Fault detection optimization method for honeycomb-shaped smart power grid

Abstract

The invention relates to the technical field of power grid fault positioning, in particular to a fault detection optimization method for a honeycomb-shaped smart power grid. The method comprises the following steps: acquiring an original data sequence in a smart power grid, decomposing the original data sequence by using a time sequence decomposition algorithm to obtain an initial residual component sequence, and determining the residual component sequence according to the data distribution characteristics in the initial residual component sequence and the data difference between the initial residual component sequence and the original data sequence. Obtaining the credibility of the initial residual component sequence; if the credibility of the initial residual component sequence is not within a preset credibility range, optimizing preset parameters in a time sequence decomposition algorithm until the credibility of the new residual component sequence is within the preset credibility range, and obtaining a target residual component sequence; and finally, performing fault detection on the smart grid according to the target residual component sequence. By optimizing the preset parameters in the time sequence decomposition algorithm, the fault detection accuracy of the residual component is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid fault location, and in particular to a fault detection optimization method for a honeycomb smart grid. Background Art

[0002] The honeycomb smart grid is a distributed, modular power grid system that uses the geometric characteristics of the honeycomb to improve the reliability and flexibility of the power grid. Each honeycomb unit is equipped with intelligent sensors and monitoring systems that can collect and analyze data in real time, use advanced algorithms to detect faults, and optimize system performance and resource utilization. The honeycomb smart grid performs fault detection to ensure power supply stability and grid security, while reducing maintenance and restoration costs, and providing early warnings to improve the overall reliability of the grid and ensure efficient, safe and reliable operation of the power system to meet social and economic needs for power supply.

[0003] The traditional fault detection method of honeycomb smart grid usually uses the smart sensors equipped in the honeycomb units to collect data in real time, and then obtains the residuals through STL decomposition. The residuals after STL decomposition are used for fault detection to identify abnormal fluctuations or emergencies. In the STL decomposition process, in order to improve the accuracy and reliability of time series data analysis, it is necessary to correct the seasonal component so that data analysis and model construction can focus on non-seasonal components, thereby providing more accurate insights to improve the effect of subsequent fault detection. However, insufficient or excessive correction of seasonal components will indirectly affect the residuals, causing periodic patterns to be mixed into the residuals, making abnormal signals mixed in periodic changes, or erasing important seasonal information, resulting in increased noise, which will reduce the accuracy of residuals for fault detection.

[0004] Therefore, how to improve the correction effect of seasonal components and thus improve the accuracy of fault detection in cellular smart grids has become an urgent problem to be solved. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a fault detection optimization method for a cellular smart grid to solve the problem of how to improve the correction effect of the seasonal component and thereby improve the accuracy of fault detection in the cellular smart grid.

[0006] An embodiment of the present invention provides a fault detection optimization method for a cellular smart grid, the method comprising the following steps: Acquire any type of grid time series data in the smart grid, preprocess the grid time series data, and obtain an original data sequence; Decomposing the original data sequence by using a time series decomposition algorithm to obtain an initial trend component sequence and an initial seasonal component sequence, obtaining a corresponding initial residual component sequence according to the initial trend component sequence and the initial seasonal component sequence, and obtaining the credibility of the initial residual component sequence according to the data distribution characteristics in the initial residual component sequence and the data difference between the initial residual component sequence and the original data sequence; If the credibility of the initial residual component sequence is not within the preset credibility range, the preset parameters in the time series decomposition algorithm are optimized to obtain a new trend component sequence and a new seasonal component sequence corresponding to the original data sequence, and a corresponding new residual component sequence is obtained according to the new trend component sequence and the new seasonal component sequence, and the credibility acquisition method is repeated until the credibility of the new residual component sequence is within the preset credibility range, thereby obtaining a target residual component sequence; According to the target residual component sequence, fault detection is performed on the smart grid.

[0007] Preferably, obtaining the credibility of the initial residual component sequence according to the data distribution characteristics in the initial residual component sequence and the data difference between the initial residual component sequence and the original data sequence includes: According to the data distribution characteristics in the initial residual component sequence, obtaining the normal distribution degree of the initial residual component sequence; According to the data difference between the initial residual component sequence and the original data sequence, obtaining the degree of difference between the initial residual component sequence and the original data sequence; The credibility of the initial residual component sequence is obtained according to the normal distribution degree and the difference degree.

