Power consumption abnormity detection equipment and method based on dynamic tracking of electrical parameters

Abstract

The invention discloses a power utilization abnormity detection device and method based on electric parameter dynamic tracking, and particularly relates to the technical field of power utilization monitoring, the method comprises the steps that the electric parameter data flow of the nth electric device in a target area is acquired, and the electric parameter data flow comprises the current value, the voltage value, the active power and the reactive power; acquiring instantaneous detection data based on the electrical parameter data stream, analyzing the instantaneous detection data, and judging whether the electric equipment is marked as short-time abnormal power utilization; obtaining harmonic detection data based on the electric parameter data flow, analyzing the harmonic detection data, and judging whether the electric equipment is marked as harmonic abnormal power; acquiring long-term detection data based on the electric parameter data stream, analyzing the long-term detection data, and judging whether the electric equipment is marked as long-term abnormal power consumption or not; the intelligent level of abnormal power consumption detection is improved, the prediction effect can be continuously optimized through model training, and automatic and accurate analysis of abnormal power consumption is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of power consumption monitoring. More specifically, the present invention relates to an equipment and method for detecting abnormal power consumption based on dynamic tracking of electrical parameters. Background Art

[0002] With the continuous expansion of the scale of power consumption and the increasing diversification of power-consuming equipment, detecting abnormal power consumption faces many challenges. Traditional power consumption detection mainly relies on simple threshold setting, such as only judging whether the current and voltage amplitudes exceed the limit. However, in reality, the working states of power-consuming equipment are complex, and the electrical parameters change dynamically during normal operation. For example, for large motors in factories, the current surges significantly at the moment of startup and then drops back after stabilization, and the electrical parameters vary significantly under different working conditions; moreover, the power consumption patterns of equipment with different specifications and types are different, and fixed thresholds are difficult to accurately distinguish normal changes from abnormalities. At the same time, the phenomenon of electricity theft occurs frequently. By using complex technical means to interfere with the electricity meter or circuit, subtle abnormal changes in electrical parameters are caused, and traditional detection methods are extremely prone to missed detections. In addition, the development of smart grids has led to a huge increase in power data, and there is an urgent need to accurately detect this data.

[0003] In existing methods, for example, the patent application with the publication number CN111103497A discloses a detection device and method for abnormal power consumption of users. The detection device includes a master station terminal installed at the starting position of the branch line and a portable terminal at the incoming line end of the meter box. The portable terminal scans the asset number of the electricity meter in the meter box, obtains the communication address, and forms an electricity meter file in units of meter boxes; the master station terminal collects the voltage and current of the branch line, generates daily frozen data, and reads the daily frozen data and files of the portable terminal and its subordinate electricity meters. By calculating the daily frozen electricity increment and the user electricity increment by the master station terminal, it is judged whether there is a line loss abnormality in the branch line and the meter box. Although the above method can improve the convenience of detecting abnormal power consumption, through research and application of the above method and the existing technology, it is found that the above method and the existing technology have at least the following partial defects:

[0004] The above method judges the line loss abnormality only based on numerical comparison, rather than intelligent pattern recognition, and it is difficult to analyze complex abnormal power consumption behaviors.

[0005] Therefore, the present invention provides an equipment and method for detecting abnormal power consumption based on dynamic tracking of electrical parameters. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides an equipment and method for detecting abnormal power consumption based on dynamic tracking of electrical parameters to solve the problems raised in the above background art.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] In a first aspect, the present invention provides a method for detecting abnormal electricity consumption based on dynamic tracking of electrical parameters, including:

[0009] Obtain the electrical parameter data stream of the nth electrical device in the target area, where the electrical parameter data stream includes current value, voltage value, active power, and reactive power;

[0010] Based on the electrical parameter data stream, obtain instantaneous detection data, analyze the instantaneous detection data, and determine whether to mark the electrical device as short-term abnormal electricity consumption;

[0011] Based on the electrical parameter data stream, obtain harmonic detection data, analyze the harmonic detection data, and determine whether to mark the electrical device as harmonic abnormal electricity consumption;

[0012] Based on the electrical parameter data stream, obtain long-term detection data, analyze the long-term detection data, and determine whether to mark the electrical device as long-term abnormal electricity consumption;

[0013] Obtain abnormal feature data based on short-term abnormal electricity consumption, harmonic abnormal electricity consumption, and long-term abnormal electricity consumption, and input the abnormal feature data into a pre-constructed abnormal electricity consumption analysis model to obtain an analysis result.

[0014] Further, the instantaneous detection data includes instantaneous voltage, instantaneous current, and power factor;

[0015] The method for analyzing the instantaneous detection data includes:

[0016] Step a1: Calculate the instantaneous current rate, instantaneous voltage rate, and real-time power factor through a sliding window, and the calculation formulas are:

[0017] In the formula, ΔI represents the instantaneous current rate, ΔU represents the instantaneous voltage rate, Δt represents the time interval, PF represents the real-time power factor, P represents the active power, S represents the apparent power; I(t) represents the current value, and U(t) represents the voltage value;

[0018] Step a2: Set the instantaneous current rate threshold ΔI th 、the instantaneous voltage rate threshold ΔU th and the lower limit threshold PF of the power factor th ;

[0019] Compare the instantaneous current rate ΔI with the instantaneous current rate threshold ΔI th , compare the instantaneous voltage rate ΔU with the instantaneous voltage rate threshold ΔU th , and compare the real-time power factor PF with the lower limit threshold PF of the power factor th ;

[0020] If ΔI > ΔI th and ΔU < ΔUth and PF < PF th , then mark the current electrical equipment as short - term abnormal power consumption;

[0021] If ΔI ≤ ΔI th and ΔU ≥ ΔU th and PF ≥ PF th , then do not process the current electrical equipment;

[0022] Step a3: Record the short - term timestamp and the mutant electrical parameter values according to the short - term abnormal power consumption.

