CN115423128A - Non-invasive abnormal load behavior monitoring method, electronic equipment and storage medium - Google Patents
Non-invasive abnormal load behavior monitoring method, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN115423128A CN115423128A CN202211066557.6A CN202211066557A CN115423128A CN 115423128 A CN115423128 A CN 115423128A CN 202211066557 A CN202211066557 A CN 202211066557A CN 115423128 A CN115423128 A CN 115423128A
- Authority
- CN
- China
- Prior art keywords
- load
- monitoring
- image
- training
- network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 141
- 238000000034 method Methods 0.000 title claims abstract description 69
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 43
- 238000003860 storage Methods 0.000 title claims abstract description 11
- 238000012549 training Methods 0.000 claims abstract description 69
- 238000006243 chemical reaction Methods 0.000 claims abstract description 9
- 238000012806 monitoring device Methods 0.000 claims abstract description 6
- 238000012545 processing Methods 0.000 claims abstract description 5
- 230000006870 function Effects 0.000 claims description 57
- 239000011159 matrix material Substances 0.000 claims description 39
- 230000006399 behavior Effects 0.000 claims description 25
- 230000007704 transition Effects 0.000 claims description 25
- 239000002775 capsule Substances 0.000 claims description 23
- 238000012795 verification Methods 0.000 claims description 22
- 238000009826 distribution Methods 0.000 claims description 21
- 238000001514 detection method Methods 0.000 claims description 18
- 238000005070 sampling Methods 0.000 claims description 16
- 238000004364 calculation method Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 10
- 230000008859 change Effects 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 9
- 238000010183 spectrum analysis Methods 0.000 claims description 7
- 238000012935 Averaging Methods 0.000 claims description 3
- 230000000977 initiatory effect Effects 0.000 claims description 3
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 5
- 230000032683 aging Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 230000005611 electricity Effects 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 239000003086 colorant Substances 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000003042 antagnostic effect Effects 0.000 description 1
- 230000008485 antagonism Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000007711 solidification Methods 0.000 description 1
- 230000008023 solidification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/20—Administration of product repair or maintenance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- General Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Water Supply & Treatment (AREA)
- Biophysics (AREA)
- Public Health (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Primary Health Care (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Quality & Reliability (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Remote Monitoring And Control Of Power-Distribution Networks (AREA)
Abstract
The invention relates to a non-invasive abnormal load behavior monitoring method, electronic equipment and a storage medium, wherein the method comprises the steps of obtaining real-time monitoring data of non-invasive load monitoring equipment and denoising the data; the monitoring data includes: total voltage and total current data of a preselected power circuit monitored by the non-intrusive load monitoring device; regarding the de-noised monitoring data, taking the monitoring data with load state conversion as effective monitoring data; based on a pre-constructed power strategy, carrying out color coding processing on effective monitoring data to obtain a V-I track image of total voltage and total current corresponding to each power circuit; and inputting the V-I track image into a training condition to generate a countermeasure network, and judging whether each power circuit has abnormal load or not based on the generated feature reconstruction image. The method has the beneficial effects that the technical problems of low expansibility, low flexibility and high monitoring error of non-intrusive load monitoring in the prior art can be solved.
Description
Technical Field
The present invention relates to the field of load monitoring technologies, and in particular, to a non-intrusive abnormal load behavior monitoring method, an electronic device, and a storage medium.
Background
In modern social production life, a large amount of renewable energy is currently generated and utilized at the consumer end, and user behavior can facilitate efficient integration of highly weather-dependent distributed energy. Therefore, it is crucial to observe the activities and actions of the user using electricity. It is particularly important to obtain real-time power consumption information of each electric appliance inside a user.
Different from the current intelligent electric meter and the load total power consumption information acquisition, the load power consumption detail monitoring is to acquire the real-time power consumption information of each electric appliance in the power consumer by a certain technical means, including the working state, the power consumption power and the accumulated electric quantity of the electric appliance, so as to obtain fault information and the like.
In the prior art, load monitoring mainly comprises an invasive type and a non-invasive type, wherein the invasive type load monitoring needs to invade the interior of a power load, install a data measurement sensor with a communication function for each electric appliance respectively, and locally collect and send out electricity utilization information. To achieve the same purpose, the non-invasive load monitoring only needs to install a data measurement sensor with a communication function at a power supply inlet of a power load, and the power consumption information of each electric appliance in the user can be obtained by analyzing the load total data. Compared with an invasive method, the non-invasive method has low cost and convenient installation, and is well applied through detailed data obtained from non-invasive load monitoring.