[0008] Preferably, obtaining the normal distribution degree of the initial residual component sequence according to the data distribution characteristics in the initial residual component sequence includes: Obtain the data mean of the initial residual component sequence, obtain the number of data in the initial residual component sequence that is within a preset normal distribution area, calculate the ratio of the number of data in the initial residual component sequence that is within a preset normal distribution area to the number of data in the initial residual component sequence to obtain a first ratio, and obtain the sum of the absolute value of the data mean of the initial residual component sequence and the first ratio as the degree of normal distribution of the initial residual component sequence.

[0009] Preferably, the obtaining the degree of difference between the initial residual component sequence and the original data sequence according to the data difference between the initial residual component sequence and the original data sequence comprises: Fitting the initial residual component sequence to obtain a residual component fitting curve, obtaining an original data curve composed of the original data sequence, obtaining each extreme point in the original data curve, and using each extreme point in the original data curve to divide the original data curve and the residual component fitting curve into at least two sub-curves, respectively, and the sub-curves of the original data curve and the sub-curves of the residual component fitting curve correspond to each other one by one; The curvature of each sub-curve in the original data curve and the residual component fitting curve is obtained, and the absolute value of the curvature difference between each sub-curve in the original data curve and the corresponding sub-curve in the residual component fitting curve is calculated respectively, and the accumulated value of the absolute value is obtained accordingly, and the accumulated value of the absolute value is used as the independent variable of the sigmoid function to obtain the corresponding sigmoid function value as the difference degree between the initial residual component sequence and the original data sequence.

[0010] Preferably, obtaining the credibility of the initial residual component sequence according to the normal distribution degree and the difference degree includes: The normal distribution degree and the difference degree are weighted and summed, and the obtained weighted summation result is used as the credibility of the initial residual component sequence.

[0011] Preferably, the preset parameters in the time series decomposition algorithm include: a seasonal smoothing window and a seasonal cycle.

[0012] Preferably, if the credibility of the initial residual component sequence is not within a preset credibility range, optimizing the preset parameters in the time series decomposition algorithm includes: If the credibility of the initial residual component sequence is not within the preset credibility range, the length of the seasonal smoothing window is increased or decreased within the preset length range of the seasonal smoothing window, and the length of the seasonal cycle is increased or decreased within the preset length range of the seasonal cycle.

[0013] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: The present invention obtains any type of grid time series data in a smart grid, pre-processes the grid time series data, and obtains an original data sequence; decomposes the original data sequence using a time series decomposition algorithm to obtain an initial trend component sequence and an initial seasonal component sequence; obtains a corresponding initial residual component sequence according to the initial trend component sequence and the initial seasonal component sequence; obtains the credibility of the initial residual component sequence according to the data distribution characteristics in the initial residual component sequence and the data difference between the initial residual component sequence and the original data sequence; if the credibility of the initial residual component sequence is not within a preset credibility range, optimizes the preset parameters in the time series decomposition algorithm to obtain a new trend component sequence and a new seasonal component sequence corresponding to the original data sequence; obtains a corresponding new residual component sequence according to the new trend component sequence and the new seasonal component sequence; repeats the credibility acquisition method until the credibility of the new residual component sequence is within the preset credibility range, and obtains a target residual component sequence; and performs fault detection on the smart grid according to the target residual component sequence. The present invention first obtains the credibility of the initial residual component sequence. For the initial residual component sequence with low credibility, the preset parameters in the time series decomposition algorithm are optimized to correct the initial seasonal component sequence, thereby improving the correction effect of the seasonal component sequence, thereby improving the credibility of the target residual component sequence, reducing the problem of low credibility of the residual component caused by insufficient or excessive correction of the seasonal component sequence, and improving the accuracy of residual components in fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0015] Figure 1 It is a method flow chart of a fault detection optimization method for a cellular smart grid provided in Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0016] Embodiments of the present disclosure are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, but should not be understood as limiting the present disclosure.

[0017] It should be noted that the terms "first", "second", etc. in the specification of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure.