[0023] Furthermore, the harmonic detection data includes current, voltage values, and harmonic distortion rate; the harmonic abnormal power consumption includes the first - type harmonic abnormal power consumption and the second - type harmonic abnormal power consumption;

[0024] The method for analyzing the harmonic detection data based on the harmonic analysis method includes:

[0025] Step b1: Use the fast Fourier transform to perform spectral analysis on the current value and extract the harmonic component amplitudes; its calculation formula is:

[0026] H b = FFT(I(t)), b = 1, 2, 3,..., B;

[0027] In the formula, H b represents the amplitude of the b - th harmonic component, and B is the highest harmonic order.

[0028] Step b2: Calculate the harmonic distortion rate according to the harmonic component amplitude H b , and its calculation formula is:

[0029]

[0030] In the formula, THD represents the harmonic distortion rate, and H 1 represents the amplitude of the 1 - st harmonic component;

[0031] Step b3: Preset a distortion rate threshold, compare the harmonic distortion rate with the preset distortion rate threshold. If the harmonic distortion rate is greater than or equal to the preset distortion rate threshold, then mark the electrical equipment as the first - type harmonic abnormal power consumption; if the harmonic distortion rate is less than the preset distortion rate threshold, then generate an analysis instruction;

[0032] Step b4: Receive the analysis instruction, obtain the harmonic characteristic coefficient, and compare the harmonic characteristic coefficient with the preset first - coefficient threshold to determine whether to mark the electrical equipment as the second - type harmonic abnormal power consumption;

[0033] Step b5: Record the harmonic abnormal electrical parameter values and harmonic timestamps according to the first - type harmonic abnormal power consumption or the second - type harmonic abnormal power consumption.

[0034] Further, the method for obtaining the harmonic feature coefficient includes:

[0035] Step b41: Obtain harmonic feature data, where the harmonic feature data includes harmonic component amplitude, harmonic distortion rate, and harmonic duration;

[0036] Step b42: Normalize the harmonic feature data and calculate the harmonic feature coefficient, and its calculation formula is:

[0037] XBT = H b ×XB 1 + THD × XB 2 + Sj × XB 3 ;

[0038] In the formula, XBT represents the harmonic feature coefficient, Sj represents the harmonic duration; XB 1 、XB 2 and XB 3 are weighting factors.

[0039] Further, the method for determining whether to mark an electrical device as a second-harmonic abnormal power consumption device includes:

[0040] If the harmonic feature coefficient is greater than or equal to a preset first coefficient threshold, mark the electrical device as a second-harmonic abnormal power consumption device;

[0041] If the harmonic feature coefficient is less than the preset first coefficient threshold, do not process the electrical device.

[0042] Further, the method for obtaining the harmonic duration includes:

[0043] Set a harmonic distortion rate threshold. When THD is greater than or equal to the set harmonic distortion rate threshold, record the start time as t 1 , when THD is less than the set harmonic distortion rate threshold, record the end time as t 2 ;

[0044] Calculate the harmonic duration, and its calculation formula is Sj = t 2 - t 1 ;

[0045] The method for determining whether to mark an electrical device as a second-harmonic abnormal power consumption device further includes:

[0046] Pre-store multiple sets of fingerprint feature data recorded by historical electricity theft devices in the harmonic fingerprint database to form a matching template;

[0047] The matching template is: device type - harmonic component amplitude - harmonic distortion rate - harmonic duration;

[0048] Perform a similarity match between the harmonic feature data and the fingerprint feature data to obtain a similarity value;

[0049] Preset a similarity threshold, compare the similarity value with the preset similarity threshold. If the similarity value is greater than or equal to the preset similarity threshold, mark the electrical equipment as having abnormal second-harmonic power consumption; if the similarity value is less than the preset similarity threshold, do not process the electrical equipment.

[0050] Furthermore, the long-term detection data includes active power, reactive power, and power factor;

[0051] The method for analyzing the long-term detection data includes:

[0052] Step c1: Extract the long-term load factor during the historical load monitoring process and mark it as the historical long-term load factor. Use the extracted historical long-term load factor to establish a load time series set. The load time series set includes i historical long-term load factors. The time intervals for collecting the i historical long-term load factors are equal, and the i historical long-term load factors correspond to a long-term analysis period;

[0053] The method for obtaining the long-term load factor includes:

[0054] Perform a formula-based calculation on the active power, reactive power, and power factor in the long-term detection data to obtain the long-term load factor. The calculation formula is:

[0055]

[0056] In the formula, CFX represents the long-term load factor, and W represents the reactive power.

[0057] Step c2: Input the historical long-term load factors in the load time series set into the load prediction model to predict the long-term load factor at the future T moment.

[0058] Furthermore, the method for determining whether to mark the electrical equipment as having long-term abnormal power consumption includes:

[0059] Step d1: Compare the predicted long-term load factor CFX at the future T moment t+1 with the preset load factor threshold FX;

[0060] Step d2: Determine whether to mark the electrical equipment at the future T moment;

[0061] If CFX t+1 ≥FX, mark the electrical equipment at the future T moment as having long-term abnormal power consumption;

[0062] If CFX t+1<FX, then no processing is performed on the electrical equipment at the future time T, and the monitoring of the electrical equipment at the future time T continues;

[0063] Step d3: Based on the long-term abnormal power consumption records, long-term abnormal parameter values, and long-term timestamps.