The existing non-invasive load monitoring is only suitable for the ideal state that the type of the electric load of the power consumer is fixed and does not change due to algorithm solidification, and when the electric equipment is changed or aged, a larger monitoring error is generated, so that the defects of inflexibility and inextensibility exist.
Disclosure of Invention
Technical problem to be solved
In view of the above disadvantages and shortcomings of the prior art, the present invention provides a method, an electronic device, and a storage medium for monitoring non-intrusive abnormal load behavior, which solve the technical problems of low expansibility, low flexibility, and high monitoring error of non-intrusive load monitoring in the prior art.
(II) technical scheme
In order to achieve the above object, in a first aspect, the present invention provides a method for monitoring non-invasive abnormal load behavior, comprising the following steps:
s1, acquiring real-time monitoring data of non-invasive load monitoring equipment and denoising; the monitoring data includes: total voltage and total current data of a preselected power circuit monitored by the non-intrusive load monitoring device;
s2, regarding the de-noised monitoring data, taking the monitoring data with load state conversion as effective monitoring data;
s3, carrying out color coding processing on effective monitoring data based on a pre-constructed power strategy to obtain a V-I track image of each single load in the power circuit;
s4, inputting the V-I track image into a training condition to generate a countermeasure network, and judging whether each power circuit has abnormal load or not based on the generated feature reconstruction image;
wherein the conditionally generating the countermeasure network comprises: the system comprises a conditional self-encoder, a capsule network and a classifier, wherein the conditional self-encoder is used for converting Gaussian prior probability of loads in a V-I track image into Gaussian posterior probability, and the capsule network is used for realizing compactness of the same type of characteristics near a Gaussian distribution center so that the classifier detects the loads.
Optionally, before S1, the method further comprises: s0, training the conditional generation countermeasure network:
the S0 comprises:
s01, acquiring training monitoring data samples and checking monitoring data samples for training the condition generation countermeasure network; the training monitoring data sample and the verification monitoring data sample are historical monitoring total voltage and total current data of the same power circuit;
s02, coding the training monitoring data sample and the verification detection data sample based on a pre-constructed power strategy to obtain a training V-I track image and a verification V-I track image of each single load in the power circuit;
s03, inputting the training V-I track image to the conditional countermeasure generation network to reconstruct and generate a feature reconstruction image of the training V-I track aiming at each single load;
s04, inputting the verification V-I track image and the feature reconstruction image corresponding to each single load into a pre-constructed discriminator respectively, and judging whether the feature reconstruction image is matched with the verification V-I track image or not;
and S05, adjusting training parameters of the conditional countermeasure generation network, and alternately generating a feature reconstruction image and inputting the feature reconstruction image into a discrimination network so as to match the feature reconstruction image finally generated by the conditional countermeasure monitoring network with the verification V-I track image and obtain the trained conditional countermeasure monitoring network.
Optionally, S01 comprises:
acquiring historical monitoring data of at least one load state conversion event in a power circuit monitored by non-invasive load monitoring equipment; the load state transition event is a circuit load transition process caused when a single load in the preselected power circuit is turned on and/or turned off;
calculating a time period of occurrence of a load state transition event based on a predefined event detection window; the method specifically comprises the following steps:
calculating a total real power S of the preselected power circuit t Determining Δ S t >S on1 Time t;
based on a pre-constructed event detection window, calculating and determining the total real power change quantity deltaS when t = t + TR t+TR <S on1 (ii) a R is the step length of the event detection window, delta S t =S t+1 -S t ;
If S t+TR -S t <S on2 Judging that a load state transition event occurs in the time period of t-t + TR;
acquiring total voltage and total current data of T time period periods before and after the load state transition event occurs, and acquiring a training monitoring data sample and a verification monitoring data sample;
said S on1 For a predefined load state transition event initiation threshold, S on2 The end threshold for a predefined load state transition event.
Optionally, S3 includes:
s30, sampling the effective monitoring data based on a pre-constructed spectrum analysis method to obtain the voltage and current values of each single load in the power circuit;
s31, determining the active component current i of the current i (t) of each single load based on a pre-constructed Fryze power strategy a (t) and the reactive component current i f (t);
Based on the active component current i a (t) and the reactive component current i f (t) calculating and obtaining a power factor matrixThe power factor is the ratio of the power of the active component current to the power of the reactive component current;
k is the total number of sampling points, P apparent To be real power, V rms 、I rms Effective values of load voltage and current respectively;
s32, aiming at each single load, constructing a hue matrix of the V-I track based on a pre-constructed HSV color spaceAnd a voltage period matrix V;
s33, connecting the power factor matrix in a standard three-dimensional coordinate system for each single loadHue matrixAnd a voltage period matrix V for acquiring a V-I track image of the single load.
