Transformer area line loss data anomaly identification method based on multi-dimensional features

Abstract

The invention relates to a transformer area line loss data anomaly identification method based on multi-dimensional features. The method comprises the following steps: acquiring original line loss data and geographic and topological data of each monitoring point in a target transformer area; constructing a two-dimensional grayscale image matrix based on the original line loss data, and performing two-dimensional wavelet threshold denoising to obtain line loss data; constructing a weighted multi-dimensional distance space of the target station area, and generating a data similarity matrix through distance measurement; a data similarity matrix is divided into a normal mode and an abnormal mode based on a hierarchical clustering algorithm, an identification result of the abnormal mode is output, noise interference is eliminated through two-dimensional wavelet threshold denoising, abnormal signal loss caused by smooth transition is avoided, and a high-quality data basis is provided for subsequent feature extraction. The spatial position, the change speed and the trend direction are taken into unified measurement, the problem of recognition blind areas of regional abnormity and gradual change abnormity in a traditional method is solved, early warning of the initial symptoms of equipment faults is achieved, and efficient technical support is provided for precise operation and maintenance of a smart power grid.

Description

Technical Field

[0001] The present application relates to the technical field of distribution network, and in particular to a method for identifying abnormalities in substation line loss data based on multi-dimensional features. Background Art

[0002] In power systems, a substation specifically refers to the power supply area or region of a transformer and is a key concept in power economic operation management. Substation line loss is the difference between the line loss measured by the meter installed on the low-voltage outlet of a transformer and the line loss measured by all users in the substation during the same period. It reflects the energy loss caused by factors such as resistance and inductance during power transmission and distribution.

[0003] Identifying abnormal line loss data in substations is a core part of power grid operation and maintenance. Existing technical methods include: setting thresholds based on the absolute value of the line loss rate or simple statistics to directly mark excessive data as abnormal; using sliding windows, autoregressive models (AR / ARIMA), etc. to analyze data trends and identify abnormal points through residual deviation; clustering data points based on a single metric such as Euclidean distance, relying on manually setting the number of clusters or thresholds.

[0004] In practical applications, existing methods suffer from the following drawbacks: They lack effective processing of noise data collected by monitoring terminals. Substation line loss data, in particular, is susceptible to electromagnetic interference and communication interruptions, leading to high rates of false positives in anomaly identification, such as misidentifying noise fluctuations as true anomalies. Models rely solely on time series characteristics of line loss values (such as trends and volatility) or static electrical parameters (current, voltage, power factor, and user load), ignoring multidimensional dynamic spatiotemporal information such as spatial location, rate of change, and direction of fluctuation. This makes it difficult to characterize complex anomaly patterns, such as topological association errors and regional anomalies caused by distributed power generation (DG) access. Traditional clustering algorithms require a preset number of clusters or rely on manually labeled samples. These algorithms are unsuitable for scenarios where substation line loss data lacks prior labels and anomaly types change dynamically, such as new power theft schemes and equipment parameter drift, leading to missed detection of unknown anomalies. Summary of the Invention

[0005] In order to solve or partially solve the problems existing in the related technologies, the present application provides a method for identifying anomalies in line loss data in a substation based on multi-dimensional features. Through two-dimensional wavelet threshold denoising and three-dimensional feature fusion of position, speed and direction, a weighted multi-dimensional distance space is constructed to quantify the spatiotemporal differences of data points, breaking through the limitations of traditional line loss analysis and forming a complete anomaly identification technology chain.

[0006] The first aspect of the present application provides a method for identifying abnormalities in line loss data in a substation area based on multi-dimensional features, comprising the following steps: Obtain original line loss data, geographic and topological data of each monitoring point in the target substation area; A two-dimensional grayscale image matrix is constructed based on the original line loss data, and the line loss data is obtained by two-dimensional wavelet threshold denoising. Construct a weighted multidimensional distance space for the target substation area and generate a data similarity matrix through distance measurement. The weighted multidimensional distance space includes location features based on geographic and topological data and speed and direction features based on line loss data. The data similarity matrix is divided into normal and abnormal modes based on the hierarchical clustering algorithm, and the recognition results of abnormal modes are output.