[0064] Furthermore, the abnormal feature data includes short-term timestamps and mutated parameter values of short-term abnormal power consumption records, harmonic abnormal parameter values and harmonic timestamps of harmonic abnormal power consumption records, and long-term abnormal parameter values and long-term timestamps of long-term abnormal power consumption records; the analysis results include power theft abnormal power consumption, equipment failures, and line failures;

[0065] The training method of the abnormal power consumption analysis model includes:

[0066] Pre-collect K sets of abnormal feature data, where K is an integer greater than 1. Set corresponding prediction results for the abnormal feature data, and set different letter labels for power theft abnormal power consumption, equipment failures, and line failures. Mark the letter label of the prediction result as the prediction label, and convert the abnormal feature data and the corresponding prediction label into a corresponding set of feature vectors;

[0067] Use each set of feature vectors as the input of the abnormal power consumption analysis model. The abnormal power consumption analysis model takes a set of prediction labels corresponding to each set of abnormal feature data as the output, and takes the actual prediction label corresponding to each set of abnormal feature data as the prediction target. The actual prediction label is the prediction label preset corresponding to the abnormal feature data; take minimizing the sum of the prediction errors of all abnormal feature data as the training target; train the abnormal power consumption analysis model until the sum of the prediction errors reaches convergence and then stop training. The abnormal power consumption analysis model is specifically a deep neural network model.

[0068] In a second aspect, the present invention provides an abnormal power consumption detection device based on dynamic tracking of electrical parameters; the device includes:

[0069] A data acquisition module that acquires the electrical parameter data stream of the nth electrical equipment in the target area. The electrical parameter data stream includes current value, voltage value, active power, and reactive power;

[0070] An instantaneous detection module that obtains instantaneous detection data based on the electrical parameter data stream, analyzes the instantaneous detection data, and determines whether to mark the electrical equipment as short-term abnormal power consumption;

[0071] A harmonic detection module that obtains harmonic detection data based on the electrical parameter data stream, analyzes the harmonic detection data, and determines whether to mark the electrical equipment as harmonic abnormal power consumption;

[0072] The long-term detection module obtains long-term detection data based on the electrical parameter data stream, analyzes the long-term detection data, and determines whether to mark the electrical equipment as long-term abnormal power consumption.

[0073] The comprehensive analysis module obtains abnormal feature data based on short-term abnormal power consumption, harmonic abnormal power consumption, and long-term abnormal power consumption, and inputs the abnormal feature data into a pre-constructed abnormal power consumption analysis model to obtain an analysis result.

[0074] The technical effects and advantages of the present invention:

[0075] By obtaining the electrical parameter data stream such as the current value, voltage value, active power, and reactive power of the electrical equipment, and combining multi-level analysis of instantaneous detection, harmonic detection, and long-term detection, different types of abnormal power consumption behaviors can be quickly and accurately identified. Instantaneous detection can analyze the instantaneous current, voltage rate, and power factor in real time through a sliding window to quickly identify short-term abnormal power consumption; harmonic detection is based on fast Fourier transform and harmonic feature analysis, which can effectively capture harmonic abnormalities caused by accessing electricity stealing equipment; long-term detection combines historical load data and prediction models to early warn potential long-term abnormal power consumption problems. Through multi-dimensional detection methods, this method can not only identify electricity stealing behaviors, but also accurately classify equipment failures and line failures, improving the comprehensiveness and reliability of abnormal power consumption analysis. Description of the Drawings

[0076] Figure 1 It is a flowchart of the abnormal power consumption detection method based on electrical parameter dynamic tracking in Embodiment 1;

[0077] Figure 2 It is a flowchart of the method for analyzing harmonic detection data based on the harmonic analysis method in Embodiment 1;

[0078] Figure 3 It is a flowchart of the method for analyzing long-term detection data in Embodiment 1;

[0079] Figure 4 It is a structural schematic diagram of the abnormal power consumption detection device based on electrical parameter dynamic tracking in Embodiment 2. Detailed Embodiments

[0080] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0081] In addition, the attached drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities may be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0082] It should be understood that although terms such as "first" and "second" may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0083] Embodiment 1

[0084] Please refer to Figure 1 As shown, this embodiment discloses and provides a method for detecting abnormal electricity consumption based on dynamic tracking of electrical parameters. The method includes:

[0085] Step 1: Obtain the electrical parameter data stream of the nth electrical device in the target area. The electrical parameter data stream includes current value, voltage value, active power, and reactive power.

[0086] It should be noted that: the target area may be the incoming line end of the power line, the branch line end, or the electrical device end of the user. In this embodiment, the electrical device of the user is taken as an example for a detailed description of the method for detecting abnormal electricity consumption. High-precision current transformers (CTs) and voltage transformers (PTs) are respectively installed on the electrical devices, and electrical parameter data is collected through high-frequency synchronous sampling technology. The sampling frequency can reach 50 - 100 Hz to ensure accurate capture of transient electrical parameter data. The collected electrical parameter data is converted into digital signals by a data acquisition module (ADC) and transmitted to the system database in real time through a communication interface (such as RS-485, Ethernet, or wireless module). The data is recorded and stored in timestamp order to form a continuous electrical parameter data stream, providing complete data support for subsequent short-term analysis, long-term analysis, and harmonic analysis.