Optionally, the S32 specifically includes;
s321, acquiring the motion direction H of the V-I track by using hue attribute hue based on the HSV color space j ;
Based on the direction of movement H j Storing the hue of the jth sampling point into a2 Nx 2N matrix to obtain a hue matrix
The direction of motion H j The calculation expression is:
the arg is an arc tangent function of four quadrants;
s322, based on the pre-constructed binary image W m (1,2,. Once, M), averaging M cycles of a single load voltage to obtain a voltage cycle matrix V;
the expression of the voltage period matrix V is:
optionally, adjusting the training parameters of the conditional antagonistic monitoring network, in particular,
calculating the minimum value of the training loss of the conditional countermeasure monitoring network based on a loss function constructed in advance;
adjusting the minimum value weighting calculation to generate parameters of a countermeasure network;
the loss functions include feature matching loss functions, reconstruction loss functions, additional encoder loss functions, center constrained loss functions, and/or contrast loss functions.
Optionally, the feature matching loss function expression is:
f (x) is given input V-I track x, the output of the discriminator intermediate layer;
the reconstruction loss function expression is as follows:
the above-mentionedMu is the average intensity of the training V-I track image, delta is the standard deviation of the training V-I track image,reconstructing the covariance of the image for the training V-I trajectory and features; c1 and c2 are constants;
the additional encoder penalty function expression is:
the z is a sampling vector characteristic output by the capsule network when the V-I track image is trained,reconstructing the coding features of the image for the features;
the central constraint loss function expression is as follows:
L KL =d(C,sg[P y ]);
the above-mentionedC is a probability capsule, and P is Gaussian distribution of the target load cluster;
the contrast loss function expression is:
said [ ·] + Is a function of the positive number of the return parameter;
the formula for calculating the minimum value weight is as follows:
L=αL KL +βL rec +γL contr +σL enc +λL adv and the alpha, the beta, the gamma, the sigma and the lambda are constants.
Optionally, the S4 specifically includes:
inputting the real-time V-I track of the power circuit into the trained condition generation countermeasure network to generate a real-time feature reconstruction image;
calculating the minimum distance between the real-time feature reconstruction image and a final training condition to generate a historical feature reconstruction image of the confrontation network;
and if the minimum distance between the real-time characteristic reconstruction image and the historical characteristic reconstruction image is greater than the threshold tau meeting the preset requirement, judging that an abnormal load occurs in the preselected power circuit.
In a second aspect, the present invention provides an electronic device, including a memory and a processor, where the memory stores a computer program, and the processor executes the computer program stored in the memory, so as to implement the steps of the method for monitoring non-invasive abnormal load behavior according to any one of the above first aspects.
In a third aspect, a computer readable storage medium has stored thereon a computer program which, when being executed by a processor, implements the steps of the method for non-intrusive abnormal load behavior monitoring as defined in any of the first aspects above.
(III) advantageous effects
The invention provides a non-invasive abnormal load behavior monitoring method, electronic equipment and a storage medium, wherein in the method, a condition generation countermeasure network is trained in advance, historical load voltage and current values causing load state conversion events are judged and sampled, and color coding is carried out on the voltage and current values by combining a power strategy to obtain a color load V-I track image, so that visual identification is facilitated, the V-I track image and a characteristic reconstruction image are input to the condition generation countermeasure network constructed in advance for multiple times, judgment by a discriminator is repeated, and the condition generation countermeasure network meeting the correct ratio of the preset identification is realized.
And comparing the characteristic reconstructed image of the V-I track image of the voltage and current values acquired in real time with the minimum distance of the historical characteristic reconstructed image required by load presetting, judging whether an abnormal load exists or not, and monitoring the abnormal load of the circuit to be detected.
Compared with the prior art, the technical scheme can realize monitoring according to the actual load of the power consumer, achieves the purpose of flexible monitoring when the power consumer changes the electric appliance, improves the flexibility and expansibility of non-invasive abnormal load behavior monitoring, and reduces the detection error.
Drawings
Fig. 1 is a schematic flow chart of a non-invasive abnormal load behavior monitoring method according to an embodiment of the present invention;
FIG. 2 is a training flow diagram of the conditional generation countermeasure network according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a model of a training conditional generative confrontation network according to an embodiment of the present invention;
fig. 4 is a logic flow diagram of detecting a load state switching event according to an embodiment of the present invention.