[0007] Among them, the line loss data obtained by two-dimensional wavelet threshold denoising includes: Decompose the two-dimensional signal using wavelet transform method, select the best wavelet basis according to the decomposition result and perform layer decomposition; The wavelets of each layer are preprocessed separately, and the estimated coefficients are obtained after weighted average threshold function processing; The image signal is reconstructed by two-dimensional wavelet and the line loss data is obtained after denormalization.

[0008] Among them, the location feature is formed by extracting the spatial coordinates or network connection relationship of the monitoring points based on geographic and topological data: given any two data points and , the distance between the two Expressed as: ; Where: represents the total number of datasets; and Represents the coordinate position of the data point; Represents the direction of the data column.

[0009] Among them, the speed feature is based on the line loss data to calculate the rate of change in each period, and the Represents the velocity Euclidean distance between two data points, where the velocity of the point can be decomposed into vertical and horizontal velocities.

[0010] Among them, the directional characteristics are based on the line loss data to calculate the increase and decrease trends of each period. Represents the degree of change in the internal direction of two data points and the fluctuation of line loss data.

[0011] Among them, generating a data similarity matrix through distance measurement includes: The multi-dimensional feature distance of line loss data is calculated based on position, speed and direction features , as shown below: ; Where: 、 、 They are position, speed and direction characteristics respectively, the sum of the three parameters is equal to 1 and all are greater than or equal to 0; The multidimensional similarity distance between any two line loss data is calculated using the multidimensional feature distance method based on time series , as shown below: ; Where: ; ; Based on the multidimensional feature distance and multidimensional similarity distance, the direct similarity distance of the line loss data and the similarity between the line loss data are calculated to generate a data similarity matrix.

[0012] Among them, the data similarity matrix Expressed as: ; Where: Representative Line loss data and The similarity distance between line losses is calculated by multi-dimensional similarity distance between line loss data. Can be obtained directly ; 0 represents the similarity distance between the line loss data themselves.

[0013] Among them, the recognition results of abnormal patterns include single line loss abnormalities and multiple line loss abnormalities.

[0014] The technical solution provided by this application may have the following beneficial effects: This application provides a method for identifying abnormalities in substation line loss data based on multi-dimensional features. It eliminates noise interference through two-dimensional wavelet threshold denoising, breaks through the smoothing effect of traditional filtering methods on sudden changes in features, retains key detail features such as load spikes and trend turning points, avoids the loss of abnormal signals due to excessive smoothing, and provides a high-quality data foundation for subsequent feature extraction. Incorporating spatial position, change speed, and trend direction into a unified metric solves the blind spots of traditional methods in identifying regional anomalies and gradual anomalies, captures line loss trend reversals, and provides early warning of early signs of equipment failure, providing efficient technical support for the precise operation and maintenance of smart grids.

[0015] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and other objects, features and advantages of the present application will become more apparent through a more detailed description of exemplary embodiments of the present application in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments of the present application.

[0017] Figure 1 1 is a flow chart of an abnormality identification method of a device shown in an embodiment of the present application; Figure 2 1 is a schematic diagram of an abnormality identification logic of an abnormality identification method of a device shown in an embodiment of the present application; Figure 3 Schematic diagram of the spatiotemporal coupling of factors affecting line loss rate in the abnormality identification method of the device shown in the embodiment of the present application. DETAILED DESCRIPTION

[0018] The following describes embodiments of the present application in more detail with reference to the accompanying drawings. Although the accompanying drawings illustrate embodiments of the present application, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. Rather, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0019] It should be understood that although the terms "first", "second", "third", etc. may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0020] In the description of this application, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on this application.

[0021] Unless otherwise expressly specified or limited, terms such as "mounted," "connected," "connect," and "fixed" should be interpreted broadly. For example, they may refer to fixed or detachable connections, or integration; mechanical or electrical connections; direct or indirect connections through an intermediary; and internal communication between two components or interaction between two components. Those skilled in the art will understand the specific meanings of these terms in this application based on the specific circumstances.

[0022] The technical solutions of the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0023] like Figure 1 The method for identifying abnormal line loss data in a substation area based on multi-dimensional features includes the following steps: S1. Obtain the original line loss data, geographic and topological data of each monitoring point in the target substation area.

[0024] The line loss data of the power grid area is composed of a series of data points that change over time. Among them, the line loss data set represents The collection of line loss data, that is ; Where, Representative Line loss data.