[0087] Step 2: Obtain instantaneous detection data based on the electrical parameter data stream, analyze the instantaneous detection data, and determine whether to mark the electrical device as having short-term abnormal electricity consumption.

[0088] It should be noted that the instantaneous detection data includes instantaneous voltage, instantaneous current, and power factor;

[0089] In implementation, the method for analyzing the instantaneous detection data includes:

[0090] Step a1: Calculate the instantaneous current rate, instantaneous voltage rate, and real-time power factor through a sliding window. The calculation formulas are as follows:

[0091] In the formula, ΔI represents the instantaneous current rate, ΔU represents the instantaneous voltage rate, Δt represents the time interval, PF represents the real-time power factor, P represents the active power, and S represents the apparent power;

[0092] The current value I(t) reflects the magnitude of the real-time current flowing through the electrical equipment; the voltage value U(t) measures the voltage across the electrical equipment, and the active power P represents the power that the electrical energy is effectively converted into actual work;

[0093] It should be noted that: the unit of the time interval is seconds or minutes, which is specifically set by those skilled in the art according to experience and will not be specifically limited here.

[0094] Step a2: Set the instantaneous current rate threshold ΔI th 、the instantaneous voltage rate threshold ΔU th and the power factor lower limit threshold PF th ;

[0095] Compare the instantaneous current rate ΔI with the instantaneous current rate threshold ΔI th , compare the instantaneous voltage rate ΔU with the instantaneous voltage rate threshold ΔU th , and compare the real-time power factor PF with the power factor lower limit threshold PF th ;

[0096] If ΔI > ΔI th and ΔU < ΔU th and PF < PF th , then mark the current electrical equipment as short-term abnormal power consumption;

[0097] If ΔI ≤ ΔI th and ΔU ≥ ΔU th and PF ≥ PF th , then do not process the current electrical equipment, indicating that the electrical equipment is normally disconnected from the line, and continue with short-term analysis.

[0098] Step a3: Record the short-time timestamp and mutant electrical parameter values according to the short-term abnormal power consumption.

[0099] The mutated electrical parameter value refers to the abnormally changing parameter recorded during the short-term analysis of the electrical equipment, including the instantaneous current change rate, the instantaneous voltage change rate, and the abnormal power factor value.

[0100] Step 3: Obtain harmonic detection data based on the electrical parameter data stream, analyze the harmonic detection data, and determine whether to mark the electrical equipment as harmonically abnormal power consumption:

[0101] It should be noted that: the harmonic detection data includes current, voltage values, and the harmonic distortion rate; the harmonically abnormal power consumption includes the first harmonically abnormal power consumption and the second harmonically abnormal power consumption.

[0102] Please refer to Figure 2 As shown, in the implementation, the method for analyzing the harmonic detection data based on the harmonic analysis method includes:

[0103] Step b1: Use the fast Fourier transform to perform spectral analysis on the current value and extract the harmonic component amplitude; its calculation formula is:

[0104] H b = FFT(I(t)), b = 1, 2, 3,..., B;

[0105] In the formula, H b represents the amplitude of the b-th harmonic component, B is the highest harmonic order, and FFT(·) is the fast Fourier transform algorithm.

[0106] Exemplarily, H 2 represents the 2nd harmonic, and its frequency is 2 times the amplitude of the 1st harmonic component. H 3 represents the 3rd harmonic, and its frequency is 3 times the amplitude of the 1st harmonic component, and so on.

[0107] Step b2: Calculate the harmonic distortion rate according to the harmonic component amplitude H b The calculation formula is:

[0108]

[0109] In the formula, THD represents the harmonic distortion rate, and H 1 represents the amplitude of the 1st harmonic component.

[0110] Step b3: Preset a distortion rate threshold, compare the harmonic distortion rate with the preset distortion rate threshold. If the harmonic distortion rate is greater than or equal to the preset distortion rate threshold, mark the electrical equipment as the first harmonically abnormal power consumption; if the harmonic distortion rate is less than the preset distortion rate threshold, generate an analysis instruction.

[0111] It should be noted that: the amplitude of the harmonic component increases with the increase of the harmonic order, and the order is odd. The electrical equipment in the target area can also be marked as abnormal harmonic power consumption.

[0112] Step b4: Receive the analysis instruction, obtain the harmonic feature coefficient, compare the harmonic feature coefficient with the preset first coefficient threshold, and determine whether to mark the electrical equipment as the second abnormal harmonic power consumption;

[0113] Step b5: Record the abnormal harmonic electrical parameter value and the harmonic time stamp according to the first abnormal harmonic power consumption or the second abnormal harmonic power consumption.

[0114] It should be noted that: the abnormal harmonic electrical parameter value refers to the abnormal harmonic features recorded during the harmonic analysis of the electrical equipment, including the abnormal harmonic component amplitude, the total harmonic distortion rate (THD value), and the related current value and voltage value.

[0115] Specifically, the method for obtaining the harmonic feature coefficient includes:

[0116] Step b41: Obtain the harmonic feature data, where the harmonic feature data includes the harmonic component amplitude, the harmonic distortion rate, and the harmonic duration;

[0117] It should be noted that: the harmonic duration is obtained by recording the electrical parameter data stream with a harmonic analyzer. Specifically, the method for obtaining the harmonic duration includes:

[0118] Set the harmonic distortion rate threshold. When THD is greater than or equal to the set harmonic distortion rate threshold, record the start time as t 1 , when THD is less than the set harmonic distortion rate threshold, record the end time as t 2 ;

[0119] Calculate the harmonic duration, and its calculation formula is Sj = t 2 -t 1 .