Detailed Description
For the purpose of better explaining the present invention and to facilitate understanding, the present invention will be described in detail by way of specific embodiments with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
In the modern society, electric energy becomes one of the most important energy sources in the modern society, and the electricity consumption demand of residents is increasing day by day. Therefore, household energy management is a way capable of effectively reducing power waste, in the prior art, a non-invasive load monitoring method can be only well applied under the condition that the load type is fixed and does not change, so that the invention provides the non-invasive abnormal load behavior monitoring method, and the load monitoring method can effectively cope with various situations of adding and deleting electrical appliances in a household power circuit or aging of the electrical appliances.
As shown in fig. 1, fig. 1 is a method for monitoring a non-invasive abnormal load behavior according to an embodiment of the present invention, in which a data measurement sensor with a communication function is only required to be installed at a power inlet of an electrical load for non-invasive load monitoring, so as to obtain power consumption information of each electrical appliance inside a user by analyzing load total data, the method includes the following steps:
s1, acquiring real-time monitoring data of non-invasive load monitoring equipment and denoising; the monitoring data includes: total voltage and total current data of a preselected power circuit monitored by the non-intrusive load monitoring device.
And S2, regarding the de-noised monitoring data, taking the monitoring data with load state conversion as effective monitoring data.
And S3, carrying out color coding processing on the effective monitoring data based on a pre-constructed power strategy to obtain a V-I track image of each single load in the power circuit.
In practice, the individual loads may be any consumer in a domestic power circuit, and since the shape of the V-I trace depends largely on the load current reflecting the physical characteristics of the power load, the proportion of active current in the power factor is greater than reactive current, resulting in loads of the same class differing less in the shape of the V-I trace, for which reason, in some embodiments, the load current may be broken down into active and reactive current components representing resistive and non-resistive information using the Fryze power strategy, thereby enhancing the uniqueness of the V-I trace.
And S4, inputting the V-I track image into a training condition to generate a countermeasure network, and judging whether each power circuit has abnormal load or not based on the generated feature reconstruction image.
Wherein the conditionally generating the countermeasure network comprises: the system comprises a conditional self-encoder, a capsule network and a classifier, wherein the conditional self-encoder is used for converting Gaussian prior probability of loads in a V-I track image into Gaussian posterior probability, and the capsule network is used for realizing compactness of the same type of characteristics near a Gaussian distribution center so that the classifier detects the loads.
According to the non-invasive abnormal load behavior monitoring method provided by the embodiment of the invention, the real-time monitoring data of non-invasive load monitoring equipment is obtained in real time, the V-I track image is generated, the V-I track image is input into a pre-trained condition generation countermeasure network, whether abnormal load exists or not is judged, the abnormal load is monitored, the influence of change and aging of electric equipment is avoided, the expansibility is good, the use is flexible, and the applicability is high.
Specifically, in the foregoing method for monitoring non-intrusive abnormal load behavior, S3 implemented in another embodiment may include:
s30, sampling the effective monitoring data based on a pre-constructed spectrum analysis method to obtain the voltage and current values of each single load in the power circuit;
s31, determining the active component current i of i (t) of each single load based on a pre-constructed Fryze power strategy aiming at each single load a (t) and the reactive component current i f (t);
Based on the active component current i a (t) and the reactive component current i f (t) calculating and obtaining a power factor matrix
In the present embodiment, the active current is defined as the orthogonal projection of the load current in the direction of the voltage v (t), i.e. i a (t) is proportional to v (t), and conveys resistance information, the active current i a (t) and active Power P active The expression of (a) is:
wherein, V rms Is the effective value of voltage, and T is the power supply period. The reactive component current and voltage are orthogonal, and the instantaneous voltage and current of the load can be used to the reactive current i f (t) represents:
in practical application, the proportion of active current is large, so thatSo that loads of the same class differ less in the shape of the V-I trajectory and therefore only the reactive current I can be used f (t) replacing the non-component current data i (t) based on said reactive current i f (t) obtaining a V-I trajectory while avoiding losing information between the real and reactive components, saturation may be used to represent the ratio of real power to reactive power, i.e. the power factor, over a number of cycles.
Namely, the power factor is the ratio of the power of the active component current to the power of the reactive component current.
k is the total number of sampling points, P apparent To be real power, V rms 、I rms The effective values of the load voltage and the current are respectively.
S32, aiming at each single load, constructing a hue matrix of the V-I track based on a pre-constructed HSV color spaceAnd a voltage period matrix V.