[0025] Raw line loss data refers to unprocessed basic data collected directly from substation monitoring equipment (such as electricity meters and distribution transformer terminals). It is usually stored in the form of discrete points and includes: Timestamp: the specific time when the data was collected (e.g., 2025-06-05 08:00); Spatial identification: spatial attributes such as substation number, distribution transformer ID, and line node location; Physical quantity value: Real-time line loss rate (%), current (A), voltage (V), power factor, etc.; Derived quantities such as load power (kW), accumulated electricity (kWh), etc.

[0026] Geographic and topological data include: Geographic data, including the geographic location coordinates of the substation and the monitoring points, are used to calculate the physical distance between monitoring points or analyze spatial distribution characteristics. For example, substations in the same area may be affected by the same weather; and to calculate the physical distance between monitoring points or analyze spatial distribution characteristics, such as whether meters in the same area collectively experience high losses due to aging lines.

[0027] Topological data includes the connections between substations within the power grid, such as the feeder hierarchy, electrical connection paths with adjacent substations, and the connections between monitoring points within a substation, such as the hierarchy of transformer → trunk line → branch line → user meter. This grid connection is used to calculate topological distances between substations, such as level differences or electrical impedance. The connections between monitoring points within a substation are also used to calculate topological distances between monitoring points. For example, the distance from a transformer to a meter is 3, as it requires three branches.

[0028] S2. Construct a two-dimensional grayscale image matrix based on the original line loss data, and perform two-dimensional wavelet threshold denoising to obtain line loss data.

[0029] Since the amount of data monitored by the relevant monitoring terminals in the power grid substation is large and contains noise, the abnormal identification results of the power grid line loss data will be inaccurate. Therefore, a two-dimensional wavelet threshold denoising method is used to denoise the power grid line loss data in the power grid substation, thereby improving the accuracy of data anomaly identification.

[0030] First, the line loss data of the substation containing noise is Perform wavelet transform and decompose into layer, and obtain a set of wavelet coefficients , For the After layer decomposition After decomposition, threshold processing and signal reconstruction are performed on the wavelet coefficients according to the characteristics of the original data and background noise, which can eliminate the interference of background noise. In the threshold denoising of wavelet transform, the selection of threshold plays a very important role.

[0031] The following mainly uses the VisuShrink method to determine the threshold , as shown below.

[0032] ; Where: Represents the noise standard in the line loss data of the power grid area; Represents the length of the signal.

[0033] in, It can be estimated based on the absolute value of the decomposed high-frequency coefficients, as shown in the following formula.

[0034] ; Where, is the wavelet coefficient, The median of the data series, , .

[0035] The traditional threshold function can be divided into two different forms: soft threshold and hard threshold. Among them, the hard threshold is to retain the wavelet coefficients that are not lower than the threshold and set all their values to 0. The specific expression is shown in the following formula.

[0036] ; The soft threshold criterion is to set the wavelet coefficients below the threshold to 0, and combine the characteristics of the threshold itself to obtain the following calculation formula.

[0037] ; According to the above formula, the curves corresponding to the soft and hard thresholds can be obtained.

[0038] In the soft and hard thresholding methods, exist is discontinuous, which leads to The signal obtained by reconstruction will have a certain degree of oscillation; on the other hand, although the soft threshold is used to estimate It is continuous, but the corresponding derivative is not continuous, so there are certain difficulties in solving high-order derivatives.

[0039] Based on the above analysis, it can be seen that both soft and hard threshold functions will be affected to varying degrees during the denoising process, resulting in defects. Therefore, this embodiment uses a weighted average threshold function based on soft and hard functions for denoising, as specifically expressed in the following formula.

[0040] ; Where: Represents the weighting factor, and the detailed formula is shown below: ; Combined with the above analysis, it can be obtained that the two-dimensional wavelet denoising of the power grid line loss data includes the following steps: Step 1: Select the prediction The line loss data of the day is used as data samples, and the line loss data samples are formed into corresponding two-dimensional data.

[0041] Step 2: Normalize the above data to obtain two-dimensional grayscale image matrix data.

[0042] Step 3: After obtaining the two-dimensional signal, the wavelet transform method is used to decompose the signal, and the best wavelet basis is selected according to the decomposition result. Layer decomposition, we get .

[0043] Step 4: Preprocess the wavelets at different levels and determine the range of threshold values. Perform semi-soft threshold function processing to obtain the estimated coefficients.