[0120] Step b42: Normalize the harmonic feature data and calculate the harmonic feature coefficient. Its calculation formula is:

[0121] XBT = H b ×XB 1 +THD×XB 2 +Sj×XB 3 ;

[0122] In the formula, XBT represents the harmonic feature coefficient, Sj represents the harmonic duration; XB 1 , XB 2 and XB 3is a weighting factor, and all the weighting factors are set by those skilled in the art according to experience.

[0123] In implementation, the method for determining whether to mark an electrical device as a second-harmonic abnormal power user includes:

[0124] If the harmonic characteristic coefficient is greater than or equal to a preset first coefficient threshold, mark the electrical device as a second-harmonic abnormal power user;

[0125] If the harmonic characteristic coefficient is less than the preset first coefficient threshold, do not process the electrical device.

[0126] It should be noted that: the preset first coefficient threshold is calculated by those skilled in the art based on multiple sets of fingerprint feature data of historical abnormal power users and after normalizing the multiple sets of fingerprint feature data. The normalization method is the same as that of the harmonic feature data normalization, and will not be elaborated here. The fingerprint feature data is pre-stored in the harmonic fingerprint database, and the harmonic fingerprint database is an SQL database or a NoSQL database.

[0127] In another preferred embodiment, the method for determining whether to mark an electrical device as a second-harmonic abnormal power user further includes:

[0128] Pre-store multiple sets of fingerprint feature data of historical electricity theft devices in the harmonic fingerprint database to form a matching template;

[0129] The matching template is: device type - harmonic component amplitude - harmonic distortion rate - harmonic duration.

[0130] Perform similarity matching between the harmonic feature data and the fingerprint feature data to obtain a similarity value;

[0131] Preset a similarity threshold, compare the similarity value with the preset similarity threshold. If the similarity value is greater than or equal to the preset similarity threshold, mark the electrical device as a second-harmonic abnormal power user; if the similarity value is less than the preset similarity threshold, do not process the electrical device.

[0132] It should be noted that: the calculation method of the similarity value includes the cosine similarity algorithm or the Euclidean distance algorithm.

[0133] Step 4: Obtain long-term detection data based on the electrical parameter data stream, analyze the long-term detection data, and determine whether to mark the electrical device as a long-term abnormal power user;

[0134] It should be noted that: the long-term detection data includes active power, reactive power, and power factor;

[0135] Please refer to Figure 3 As shown, in implementation, the method for analyzing the long-term detection data includes:

[0136] Step c1: Extract the long-term load factor during the historical load monitoring process, and mark it as the historical long-term load factor. Use the extracted historical long-term load factor to establish a load time series set. The load time series set includes i historical long-term load factors. The time intervals for collecting the i historical long-term load factors are equal, and the i historical long-term load factors correspond to a long-term analysis period. The reference value for the duration of the long-term analysis period is from 1 hour to 30 days.

[0137] Among them, the long-term analysis period can be in hours or days, or can be set by those skilled in the art according to actual experience, and no specific limitation is made here.

[0138] It should be noted that: The method for obtaining the long-term load factor includes:

[0139] Perform formulaic calculations on the active power, reactive power, and power factor in the long-term detection data to obtain the long-term load factor. Its calculation formula is:

[0140]

[0141] In the formula, CFX represents the long-term load factor, and W represents the reactive power. The reactive power W reflects the reactive component generated by the electric energy in the electrical equipment, usually caused by inductive or capacitive loads.

[0142] Step c2: Input the historical long-term load factors in the load time series set into the load prediction model to predict the long-term load factor at the future T moment.

[0143] Among them, the training method of the load prediction model includes:

[0144] Those skilled in the art preset the sliding step size and the sliding window length according to actual experience. Convert the historical long-term load factors in the load time series set into multiple training samples using the sliding window method. Use the training samples as the input of the recurrent neural network model, predict the long-term load factor after the sliding step size as the output, and use the subsequent long-term load factors of each training sample as the prediction target. Use the preset accuracy rate as the training target to train the recurrent neural network model. After training, obtain the load prediction model. Predict the long-term load factor at the future T moment according to the historical long-term load factors in the load time series set. Among them, the recurrent neural network model can be an RNN neural network model.

[0145] In implementation, the method for determining whether an electrical equipment is marked as long-term abnormal power consumption includes:

[0146] Step d1: Compare the predicted long-term load factor CFX at the future T moment t+1 with the preset load factor threshold FX.

[0147] Further elaboration is as follows. The long-term load factor at future time T predicted by the load prediction model, that is, the long-term load factor that has not actually occurred. At this time, the long-term load factor at future time T is retrieved and marked as CFX t+1 , a preset load factor threshold is set and marked as FX;

[0148] Step d2: Determine whether to mark the electrical equipment at future time T;

[0149] If CFX t+1 ≥FX, then mark the electrical equipment at future time T as long-term abnormal power consumption;

[0150] If CFX t+1 <FX, then do not process the electrical equipment at future time T, and continue to monitor the electrical equipment at future time T.

[0151] Step d3: Based on the long-term abnormal power consumption, record the long-term abnormal electrical parameter values and long-term timestamps.

[0152] It should be noted that: the long-term abnormal electrical parameter values refer to the active power, reactive power, and power factor data related to the abnormal behavior of electrical equipment recorded during long-term monitoring and analysis.

