2对应当前时刻修复输出序列{y(m)}中上一节拍的修复采样数据y(m-1)，x 1对应修复输出序列{y(m)}中上两个采样节拍的采样数据y(m-2)，m为拟写入到修复输出序列{y(m)}的最新修复输出值序号；
步骤(3),取u为中间变量并设初值为x 3，计算表达式[(x 2-x 3)×(x 3-x 4)]的符号，若sgn[(x 2-x 3)×(x 3-x 4)]为-1则转步骤(4),否则直接转步骤(6)，其中sgn[x]表示参数x的正负号；
步骤(4),计算表达式[(x 3-x 4)×(x 4-x 5)]的符号，若sgn[(x 3-x 4)×(x 4-x 5)]为-1，则先按公式(15)更新中间变量u后再转步骤(6)，否则直接转步骤(5)：
Figure PCTCN2021096654-appb-000025
步骤(5),计算表达式[(x 1-x 2)×(x 2-x 3)]的符号，若sgn[(x 1-x 2)×(x 2-x 3)]为-1，则先按公式(16)更新中间变量u后再转步骤(6)，否则按公式(17)更新中间变量后转步骤(6)；
Figure PCTCN2021096654-appb-000026
Figure PCTCN2021096654-appb-000027
步骤(6)，更新中间变量u到输出变量y(m)，即y(m)＝u，将y(m)作为当前修复输出值更新至修复输出序列{y(m)}，并转步骤(2)继续检查新的采样数据。
随着原始采样数据的读入，算法将获得原始采样序列的修复输出序列{y(m)},且修复后输出序列{y(m)}在时间上相比原始输入序列{x(n)}滞后两个采样间隔。
本领域内的技术人员应明白，本申请的实施例可提供为方法、系统、或计算机程序产品。因此，本申请可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本申请可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本申请是参照根据本申请实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
以上所述仅是本发明的优选实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本发明技术原理的前提下，还可以做出若干改进和变形，这些改进和变形也应视为本发明的保护范围。

Claims (10)