[0044] Step 5: Reconstruct the image signal through two-dimensional wavelet to obtain the denoised signal.

[0045] Step 6: Denormalize the reconstructed signal to achieve denoising of the power grid line loss data.

[0046] S3. Construct a weighted multidimensional distance space of the target substation and generate a data similarity matrix through distance measurement. The weighted multidimensional distance space includes location features based on geographic and topological data and speed features and direction features based on line loss data.

[0047] To better identify anomalies in power grid line loss data, it's necessary to consider the data's own characteristics, such as speed and direction, in addition to considering location and time characteristics. The following uses the Hasusdorff distance to calculate the similarity of line loss data. By combining temporal characteristics with information such as the transmission speed and direction of power grid line loss data, the Hasusdorff distance function is improved with the help of multidimensional features in time series to ensure the accuracy of the final calculation results and thus improve the algorithm's computational efficiency.

[0048] Among them, the time series-based Hasusdorff distance algorithm prioritizes incorporating the power grid line loss data into the Hasusdorff distance formula according to multidimensional features such as location and direction. The multidimensional Hasusdorff distance is combined to calculate the similarity between random data in the line loss dataset, and based on this, a similarity matrix is constructed. The specific operation process is as follows.

[0049] (1) Position characteristics Given any two data points and , then the distance between the two Can be expressed as ; Where: represents the total number of datasets; and Represents the coordinate position of the data point; Represents the direction of the data column.

[0050] (2) Speed characteristics Represents the velocity Euclidean distance between two data points, where the velocity of a point can be decomposed into vertical and horizontal velocities.

[0051] (3) Directional characteristics It represents the degree of change in the internal direction of two data points, and also reflects the fluctuation of line loss data.

[0052] Combining position, speed and direction features, the multi-feature two-point distance of line loss data can be calculated , as shown below.

[0053] ; Where: 、 、 They are position, speed, and direction features, respectively. The sum of the three parameters is equal to 1 and all are greater than or equal to 0. In this paper, they are set to 0.30, 0.30, and 0.40.

[0054] Next, we use the multidimensional feature distance method based on time series to calculate the multidimensional similarity distance between any two line loss data. .

[0055] ; Where: ; .

[0056] After obtaining the multi-dimensional feature distance and multi-dimensional similarity distance, the direct similarity distance of the line loss data can be obtained. Then the similarity between the line loss data is calculated and the similarity matrix is constructed. ,Right now ; Where: Representative Line loss data and The similarity distance between line losses is calculated by multi-dimensional similarity distance between line loss data. Can be obtained directly ; 0 represents the similarity distance between the line loss data themselves.

[0057] S4. Divide the data similarity matrix into normal and abnormal modes based on the hierarchical clustering algorithm, and output the recognition results of the abnormal mode.

[0058] Hierarchical clustering is an unsupervised non-parametric clustering method that can be applied to group and cluster data with similar characteristics, and to cut and form dendrograms at different levels without prioritizing the number of clusters.

[0059] Unsupervised learning is a type of machine learning that classifies both tangible and intangible objects without requiring any prior information. Attackers frequently change their attack patterns and types during attacks, making unsupervised learning more suitable in such situations. Therefore, the system must rapidly learn from attack data and infer new attack types to take preemptive action. Line loss data at power grid stations can accurately monitor the operating status of the power grid.

[0060] Classes can usually be represented in the following different forms: (1) Represent a class by its center or boundary point; (2) graphically represent a cluster through nodes in a cluster tree; (3) Representing clusters through logical expressions of sample attributes. The metric between the line losses in the power grid area can effectively determine the cluster division, and the clustering quality can be experienced using the number of similarities between clusters and the similarity between clusters. Among them, the more commonly used clustering distance function is the Min distance function, which is set to and Represents the calculated sample, then the Minn distance between the two It can be expressed as the following formula.

[0061] ; Where: and Representing the The samples to be calculated; Represents the absolute value distance.

[0062] On the basis of the above analysis, the hierarchical clustering method of multi-dimensional Hasusdorff distance is combined to identify abnormal line loss data in power grid areas, such as Figure 2 As shown in the figure, the multidimensional Hasusdorff distance is further applied to the hierarchical clustering algorithm, and the clustering algorithm is used to complete the abnormal identification of the line loss data of the power grid station area.