[0153] Step 5: According to the short-term abnormal power consumption, harmonic abnormal power consumption, and long-term abnormal power consumption, obtain abnormal feature data, and input the abnormal feature data into a pre-constructed abnormal power consumption analysis model to obtain an analysis result;

[0154] The abnormal feature data includes the short-term timestamps and mutated electrical parameter values recorded by the short-term abnormal power consumption, the harmonic abnormal electrical parameter values and harmonic timestamps recorded by the harmonic abnormal power consumption, and the long-term abnormal electrical parameter values and long-term timestamps recorded by the long-term abnormal power consumption; the analysis results include electricity theft abnormal power consumption, equipment failure, and line failure.

[0155] In implementation, the training method of the abnormal power consumption analysis model includes:

[0156] Pre-collect K sets of abnormal feature data, where K is an integer greater than 1. Set corresponding prediction results for the abnormal feature data, and set different letter labels for electricity theft abnormal power consumption, equipment failure, and line failure. Exemplarily, set the letter label for electricity theft abnormal power consumption as Z 1 , set the letter label for equipment failure as Z 2 , set the letter label for line failure as Z 3 ; the prediction results corresponding to the abnormal feature data are obtained by those skilled in the art during the diagnosis of historical abnormal feature data. Those skilled in the art set the corresponding prediction results for the K sets of different abnormal feature data in sequence according to the actual situation;

[0157] Mark the letter label of the prediction result as the prediction label, and convert the abnormal feature data and the corresponding prediction label into a corresponding set of feature vectors;

[0158] Take each set of feature vectors as the input of the abnormal power consumption analysis model. The abnormal power consumption analysis model takes a set of prediction labels corresponding to each set of abnormal feature data as the output, and takes the actual prediction label corresponding to each set of abnormal feature data as the prediction target. The actual prediction label is the pre-set prediction label corresponding to the abnormal feature data; take minimizing the sum of the prediction errors of all abnormal feature data as the training target; train the abnormal power consumption analysis model until the sum of the prediction errors reaches convergence and then stop training. The abnormal power consumption analysis model is specifically a deep neural network model.

[0159] It should be noted that: the judgment rules for setting different letter labels for electricity theft abnormal power consumption, equipment failures, and line failures include:

[0160] During short-time stamp analysis, if the current suddenly increases, the voltage slightly drops, and the power factor is significantly low for a short time within the short-time stamp, it is determined that there is an electricity theft abnormal power consumption behavior.

[0161] If the current suddenly decreases and the voltage slightly increases within the short-time stamp, it is determined that there is an electrical equipment failure (such as a sudden disconnection of the load or abnormal disconnection of the equipment from operation).

[0162] During harmonic time stamp analysis, if within the harmonic time stamp, the total harmonic distortion rate significantly exceeds the normal threshold, and the amplitudes of specific harmonics (such as the 3rd and 5th harmonics) increase significantly, it is determined that there is an access of electricity theft equipment.

[0163] If the harmonic abnormal state lasts for a long time, it indicates that there is equipment damage or power grid quality problems, and the cause should be further analyzed and determined as an equipment failure or a line failure.

[0164] During long-term time stamp analysis, if within the long-term time stamp, the long-term load factor is close to or equal to the preset first coefficient threshold, and the low power factor operation time is too long, it is determined that there is an electricity theft abnormal power consumption behavior (such as an illegal load with long-term inefficient operation).

[0165] If the load curve suddenly shows obvious abnormalities (such as the current and voltage fluctuations do not conform to the load pattern), it is determined that there may be a line failure (such as line aging or sudden damage).

[0166] In this embodiment, by obtaining the data streams of electrical parameters such as the current value, voltage value, active power, and reactive power of the electrical equipment, and combining multi-level analysis of instantaneous detection, harmonic detection, and long-term detection, different types of abnormal electricity consumption behaviors can be quickly and accurately identified. Instantaneous detection can analyze the instantaneous current, voltage rate, and power factor in real time through a sliding window to quickly identify short-term abnormal electricity consumption; harmonic detection is based on fast Fourier transform and harmonic feature analysis, which can effectively capture the harmonic anomalies caused by the access of electricity theft equipment; long-term detection combines historical load data and prediction models to early warn potential long-term abnormal electricity consumption problems. Through multi-dimensional detection methods, this method can not only identify electricity theft behaviors, but also accurately classify equipment failures and line failures, improving the comprehensiveness and reliability of abnormal electricity consumption analysis.

[0167] This implementation can extract high-dimensional features from multiple groups of abnormal feature data of electrical parameters to complete the task of abnormal electricity consumption classification in complex scenarios. During model training, by combining the timestamp of short-term abnormal records with the sudden change value of electrical parameters, the distortion rate and duration of harmonic abnormal records, the load coefficient of long-term abnormal records, etc., a comprehensive feature vector is formed, significantly improving the accuracy and robustness of the analysis. In addition, the abnormal electricity consumption analysis model is based on the harmonic characteristics and matching templates of historical electricity theft behaviors, and can further optimize the classification results through similarity determination methods to ensure the high efficiency of electricity theft behavior identification. This method not only improves the intelligent level of abnormal electricity consumption detection, but also can continuously optimize the prediction effect through model training to achieve automatic and accurate analysis of abnormal electricity consumption.