  1. 一种采样数据异常值修复方法，其特征在于，所述方法包括以下步骤：
    根据原始采样值序列中采样数据并结合修复输出序列中的采样数据构建数据窗；
    根据函数极值原理对数据窗内采样数据进行异常数据识别；
    通过数据窗内的采样数据对异常数据进行修复。
  2. 根据权利要求1所述的一种采样数据异常值修复方法，其特征在于，所述方法还包括以下步骤：
    对信号进行等间隔连续采样，获得原始采样值序列{x(n)}。
  3. 根据权利要求1所述的一种采样数据异常值修复方法，其特征在于，所述数据窗的构建方法包括以下步骤：
    取原始采样值序列{x(n)}的最新三个采样点以及修复输出序列{y(m)}中的最新两个采样点，构成5个采样数据的数据窗，记为{x 1,x 2,x 3,x 4,x 5}，其中，x 5对应当前时刻的最新采样点x(n)，n为采样序列{x(n)}中当前最新采样点的序号，x 4对应原始采样序列中x(n)的上一采样节拍的值x(n-1)，x 3对应原始采样序列中x(n)的上两个采样节拍的值x(n-2)，而x 2对应当前时刻修复输出序列{y(m)}中上一节拍的修复采样数据y(m-1)，x 1对应修复输出序列{y(m)}中上两个采样节拍的修复采样数据y(m-2)，m为拟写入到修复输出序列{y(m)}的最新修复输出值序号。
  4. 根据权利要求3所述的一种采样数据异常值修复方法，其特征在于，识别异常数据的方法包括以下步骤：
    若(x 2-x 3)×(x 3-x 4)为负数，则认为所述数据窗内采样值异常，否则判别所述数据窗内采样值正常，并把x 3更新到输出变量y(m)，即y(m)＝x 3，同时将y(m)更新 至修复输出序列{y(m)}。
  5. 根据权利要求4所述的一种采样数据异常值修复方法，其特征在于，异常数据的修复的方法包括如下步骤：
    取中间变量u并设置初值为x 3，
    发现异常数据后，若(x 3-x 4)×(x 4-x 5)为负数，则按公式(1)重新计算中间变量u；
    Figure PCTCN2021096654-appb-100001
    若(x 3-x 4)×(x 4-x 5)不为负数，则计算表达式[(x 1-x 2)×(x 2-x 3)]的符号，若(x 1-x 2)×(x 2-x 3)为负数，则按公式(2)更新中间变量u，否则按公式(3)更新中间变量u；
    Figure PCTCN2021096654-appb-100002
    Figure PCTCN2021096654-appb-100003
    更新中间变量u到输出变量y(m)，即y(m)＝u，将y(m)作为修复输出值更新至修复输出序列{y(m)}。
  6. 根据权利要求1所述的一种采样数据异常值修复方法，其特征在于，所述方法还包括以下步骤：当前数据窗数据修复完成后移动原始采样值序列的读取指针和修复输出序列的读写指针，继续对采样数据进行处理直至原始采样数据被处理完。
  7. 一种采样数据异常值修复装置，其特征在于，所述装置包括：
    数据窗模块：用于选取原始采样值序列中采样数据并结合修复输出序列中的采样数据构建数据窗；
    识别模块：用于对数据窗内采样数据进行异常数据识别；
    修复模块：用于结合数据窗内采样数据对疑似异常的数据进行修复。
  8. 根据权利要求7所述的一种采样数据异常值修复装置，其特征在于，还包括采样模块：用于对信号进行等间隔采样，生成原始采样值序列。
  9. 一种采样数据异常值修复装置，其特征在于，包括处理器及存储介质和I/O接口设备；
    所述存储介质用于存储指令；
    所述I/O接口设备用于获取离线采集的或者在线采集的原始采样值序列；
    所述处理器用于根据所述指令进行操作，从I/O接口设备提取原始采样值序列，以执行根据权利要求1～6任一项所述方法的步骤。
  10. 计算机可读存储介质，其上存储有计算机程序，其特征在于，该程序被处理器执行时实现权利要求1～6任一项所述方法的步骤。
PCT/CN2021/096654 2020-08-17 2021-05-28 一种采样数据异常值修复方法及装置 WO2022037172A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010825283.9A CN112000519B (zh) 2020-08-17 2020-08-17 一种采样数据异常值修复方法及装置
CN202010825283.9 2020-08-17

Publications (1)

Publication Number Publication Date
WO2022037172A1 true WO2022037172A1 (zh) 2022-02-24

Family

ID=73472537

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/096654 WO2022037172A1 (zh) 2020-08-17 2021-05-28 一种采样数据异常值修复方法及装置

Country Status (3)

Country Link
CN (1) CN112000519B (zh)
WO (1) WO2022037172A1 (zh)
ZA (1) ZA202106321B (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117951455A (zh) * 2024-03-22 2024-04-30 汶上义桥煤矿有限责任公司 一种刮板输送机运行故障在线监测方法

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112000519B (zh) * 2020-08-17 2023-11-28 南京磐能电力科技股份有限公司 一种采样数据异常值修复方法及装置
CN112905958B (zh) * 2021-01-27 2024-04-19 南京国电南自电网自动化有限公司 基于测控装置的短时数据窗遥测数据状态辨识方法及系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105573853A (zh) * 2015-12-18 2016-05-11 国电南瑞科技股份有限公司 基于双重化数据窗的采样异常数据处理方法
CN107657288A (zh) * 2017-10-26 2018-02-02 国网冀北电力有限公司 一种基于孤立森林算法的电力调度流数据异常检测方法
CN110244099A (zh) * 2019-06-24 2019-09-17 河南工业大学 基于用户电压的窃电检测方法
CN112000519A (zh) * 2020-08-17 2020-11-27 南京磐能电力科技股份有限公司 一种采样数据异常值修复方法及装置