[0018] In order to illustrate the technical solution of the present invention, specific embodiments are provided below for illustration.

[0019] See also Figure 1 , is a method flow chart of a fault detection optimization method for a cellular smart grid provided in Embodiment 1 of the present invention, such as Figure 1 As shown, the method may include: Step S101, obtaining any type of grid time series data in the smart grid, and preprocessing the grid time series data to obtain an original data sequence.

[0020] Obtain different types of grid data in the honeycomb smart grid, such as load data, grid status data, equipment data, etc. Grid data is collected once a day. Each type of grid data is collected within a year. There is no restriction here. It can be set according to the specific implementation scenario to form grid time series data corresponding to each type of grid data.

[0021] For any type of grid time series data in the honeycomb smart grid, check whether there are duplicate values ​​and missing values ​​in the grid time series data, delete the duplicate values, fill in the missing values ​​or delete the grid time series data corresponding to the missing values ​​to obtain the original data sequence.

[0022] Step S102, using a time series decomposition algorithm to decompose the original data sequence to obtain an initial trend component sequence and an initial seasonal component sequence, and obtaining a corresponding initial residual component sequence based on the initial trend component sequence and the initial seasonal component sequence, and obtaining the credibility of the initial residual component sequence based on the data distribution characteristics in the initial residual component sequence and the data difference between the initial residual component sequence and the original data sequence.

[0023] The STL time series decomposition algorithm is a powerful tool for processing time series data. It can process different types of complex seasonal data, has strong robustness to outliers, has a clear decomposition structure, and its results can be displayed by visualization. It is very suitable for processing various types of power grid data. Since the STL time series decomposition algorithm can decompose time series data into trend components, seasonal components, and residual components, by analyzing the residual components after STL decomposition, abnormal fluctuations or emergencies in time series data can be effectively identified. Therefore, after obtaining the original data sequence, the STL time series decomposition algorithm is used to decompose the original data sequence to obtain the initial trend component sequence and the initial seasonal component sequence. According to the initial trend component sequence and the initial seasonal component sequence, the corresponding initial residual component sequence is obtained. The STL time series decomposition algorithm belongs to the prior art and will not be repeated here.

[0024] In one embodiment, the parameters of the STL time series decomposition algorithm are set, and the number of observation points in the seasonal cycle is set to 90, the length of the seasonal smoothing window is 23, the length of the trend smoothing window is 21, the step size of the seasonal smoothing window is 13, and the step size of the trend smoothing window is 11. There is no limitation here. According to the specific implementation scenario setting, the original data sequence can be decomposed by the STL time series decomposition algorithm to obtain an initial trend component sequence and an initial seasonal component sequence. According to the initial trend component sequence and the initial seasonal component sequence, the corresponding initial residual component sequence is obtained.

[0025] Because the initial seasonal component sequence will introduce periodic fluctuations, masking the long-term trends and abnormal signals in the data, resulting in the low credibility of the obtained initial residual component sequence and inaccurate final fault detection results, it is necessary to correct the initial seasonal component sequence to reduce the impact of these periodic fluctuations on data analysis. At the same time, it can also make the data more stable and improve the accuracy of the prediction model.

[0026] In the process of correcting the initial seasonal component sequence, since the correction of the initial seasonal component sequence changes the smoothness and detail capturing ability of the initial seasonal component sequence, if the initial seasonal component sequence is corrected insufficiently or excessively, the part of the original data sequence allocated to the initial seasonal component sequence will change, causing the initial residual component sequence to change accordingly. For example, periodic patterns may be mixed into the initial residual component sequence, causing abnormal signals to be mixed with periodic changes, or important seasonal information may be erased, resulting in an increase in noise. These may cause the data in the initial residual component sequence to have non-normal characteristics, be too similar to the original data sequence, and other characteristics, seriously affecting the accuracy of subsequent fault detection.

[0027] Therefore, it is necessary to analyze the data characteristics of the initial residual component sequence. Specifically, the credibility of the initial residual component sequence is obtained according to the data distribution characteristics in the initial residual component sequence and the data difference between the initial residual component sequence and the original data sequence. According to the credibility of the initial residual component sequence, the initial seasonal component sequence is corrected to ensure the correction effect of the initial seasonal component sequence, thereby ensuring the accuracy of subsequent fault detection.