[0168] Embodiment 2

[0169] Please refer to Figure 4 As shown in the figure, this embodiment provides an abnormal electricity consumption detection device based on dynamic tracking of electrical parameters. The device includes a data acquisition module, an instantaneous detection module, a harmonic detection module, a long-term detection module, and a comprehensive analysis module; each module is connected by wired and / or wireless means to realize data transmission between modules;

[0170] The data acquisition module obtains the data stream of electrical parameters of the nth electrical equipment in the target area, and the data stream of electrical parameters includes current value, voltage value, active power, and reactive power;

[0171] The instantaneous detection module obtains instantaneous detection data based on the data stream of electrical parameters, analyzes the instantaneous detection data, and determines whether to mark the electrical equipment as short-term abnormal electricity consumption;

[0172] The harmonic detection module obtains harmonic detection data based on the data stream of electrical parameters, analyzes the harmonic detection data, and determines whether to mark the electrical equipment as harmonic abnormal electricity consumption;

[0173] The long-term detection module obtains long-term detection data based on the electrical parameter data stream, analyzes the long-term detection data, and determines whether to mark the electrical equipment as long-term abnormal power consumption;

[0174] The comprehensive analysis module obtains abnormal feature data according to short-term abnormal power consumption, harmonic abnormal power consumption and long-term abnormal power consumption, and inputs the abnormal feature data into a pre-constructed abnormal power consumption analysis model to obtain an analysis result.

[0175] As mentioned above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered by the protection scope of this application. Therefore, the protection scope of this application shall be subject to the protection scope of the claims.

[0176] Finally: The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for detecting abnormal power consumption based on dynamic tracking of power parameters, characterized in that: include: Acquire an electrical parameter data stream of an nth electrical device in a target area, wherein the electrical parameter data stream includes a current value, a voltage value, an active power, and a reactive power; Acquire instantaneous detection data based on the electrical parameter data stream, analyze the instantaneous detection data, and determine whether to mark the electrical equipment as short-term abnormal power consumption; Obtain harmonic detection data based on the electrical parameter data stream, analyze the harmonic detection data, and determine whether to mark the electrical equipment as abnormal harmonic power consumption; Acquire long-term detection data based on the electrical parameter data stream, analyze the long-term detection data, and determine whether to mark the electrical equipment as long-term abnormal power consumption; Abnormal feature data is obtained based on short-term abnormal power consumption, harmonic abnormal power consumption and long-term abnormal power consumption, and the abnormal feature data is input into a pre-built abnormal power consumption analysis model to obtain analysis results.

2. The power consumption anomaly detection method based on dynamic tracking of power parameters according to claim 1 is characterized in that: The instantaneous detection data includes instantaneous voltage, instantaneous current and power factor; Methods for analyzing instantaneous detection data include: Step a1: Calculate the instantaneous current rate ΔI, the instantaneous voltage rate ΔU and the real-time power factor PF through a sliding window; Step a2: Set the instantaneous current rate threshold ΔI th , instantaneous voltage rate threshold ΔU th And the power factor lower limit threshold PF th ; The instantaneous current rate ΔI and the instantaneous current rate threshold ΔI th For comparison, the instantaneous voltage rate ΔU and the instantaneous voltage rate threshold ΔU th Compare the real-time power factor PF with the power factor lower limit threshold PF th Make a comparison; If ΔI>ΔI th And ΔU<ΔU th And PF<PF th , then the current power-consuming equipment is marked as short-term abnormal power consumption; If ΔI≤ΔI th And ΔU ≥ ΔU th And PF ≥ PF th , then no processing will be done on the current electrical equipment; Step a3: Record the short-term timestamp and sudden change power parameter value according to the short-term abnormal power consumption.

3. The power consumption anomaly detection method based on dynamic tracking of power parameters according to claim 2 is characterized in that: Harmonic detection data includes current, voltage values ​​and harmonic distortion rate; The abnormal harmonic power consumption includes the first abnormal harmonic power consumption and the second abnormal harmonic power consumption; Methods for analyzing harmonic detection data based on harmonic analysis include: Step b1: Use fast Fourier transform to perform spectrum analysis on the current value and extract the amplitude of the harmonic component H b ; Step b2: According to the harmonic component amplitude H b Calculate the harmonic distortion THD; Step b3: preset a distortion rate threshold, compare the harmonic distortion rate with the preset distortion rate threshold, if the harmonic distortion rate is greater than or equal to the preset distortion rate threshold, mark the electrical equipment as first harmonic abnormal power consumption; if the harmonic distortion rate is less than the preset distortion rate threshold, generate an analysis instruction; Step b4: receiving an analysis instruction, obtaining a harmonic characteristic coefficient, comparing the harmonic characteristic coefficient with a preset first coefficient threshold, and determining whether to mark the electrical equipment as second harmonic abnormal power consumption; Step b5: Record the harmonic abnormal power parameter value and the harmonic timestamp according to the first harmonic abnormal power consumption or the second harmonic abnormal power consumption.

4. The power consumption anomaly detection method based on dynamic tracking of power parameters according to claim 3 is characterized in that: Methods for obtaining harmonic characteristic coefficients include: Step b41: Acquire harmonic characteristic data, wherein the harmonic characteristic data includes harmonic component amplitude, harmonic distortion rate and harmonic duration; Step b42: normalize the harmonic characteristic data and calculate the harmonic characteristic coefficient XBT.