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5465321A (en) * 1993-04-07 1995-11-07 The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration Hidden markov models for fault detection in dynamic systems
CN102707132B (zh) * 2012-05-21 2014-07-02 国电南瑞科技股份有限公司 一种数字化保护测控装置异常采样值动态处理方法
CN105138843A (zh) * 2015-08-26 2015-12-09 长园深瑞继保自动化有限公司 电力系统采样飞点检测及其修复方法
CN110211099A (zh) * 2019-05-18 2019-09-06 安徽大学 综合麦穗正反面病情的小麦赤霉病检测方法及装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105573853A (zh) * 2015-12-18 2016-05-11 国电南瑞科技股份有限公司 基于双重化数据窗的采样异常数据处理方法
CN107657288A (zh) * 2017-10-26 2018-02-02 国网冀北电力有限公司 一种基于孤立森林算法的电力调度流数据异常检测方法
CN110244099A (zh) * 2019-06-24 2019-09-17 河南工业大学 基于用户电压的窃电检测方法
CN112000519A (zh) * 2020-08-17 2020-11-27 南京磐能电力科技股份有限公司 一种采样数据异常值修复方法及装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
WU WENJU, JIEWEN ZHAN: "A flying spot detection algorithm for continuous sampling based on removable discontinuous points", DIANLI XITONG BAOHU YU KONGZHI/POWER SYSTEM PROTECTION AND CONTROL, vol. 43, no. 4, 16 February 2015 (2015-02-16), pages 18 - 24, XP055901443, ISSN: 1674-3415 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117951455A (zh) * 2024-03-22 2024-04-30 汶上义桥煤矿有限责任公司 一种刮板输送机运行故障在线监测方法
CN117951455B (zh) * 2024-03-22 2024-06-07 汶上义桥煤矿有限责任公司 一种刮板输送机运行故障在线监测方法

Also Published As

Publication number Publication date
CN112000519A (zh) 2020-11-27
CN112000519B (zh) 2023-11-28
ZA202106321B (en) 2021-10-27

Similar Documents

Publication Publication Date Title
WO2022037172A1 (zh) 一种采样数据异常值修复方法及装置
JP6050754B2 (ja) 信号処理に基づく故障検出、分離、及び修正のためのシステム、方法、及び装置
US8954294B2 (en) Measuring device and measuring method for continuous physical quantity
CA2931624A1 (en) Systems and methods for event detection and diagnosis
CN107978147B (zh) 一种基于knn算法的交通流异常数据双向检测修复方法
US20180191241A1 (en) A Method for Correcting Effect of Saturation in Current Transformer and an Intelligent Electronic Device Therefor
CN110147827B (zh) 一种基于iaalo-svm与相似性度量的故障预测方法
Han et al. Receding-horizon unbiased FIR filters for continuous-time state-space models without a priori initial state information
CN115800272B (zh) 基于拓扑识别的电网故障分析方法、系统、终端及介质
CN112235043A (zh) 一种基于自适应长短期记忆的分布式光纤异常数据修复模型
Andrieu et al. Continuous discrete observer with updated sampling period
CN110134676B (zh) 一种传感器数据质量的监控方法
CN102571236A (zh) 最大时间间隔误差的测量方法和系统
CN110677140A (zh) 一种含未知输入和非高斯量测噪声的随机系统滤波器
CN109186571B (zh) 一种陀螺仪滤波去噪方法
CN114153788A (zh) 一种执行器的可追溯控制方法、执行器和控制系统
JP5219783B2 (ja) 不正アクセス検知装置及び不正アクセス検知プログラム及び記録媒体及び不正アクセス検知方法
CN109462242B (zh) 基于iir数字滤波和esprit辨识算法的电力系统低频振荡检测方法
CN112965966B (zh) 一种基于实测飞参数据的快速预处理方法、系统及计算机相关产品
CN115660957A (zh) 一种波形数据的重采样方法、装置、设备及介质
CN113139158B (zh) 基于高斯过程回归的comtrade异常录波数据监测和修正方法及系统
CN102435209B (zh) 一种消除倾角传感器信号基线漂移的方法
CN112965964A (zh) 一种实测飞参数据的野值检测方法、系统及计算机相关产品
CN112711081B (zh) 基于偏航误差检测极端阵风的方法及装置
CN114781760B (zh) 一种基于大数据的故障预测方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21857273

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21857273

Country of ref document: EP

Kind code of ref document: A1