[0063] The formula for calculating line loss rate is: ; For the 10kV branch line loss rate, the power supply refers to the metered electricity consumption at the substation's 10kV outgoing line switch. The power sales account for the load at the same voltage level, including the power consumption of the 10kV distribution transformer and medium-voltage users. When the daily line loss rate is used as the calculation indicator, the power supply and sales values are based on the power collected within a day. When the monthly line loss rate is used as the assessment indicator, the power supply and sales values are based on the power collected within a month.

[0064] At present, considering the theoretical loss of the line and the metering accuracy error on the power supply and sales side, the theoretical reasonable range of the 10kV line loss rate is set to [0, 6%]. When the line loss rate exceeds the reasonable theoretical range, it is necessary to check the abnormal factors in the collection, association, and calculation links. The calculation process of the line loss rate involves three major links: power supply metering, power sales metering, and topological association of the power supply and sales side, covering all kinds of meter information. Spatial factors such as the power supply range of the 10kV line, distribution transformer association ledgers, and user association ledgers; at the same time, the collection, metering, and calculation of power supply and sales involve time factors such as real-time data, daily frozen data, and monthly frozen data. The spatiotemporal coupling relationship of various factors is as follows: Figure 3 shown.

[0065] Depend on Figure 3It can be seen that the line loss rate of 10kV sub-lines is influenced by numerous factors, with complex coupling relationships. Even a slight change in any factor can lead to an unreasonable line loss rate. Some of these factors are fixed; if an abnormality occurs, the line loss rate will remain unreasonable, and after the abnormality is addressed, these factors themselves will not fluctuate. Other factors are variable; abnormalities in these factors can cause the line loss rate to change from a reasonable level to an unreasonable level. Furthermore, these factors themselves are inherently volatile. To further analyze the specific causes of line loss rate fluctuations, it is necessary to conduct qualitative and quantitative analysis of these various influencing factors and identify the variable components.

[0066] Time influencing factors mainly manifest as metering issues on the power supply and sales side, which may appear at any time during line operation and line loss calculation and are all variable components.

[0067] Spatial influencing factors involve issues such as wiring, transformation ratios, and topological relationships on the power supply and retail side. Errors in power supply wiring, transformation ratios, meter relationships, and pole-mounted transformer wiring and documentation can lead to persistently irrational line losses. There is no abnormal change from reasonable to unreasonable line losses. Therefore, these factors are inherent components of spatial influencing factors. Voltage and current losses on the power supply and retail side, as well as changes in their topological relationships, can occur during line operation and line loss calculations and are therefore variable components.

[0068] The recognition results of abnormal patterns include single line loss abnormality and multiple line loss abnormality.

[0069] In the case of a single line loss anomaly, abnormal spatiotemporal factors on a single power supply side or power sales side generally only affect this line and are variable components of the single line loss anomaly, as shown in the following table.

[0070] Table 1 Variable components of temporal and spatial factors causing single line loss fluctuations By analyzing the line loss rate fluctuation characteristics caused by different temporal and spatial influencing factors, the causes of the fluctuation can be accurately identified.

[0071] (1) Characteristics of abnormal changes in line loss rate of a single line Features <0, changes from positive rationality to negative irrationality.

[0072] ① When the line loss data characteristics of the abnormal line meet the following formula, the abnormal cause is the interruption of data collection on the power supply side: ; In the above formula, 、 Line The original power supply and power sales, For the line The average value of power supply at each moment, is the number of communication interruption moments in the line loss rate calculation period, For the line Corrected line loss rate.

[0073] ② When the following equation is satisfied, the cause of the abnormality is the missing load curve on the power supply side: ; Where, The number of missing moments of the load curve in the line loss rate calculation period.

[0074] ③ When the following formula is satisfied, the abnormality is caused by the loss of current in the power supply side meter: ; Where, For the line Power supply side switch Power. =0, the abnormal cause is three-phase current loss; when or( - ) / ≈33%, the abnormal cause is single-phase current loss on the power supply side; when or( - ) / ≈66% of the abnormalities are caused by two-phase current loss on the power supply side; =0 or ( - ) / ≈100% Under other conditions, the abnormal cause is partial loss of flow.

[0075] (2) Characteristics of abnormal changes in line loss rate of a single line Features >6%, changes from positive rationality to negative irrationality.