[0028] The method for obtaining the credibility of the initial residual component sequence includes: (1) According to the data distribution characteristics in the initial residual component sequence, the normal distribution degree of the initial residual component sequence is obtained.

[0029] Specifically, the data mean of the initial residual component sequence is obtained, the number of data in the initial residual component sequence that is within a preset normal distribution area is obtained, the ratio of the number of data in the initial residual component sequence that is within a preset normal distribution area to the number of data in the initial residual component sequence is calculated to obtain a first ratio, and the absolute value of the data mean of the initial residual component sequence and the sum of the first ratio are obtained as the degree of normal distribution of the initial residual component sequence.

[0030] In one embodiment, the degree of normal distribution of the initial residual component sequence is calculated:

[0031] in, is the normal distribution degree of the initial residual component sequence; is the i-th data in the initial residual component sequence; i is the sequence number of the data in the initial residual component sequence; n is the number of data in the initial residual component sequence; It is the number of data in the initial residual component sequence under normal distribution within the range of ±3 standard deviations of the mean of the initial residual component sequence. There is no restriction here and it can be set according to the specific implementation scenario.

[0032] It should be noted that is the data mean of the initial residual component sequence. Under normal circumstances, the initial residual component sequence will present an approximate normal distribution with a mean of 0. The more data in the initial residual component sequence under normal distribution are within the range of ±3 standard deviations of the mean of the initial residual component sequence, the more the initial residual component sequence conforms to the normal distribution, and the higher the degree of normal distribution of the initial residual component sequence (i.e. The closer it is to 1), the higher the accuracy of subsequent fault detection.

[0033] (2) Obtaining a degree of difference between the initial residual component sequence and the original data sequence according to a data difference between the initial residual component sequence and the original data sequence.

[0034] Specifically, the initial residual component sequence is fitted to obtain a residual component fitting curve, an original data curve composed of the original data sequence is obtained, each extreme point in the original data curve is obtained, and each extreme point in the original data curve is used to divide the original data curve and the residual component fitting curve into at least two sub-curves, and the sub-curves of the original data curve and the sub-curves of the residual component fitting curve correspond to each other one by one; The curvature of each sub-curve in the original data curve and the residual component fitting curve is obtained, and the absolute value of the curvature difference between each sub-curve in the original data curve and the corresponding sub-curve in the residual component fitting curve is calculated respectively, and the accumulated value of the absolute value is obtained accordingly, and the accumulated value of the absolute value is used as the independent variable of the sigmoid function to obtain the corresponding sigmoid function value as the difference degree between the initial residual component sequence and the original data sequence.

[0035] In one embodiment, an original data curve composed of an original data sequence is obtained, and each extreme point in the original data curve is obtained. The extreme point belongs to the prior art and will not be described in detail here. The initial residual component sequence is fitted to obtain a residual component fitting curve. According to the time points corresponding to the extreme points in the original data curve, the curve between each two adjacent time points constitutes a sub-curve. The original data curve and the residual component fitting curve are respectively divided into t sub-curves. The sub-curves of the original data curve and the sub-curves of the residual component fitting curve correspond one to one. The curvature of each sub-curve in the original data curve and the residual component fitting curve is obtained. The curvature belongs to the prior art and will not be described in detail here. The degree of difference between the initial residual component sequence and the original data sequence is calculated:

[0036] in, is the degree of difference between the initial residual component sequence and the original data sequence; is the curvature of the u-th sub-curve among the sub-curves divided by the original data curve; is the curvature of the u-th sub-curve in the sub-curve divided by the residual component fitting curve; F is the original data curve; f is the residual component fitting curve; u is the sequence number of the sub-curve in the original data curve and the residual component fitting curve; t is the number of sub-curves in the original data curve and the residual component fitting curve; Sig() is an S-type function, also known as an S-type growth curve, which can map the output result to (0, 1); | | is the absolute value symbol.

[0037] It should be noted that It represents the cumulative value of the absolute value of the curvature difference between each segment of the original data curve and the residual component fitting curve. The greater the difference in curvature between each segment of the original data curve and the residual component fitting curve, The larger it is, the greater the difference between the initial residual component sequence and the original data sequence, which means that the original data curve and the residual component fitting curve are less matched, the initial residual component sequence is less affected by the initial seasonal component sequence, and the subsequent fault detection accuracy is higher.