5. The power consumption anomaly detection method based on dynamic tracking of power parameters according to claim 4 is characterized in that: Methods for determining whether to mark an electrical device as abnormal second harmonic power consumption include: If the harmonic characteristic coefficient is greater than or equal to the preset first coefficient threshold, the electrical equipment is marked as abnormal second harmonic power consumption; If the harmonic characteristic coefficient is less than the preset first coefficient threshold, no processing is performed on the electrical equipment.

6. The method for detecting abnormal power consumption based on dynamic tracking of power parameters according to claim 4, characterized in that: The method for obtaining the harmonic duration includes: Set the harmonic distortion rate threshold. When THD is greater than or equal to the set harmonic distortion rate threshold, the recording start time is t1, and when THD is less than the set harmonic distortion rate threshold, the recording end time is t2; Calculate the harmonic duration Sj; The method for determining whether to mark the electrical equipment as second harmonic abnormal electrical consumption also includes: Pre-store multiple groups of fingerprint feature data recorded by historical electricity theft devices in the harmonic fingerprint library to form a matching template; The matching template is: device type - harmonic component amplitude - harmonic distortion rate - harmonic duration; Perform similarity matching between the harmonic feature data and the fingerprint feature data to obtain a similarity value; A similarity threshold is preset, and the similarity value is compared with the preset similarity threshold. If the similarity value is greater than or equal to the preset similarity threshold, the electrical equipment is marked as second harmonic abnormal power consumption; if the similarity value is less than the preset similarity threshold, the electrical equipment is not processed.

7. The method for detecting abnormal power consumption based on dynamic tracking of power parameters according to claim 6, characterized in that: The long-term detection data includes active power, reactive power and power factor; Methods for analyzing long-term test data include: Step c1: extract the long-term load factor in the historical load monitoring process and mark it as the historical long-term load factor, use the extracted historical long-term load factor to establish a load time series set, the load time series set includes i historical long-term load factors, the time intervals of i historical long-term load factors are equal, and i historical long-term load factors correspond to a long-term analysis period; The method for obtaining the long-term load factor includes: The active power, reactive power and power factor in the long-term detection data are calculated by formulating to obtain the long-term load factor CFX; Step c2: Input the historical long-term load factor in the load time series set into the load forecasting model to predict the long-term load factor at the future time T.

8. The method for detecting abnormal power consumption based on dynamic tracking of power parameters according to claim 7, characterized in that: Methods for determining whether to mark an electrical device as having long-term abnormal power consumption include: Step d1: The predicted long-term load factor CFX at time T in the future t+1 Compare with the preset load factor threshold FX; Step d2: Determine whether to mark the electrical equipment at the future time T; If CFX t+1 ≥FX, the power-consuming equipment at time T in the future will be marked as long-term abnormal power consumption; If CFX t+1 <FX, the power-consuming equipment at the next T time will not be processed, and the power-consuming equipment at the next T time will continue to be monitored; Step d3: Recording long-term abnormal electricity parameter values ​​and long-term timestamps based on the long-term abnormal electricity consumption.

9. The method for detecting abnormal power consumption based on dynamic tracking of power parameters according to claim 8, characterized in that: The abnormal characteristic data include short-term timestamps and sudden change electrical parameter values ​​of short-term abnormal power consumption records, harmonic abnormal power parameter values ​​and harmonic timestamps of harmonic abnormal power consumption records, and long-term abnormal power parameter values ​​and long-term timestamps of long-term abnormal power consumption records; The analysis results include abnormal electricity use, equipment failure, and line failure; The training method of the abnormal power consumption analysis model includes: Collect K groups of abnormal feature data in advance, where K is an integer greater than 1, set corresponding prediction results for the abnormal feature data, set different letter labels for abnormal electricity use such as electricity theft, equipment failure, and line failure, mark the letter labels of the prediction results as prediction labels, and convert the abnormal feature data and the corresponding prediction labels into a corresponding set of feature vectors; Each group of feature vectors is used as the input of the abnormal power consumption analysis model. The abnormal power consumption analysis model takes a group of prediction labels corresponding to each group of abnormal feature data as output, and takes the actual prediction labels corresponding to each group of abnormal feature data as prediction targets, where the actual prediction labels are the pre-set prediction labels corresponding to the abnormal feature data; minimizing the sum of prediction errors of all abnormal feature data is used as the training target; the abnormal power consumption analysis model is trained until the sum of prediction errors converges and the training is stopped. The abnormal power consumption analysis model is a deep neural network model.

10. An abnormal power consumption detection device based on dynamic tracking of electrical parameters, used to implement the abnormal power consumption detection method based on dynamic tracking of electrical parameters as claimed in any one of claims 1 to 9, characterized in that: include: A data acquisition module is used to obtain an electrical parameter data stream of the nth electrical device in the target area, wherein the electrical parameter data stream includes current value, voltage value, active power and reactive power; The instantaneous detection module obtains instantaneous detection data based on the electrical parameter data stream, analyzes the instantaneous detection data, and determines whether to mark the electrical equipment as short-term abnormal power consumption; The harmonic detection module obtains harmonic detection data based on the electrical parameter data stream, analyzes the harmonic detection data, and determines whether to mark the electrical equipment as abnormal harmonic power consumption; The long-term detection module obtains long-term detection data based on the electrical parameter data stream, analyzes the long-term detection data, and determines whether to mark the electrical equipment as long-term abnormal power consumption; The comprehensive analysis module obtains abnormal characteristic data based on short-term abnormal power consumption, harmonic abnormal power consumption and long-term abnormal power consumption, and inputs the abnormal characteristic data into a pre-built abnormal power consumption analysis model to obtain analysis results.

Images