[0076] ① When the line loss data characteristics of the abnormal line meet the following formula, the abnormal reason is the failure of the pole-mounted transformer meter to collect data: ; Where, For the line The pole with transformer The average power consumption at each moment, The number of cycle moments for calculating line loss rate.

[0077] ② When the line loss data characteristics of the abnormal line meet the following formula, the abnormal cause is the loss of pressure and current in the pole-mounted transformer meter: ; Where, For the line The pole with transformer The user-side electricity consumption.

[0078] ③ When the line loss data characteristics of the abnormal line meet the following formula, the abnormal reason is the user meter collection failure: ; Where, For the line Users The average power consumption at each moment, The number of cycle moments for calculating line loss rate.

[0079] ④ When the following formula is satisfied, the abnormal reason is the user's loss of pressure and flow: ; Where, For the line Users The original measured power.

[0080] ⑤ When the following equation is satisfied, the abnormality is caused by the user stealing electricity: ; Where, According to the line Reasonable users The fitted electrical quantity is calculated from the electrical quantity.

[0081] In summary, the temporal and spatial influencing factors of a single line loss fluctuation and its corresponding line loss characteristics are shown in the following table.

[0082] Table 2 Identification of causes of single line loss fluctuations In the case of multiple line loss anomalies, incorrect topology or field switch status can affect line loss calculations for lines on both sides of the switch. Similarly, issues with the association of user points and their meters can affect line loss calculations for both correctly and incorrectly associated lines. Therefore, spatiotemporal factors related to topological associations can cause simultaneous anomalies on at least two lines, representing a variable component of multiple anomalies. Furthermore, a loss of power on the supply side can lead to metering anomalies on the supply side of the entire distribution line under the busbar, also representing a variable component of multiple anomalies, as shown in the table below.

[0083] Table 3 Variable components of spatiotemporal factors causing multiple line loss fluctuations By analyzing the line loss rate fluctuation characteristics caused by different temporal and spatial influencing factors, accurate identification of the causes of the fluctuation can be achieved.

[0084] (1) The line loss rate abnormality characteristic of two lines with graphic switch connection is that the two lines change from reasonable to unreasonable at the same time, one positive and one negative, that is, .

[0085] When the line loss data characteristics of the abnormal line meet the following formula, and lines A1 and A2 are two incoming and outgoing lines of the same switch station A, the abnormal reason is that the sectionalizer remains closed after the on-site backup automatic switching device is activated, that is, the on-site switch status is abnormal: ; Where, For the line of Station power ratio coefficient, 、 They are 、 Original line loss rate 、 Line 、 Original power supply, 、 Line 、 Original electricity sales, For the corrected line 、 Baling line loss rate, 、 Line 、 Corrected line loss rate.

[0086] When the following equation is met, the cause of the change is abnormal status of the graphic switch: ; Where, 、 Line 、 Original line loss rate, Line 、 Original power supply, Line 、 Original electricity sales, For the corrected line 、 Baling line loss rate, line 、 In power supply range and The power consumption of the load between switches.

[0087] (2) The line loss rate abnormality characteristic of two lines with graphical switch connection is that both lines change from reasonable to negative unreasonable at the same time, that is, .

[0088] When the line loss data characteristics of the abnormal line meet the following formula, the abnormal cause is the line graphic ring network: ; (3) The line loss rate abnormality characteristic of two lines with graphical switch connection is that both lines change from reasonable to positive unreasonable at the same time, that is, .

[0089] When the line loss data characteristics of the abnormal line meet the following formula, the abnormal cause is the line field ring network: ; Where, 、 Line 、 Original forward power supply; 、 Line 、 The original statistical value of the original reverse power supply, 、 Line 、 Original electricity sales volume.

[0090] (4) The abnormal change characteristics of the line loss rate of two lines with user change association are that the two lines change from reasonable to unreasonable at the same time, one positive and one negative, i.e. .

[0091] When the line loss data characteristics of the abnormal line meet the following formula, the abnormal cause is the user point-line association error: ; Where, For actual and was mistakenly associated with The user's electricity consumption.

[0092] (5) For two lines that serve as backup for each other at the same user point, the line loss rate fluctuation characteristic is that the two lines change from reasonable to one positive and one negative at the same time, that is, .