[0038] (3) Obtaining the credibility of the initial residual component sequence based on the degree of normal distribution and the degree of difference.

[0039] Specifically, a weighted summation process is performed on the normal distribution degree and the difference degree, and the obtained weighted summation result is used as the credibility of the initial residual component sequence.

[0040] In one embodiment, the credibility of the initial residual component sequence is calculated:

[0041] in, is the credibility of the initial residual component sequence; is the normal distribution degree of the initial residual component sequence; is the degree of difference between the initial residual component sequence and the original data sequence; is the weight coefficient of the normal distribution degree of the initial residual component sequence, ; is the weight coefficient of the difference between the initial residual component sequence and the original data sequence, ; There is no restriction here and it can be set according to the specific implementation scenario.

[0042] It should be noted that the higher the normal distribution degree of the initial residual component sequence, the higher the credibility of the initial residual component sequence, and the higher the accuracy of subsequent fault detection; the greater the difference between the initial residual component sequence and the original data sequence, the less the initial residual component sequence is affected by the initial seasonal component sequence, the higher the credibility of the initial residual component sequence, and the higher the accuracy of subsequent fault detection.

[0043] At this point, the initial trend component sequence, initial seasonal component sequence and initial residual component sequence after decomposition by the STL time series decomposition algorithm are obtained, as well as the credibility of the initial residual component sequence.

[0044] Step S103: If the credibility of the initial residual component sequence is not within the preset credibility range, the preset parameters in the time series decomposition algorithm are optimized to obtain a new trend component sequence and a new seasonal component sequence corresponding to the original data sequence, and a corresponding new residual component sequence is obtained based on the new trend component sequence and the new seasonal component sequence. The credibility acquisition method is repeated until the credibility of the new residual component sequence is within the preset credibility range to obtain the target residual component sequence.

[0045] The higher the credibility of the initial residual component sequence, the higher the accuracy of the initial residual component sequence in fault detection. The credibility range is set to [0.65, 1]. There is no restriction here. It can be set according to the specific implementation scenario. When the credibility of the initial residual component sequence is within [0.65, 1], the initial residual component sequence is recorded as the target residual component sequence. When the target residual component sequence is used for fault detection, it is sufficient to ensure the accuracy of fault detection.

[0046] If the credibility of the initial residual component sequence is not within [0.65, 1], the normal fluctuations in the initial residual component sequence will be mistakenly marked as abnormal, masking the actual abnormal distribution of the data and affecting the accuracy of fault detection of the initial residual component sequence. Therefore, it is necessary to optimize the preset parameters in the STL time series decomposition algorithm and re-correct the initial seasonal component sequence to obtain a residual component sequence whose credibility is within the credibility range.

[0047] Among them, optimizing the preset parameters in the STL time series decomposition algorithm refers to optimizing the seasonal smoothing window and seasonal cycle in the STL time series decomposition algorithm. The specific optimization method is: If the credibility of the initial residual component sequence is not within the preset credibility range, the length of the seasonal smoothing window is increased or decreased within the preset length range of the seasonal smoothing window, and the length of the seasonal cycle is increased or decreased within the preset length range of the seasonal cycle.

[0048] In one embodiment, the length range of the seasonal smoothing window is set to [3, 43]. If the credibility of the initial residual component sequence is not within [0.65, 1], 0.35 is used as the dividing point, and the credibility range not within [0.65, 1], that is, (0, 0.65), is divided into a first range and a second range, [0.35, 0.65) is the first range, and (0, 0.35) is the second range. When the credibility of the initial residual component sequence is within the first range, the length of the seasonal smoothing window is increased or decreased by 2 each time based on 23; when the credibility of the initial residual component sequence is within the second range, the length of the seasonal smoothing window is increased or decreased by 4 each time based on 23. There is no limitation here and it can be set according to the specific implementation scenario.

[0049] The length range of the seasonal cycle is set to [80, 100]. If the credibility of the initial residual component sequence is not within [0.65, 1], it is adjusted within the length range of the seasonal cycle [80, 100]. The number of observation points in the seasonal cycle is increased or decreased by 1 each time based on 90. There is no restriction here and it can be set according to the specific implementation scenario.

[0050] For example: assuming that the seasonal smoothing window is a sliding window including 23 days of data and the seasonal cycle is 90 days of data, if the credibility of the initial residual component sequence is not in [0.65, 1], when the credibility of the initial residual component sequence is in [0.35, 0.65), the seasonal smoothing window is adjusted to a sliding window including 21 days or 25 days of data, and the seasonal cycle is adjusted to 89 days or 91 days of data; if the credibility of the initial residual component sequence is less than 0.35, the seasonal smoothing window is adjusted to a sliding window including 19 days or 27 days of data, and the seasonal cycle is adjusted to 89 days or 91 days of data.

[0051] After optimizing the preset parameters in the time series decomposition algorithm, the original data sequence is re-decomposed by STL using the optimized seasonal smoothing window and seasonal cycle to obtain a new trend component sequence and a new seasonal component sequence corresponding to the original data sequence. According to the new trend component sequence and the new seasonal component sequence, the corresponding new residual component sequence is obtained. Referring to the method for obtaining the credibility of the initial residual component sequence, the credibility of the new residual component sequence is obtained. If the credibility of the new residual component sequence is not within [0.65, 1], the preset parameters in the time series decomposition algorithm are optimized again, and so on, until the credibility of the new residual component sequence is within [0.65, 1], and the target residual component sequence is obtained.

[0052] At this point, a target residual component sequence whose credibility is within a preset credibility range is obtained.

[0053] Step S104: performing fault detection on the smart grid according to the target residual component sequence.

[0054] After obtaining the target residual component sequence whose credibility is within the preset credibility range, since the credibility of the target residual component sequence is within the preset credibility range, the accuracy of fault detection is sufficient to be guaranteed when the target residual component sequence is used for fault detection.

[0055] According to the above steps, the target residual component sequence of the power grid time series data is obtained, and the cellular smart grid is detected according to the target residual component sequence in the power grid time series data. The use of the target residual component sequence for fault detection belongs to the existing technology and will not be repeated here. The detected faults are classified according to the type of power grid time series data, such as equipment failure, load abnormality, communication problem, etc. The fault results are verified by cross-validation and other methods to ensure the accuracy of the detection. Then, on-site investigation or remote diagnosis is performed to check the equipment status to confirm the location of the fault area. At the same time, repair measures and emergency response procedures are taken to isolate, restore power supply or start the backup system, and a fault detection and processing report is written. The detection algorithm, system configuration or maintenance process is optimized according to the results and feedback to improve the safety and reliability of cellular smart grid fault detection.

[0056] In summary, this embodiment obtains any type of grid time series data in a smart grid, preprocesses the grid time series data, and obtains an original data sequence; decomposes the original data sequence using a time series decomposition algorithm to obtain an initial trend component sequence and an initial seasonal component sequence; obtains a corresponding initial residual component sequence according to the initial trend component sequence and the initial seasonal component sequence; obtains the credibility of the initial residual component sequence according to the data distribution characteristics in the initial residual component sequence and the data difference between the initial residual component sequence and the original data sequence; if the credibility of the initial residual component sequence is not within a preset credibility range, optimize the preset parameters in the time series decomposition algorithm to obtain a new trend component sequence and a new seasonal component sequence corresponding to the original data sequence; obtain a corresponding new residual component sequence according to the new trend component sequence and the new seasonal component sequence, and repeat the credibility acquisition method until the credibility of the new residual component sequence is within a preset credibility range to obtain a target residual component sequence; and perform fault detection on the smart grid according to the target residual component sequence. This embodiment first obtains the credibility of the initial residual component sequence. According to the credibility of the initial residual component sequence, for the initial residual component sequence with low credibility, the credibility of the target residual component sequence is improved by optimizing the preset parameters in the time series decomposition algorithm, thereby reducing the problem of inaccurate fault detection caused by the low credibility of the residual component and improving the accuracy of the residual component for fault detection.