[0093] When the line loss data characteristics of the abnormal line meet the following formula, the abnormal cause is the user point-meter association error: ; Where: For actual Line The user clicked and was mistakenly associated with Line The metered electricity consumption at the user's point, For actual Line The user clicked and was mistakenly associated with Line The meter measures the electricity consumption at the user's point.

[0094] (6) It is unreasonable for the 10kV lines of the same busbar section of the substation to become negative at the same time, i.e. , It is a collection of lines on the same busbar section.

[0095] When the line loss data characteristics of the abnormal line meet the following formula, the abnormal cause is voltage loss on the power supply side: ; In summary, the temporal and spatial influencing factors of multiple line loss fluctuations and their corresponding line loss characteristics are shown in the following table.

[0096] Table 4 Identification of causes of multiple line loss fluctuations Finally, it should be noted that, in this document, relationships such as first and second, etc., are used solely to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms include, comprise, or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0097] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0098] The embodiments of the present application have been described above. The above description is illustrative and not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or improvements to the technology in the market, or to enable other persons skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for identifying abnormalities in line loss data in a substation area based on multi-dimensional features, characterized in that: include: Obtain original line loss data, geographic and topological data of each monitoring point in the target substation area; Constructing a two-dimensional grayscale image matrix based on the original line loss data, and performing two-dimensional wavelet threshold denoising to obtain line loss data; Constructing a weighted multidimensional distance space of the target substation and generating a data similarity matrix through distance measurement, wherein the weighted multidimensional distance space includes location features based on the geographic and topological data and speed features and direction features based on the line loss data; The data similarity matrix is divided into normal and abnormal modes based on a hierarchical clustering algorithm, and the recognition result of the abnormal mode is output.

2. The method for identifying abnormalities in substation line loss data based on multidimensional features according to claim 1, characterized in that: The two-dimensional wavelet threshold denoising to obtain line loss data includes: Decompose the two-dimensional signal using wavelet transform method, select the best wavelet basis according to the decomposition result and perform layer decomposition; The wavelets of each layer are preprocessed separately, and the estimated coefficients are obtained after weighted average threshold function processing; The image signal is reconstructed by two-dimensional wavelet and the line loss data is obtained after denormalization processing.

3. The method for identifying abnormalities in substation line loss data based on multidimensional features according to claim 1, characterized in that: The location feature is formed based on the spatial coordinates or network connection relationship of the monitoring points extracted from the geographic and topological data: given any two data points and , the distance between the two Expressed as: ； Where: represents the total number of datasets; and Represents the coordinate position of the data point; Represents the direction of the data column.

4. The method for identifying abnormalities in substation line loss data based on multi-dimensional features according to claim 3 is characterized in that: The speed feature calculates the rate of change of each time period based on the line loss data, and uses Represents the velocity Euclidean distance between two data points, where the velocity of the point can be decomposed into vertical and horizontal velocities.

5. The method for identifying abnormalities in substation line loss data based on multi-dimensional features according to claim 4 is characterized in that: The directional characteristics are calculated based on the line loss data to calculate the increase and decrease trends in each period. Represents the degree of change in the internal direction of two data points and the fluctuation of line loss data.

6. The method for identifying abnormalities in substation line loss data based on multi-dimensional features according to claim 5, characterized in that: Generating a data similarity matrix by distance measurement includes: The multi-dimensional feature distance of the line loss data is calculated based on the position, speed and direction features , as shown below: ； Where: 、 、 They are position, speed and direction characteristics respectively, the sum of the three parameters is equal to 1 and all are greater than or equal to 0; The multidimensional similarity distance between any two line loss data is calculated using the multidimensional feature distance method based on time series , as shown below: ； Where: ; ; Based on the multidimensional feature distance and the multidimensional similarity distance, the similarity distance between the line loss data and the similarity between the line loss data are calculated to generate the data similarity matrix.

7. The method for identifying abnormalities in substation line loss data based on multi-dimensional features according to claim 1, characterized in that: The data similarity matrix Expressed as: ； Where: Representative Line loss data and The similarity distance between line losses is calculated by multi-dimensional similarity distance between line loss data. Can be obtained directly ; 0 represents the similarity distance between the line loss data themselves.

8. The method for identifying abnormalities in substation line loss data based on multi-dimensional features according to claim 1, characterized in that: The recognition results of the abnormal pattern include a single line loss abnormality situation and a multiple line loss abnormality situation.

Images