[0057] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A fault detection optimization method for a cellular smart grid, characterized in that: The method comprises: Acquire any type of grid time series data in the smart grid, preprocess the grid time series data, and obtain an original data sequence; Decomposing the original data sequence by using a time series decomposition algorithm to obtain an initial trend component sequence and an initial seasonal component sequence, obtaining a corresponding initial residual component sequence according to the initial trend component sequence and the initial seasonal component sequence, and obtaining the credibility of the initial residual component sequence according to the data distribution characteristics in the initial residual component sequence and the data difference between the initial residual component sequence and the original data sequence; If the credibility of the initial residual component sequence is not within the preset credibility range, the preset parameters in the time series decomposition algorithm are optimized to obtain a new trend component sequence and a new seasonal component sequence corresponding to the original data sequence, and a corresponding new residual component sequence is obtained according to the new trend component sequence and the new seasonal component sequence, and the credibility acquisition method is repeated until the credibility of the new residual component sequence is within the preset credibility range, thereby obtaining a target residual component sequence; According to the target residual component sequence, fault detection is performed on the smart grid.

2. A fault detection optimization method for a cellular smart grid according to claim 1, characterized in that: The obtaining the credibility of the initial residual component sequence according to the data distribution characteristics in the initial residual component sequence and the data difference between the initial residual component sequence and the original data sequence includes: According to the data distribution characteristics in the initial residual component sequence, obtaining the normal distribution degree of the initial residual component sequence; According to the data difference between the initial residual component sequence and the original data sequence, obtaining the degree of difference between the initial residual component sequence and the original data sequence; The credibility of the initial residual component sequence is obtained according to the normal distribution degree and the difference degree.

3. A fault detection optimization method for a cellular smart grid according to claim 2, characterized in that: The step of obtaining the normal distribution degree of the initial residual component sequence according to the data distribution characteristics in the initial residual component sequence includes: Obtain the data mean of the initial residual component sequence, obtain the number of data in the initial residual component sequence that is within a preset normal distribution area, calculate the ratio of the number of data in the initial residual component sequence that is within a preset normal distribution area to the number of data in the initial residual component sequence to obtain a first ratio, and obtain the sum of the absolute value of the data mean of the initial residual component sequence and the first ratio as the degree of normal distribution of the initial residual component sequence.

4. A fault detection optimization method for a cellular smart grid according to claim 2, characterized in that: The obtaining, according to the data difference between the initial residual component sequence and the original data sequence, the degree of difference between the initial residual component sequence and the original data sequence comprises: Fitting the initial residual component sequence to obtain a residual component fitting curve, obtaining an original data curve composed of the original data sequence, obtaining each extreme point in the original data curve, and using each extreme point in the original data curve to divide the original data curve and the residual component fitting curve into at least two sub-curves, respectively, and the sub-curves of the original data curve and the sub-curves of the residual component fitting curve correspond to each other one by one; The curvature of each sub-curve in the original data curve and the residual component fitting curve is obtained, and the absolute value of the curvature difference between each sub-curve in the original data curve and the corresponding sub-curve in the residual component fitting curve is calculated respectively, and the accumulated value of the absolute value is obtained accordingly, and the accumulated value of the absolute value is used as the independent variable of the sigmoid function to obtain the corresponding sigmoid function value as the difference degree between the initial residual component sequence and the original data sequence.

5. A fault detection optimization method for a cellular smart grid according to claim 2, characterized in that: The obtaining the credibility of the initial residual component sequence according to the normal distribution degree and the difference degree includes: The normal distribution degree and the difference degree are weighted and summed, and the obtained weighted summation result is used as the credibility of the initial residual component sequence.

6. A fault detection optimization method for a cellular smart grid according to claim 1, characterized in that: The preset parameters in the time series decomposition algorithm include: seasonal smoothing window and seasonal cycle.

7. A fault detection optimization method for a cellular smart grid according to claim 6, characterized in that: If the credibility of the initial residual component sequence is not within a preset credibility range, optimizing the preset parameters in the time series decomposition algorithm includes: If the credibility of the initial residual component sequence is not within the preset credibility range, the length of the seasonal smoothing window is increased or decreased within the preset length range of the seasonal smoothing window, and the length of the seasonal cycle is increased or decreased within the preset length range of the seasonal cycle.

Images