In a specific implementation, the S32 may include:
s321, aiming at a single load, acquiring the motion direction H of the V-I track by utilizing the hue attribute hue based on the HSV color space j ;
Based on the direction of movement H j Storing the hue of the jth sampling point into a new 2 Nx 2N matrix to obtain a hue matrix
The direction of motion H j The calculation expression is:
the arg is an arc tangent function of four quadrants; and calculating the phase angle of two continuous points in the V-I track, wherein the value range is 0-360 degrees.
S322, based on the pre-constructed binary image W m (1,2,... M), averaging M periods of voltage to obtain a voltage period matrix V; namely, the color generation attribute Value using the HSV color space is used to represent the repeatability of the V-I trajectory.
The expression of the voltage period matrix V is:
wherein M =1,2,3 · M.
S33, connecting the power factor matrix in a standard three-dimensional coordinate system for each single loadHue matrixAnd a voltage period matrix V, acquiring a V-I track image of the single load.
In one embodiment, the value of M is preferably 10, and is determined according to practical situations when applied, and is not limited herein.
The HSV (hue, saturation, lightness) color space is a non-linear transformation of RGB (red, green, blue color space) that better conforms to human perception of color. The HSV color space may be represented using an inverse cone model, with the colors of each hue being distributed in radial slices from red to yellow, green, cyan, blue, magenta, the hue being used to represent the category of the color. Saturation is defined as the ratio of color to brightness and increases with increasing distance from the center of the circular cross-section to represent the vividness of the color. Lightness represents brightness, and the brightness of each color is represented by the distance from the center of a circle to the vertex of a cone.
In one embodiment, the step S33 is implemented by using the power factor matrixHue matrixAnd a voltage period matrix V connected along a third dimension to convert the hue-saturation-value to the equivalent of red-green-blue, etc., so that the created color image can be perceived by humans.
Of course, in other embodiments, other converted colors may be included, and are not limited herein.
In some other embodiments, the S4 may specifically include:
and inputting the real-time V-I track of the power circuit into the trained condition generation countermeasure network to generate a real-time feature reconstruction image.
And calculating the minimum distance between the real-time feature reconstruction image and the final training condition to generate the historical feature reconstruction image of the countermeasure network.
And if the minimum distance between the real-time characteristic reconstruction image and the historical characteristic reconstruction image is larger than the threshold tau meeting the preset requirement, judging that an abnormal load occurs in the pre-selection power circuit.
In practical application, the type of the input load can be further determined by the threshold, that is, if the minimum distance between the real-time feature reconstructed image and the historical feature reconstructed image is smaller than the threshold τ meeting the preset requirement, the type of the input load is determined to be the same as the load type corresponding to the minimum distance; and if the minimum distance between the real-time characteristic reconstructed image and the historical characteristic reconstructed image is equal to the threshold tau meeting the preset requirement, judging that the input load type is an unknown load.
In some embodiments, when the minimum distance between the real-time feature reconstruction image and the historical feature reconstruction image is greater than or equal to the threshold τ meeting the preset requirement, it is determined that there is an abnormal load.
According to the non-invasive abnormal load behavior monitoring method provided by the embodiment of the invention, the active component and the reactive component of the load current are separated, and the reactive component can be only taken as the original load current to generate a V-I track with larger difference.
In some other embodiments, such as initial use of the power circuit of the non-intrusive load monitoring device, prior to S1, the method may further comprise: s0, training the conditionally generated confrontation network, as shown in fig. 2, fig. 2 is a training flowchart of the conditionally generated confrontation network according to an embodiment of the present invention.
The S0 may include:
s01, acquiring training monitoring data samples and checking monitoring data samples for training the condition generation countermeasure network; the training monitoring data sample and the verification monitoring data sample are historical monitoring total voltage and total current data of the same power circuit;
specifically, in some other embodiments, the training monitoring data samples and the verification monitoring data samples may be non-intrusive load monitoring public data of a preselected power circuit, noise exists in raw data, the extraction of load characteristics is affected, and in order to facilitate subsequent characteristic extraction, the data samples are generally subjected to denoising processing.
In a specific embodiment, the S01 is implemented as:
acquiring historical monitoring data of at least one load state conversion event in a power circuit monitored by non-invasive load monitoring equipment; the load state transition event is a circuit load transition process caused by the switching on and/or off of a device in the preselected power circuit.
Calculating a time period of occurrence of a load state transition event based on a predefined event detection window; the method specifically comprises the following steps:
calculating a total real power S of the preselected power circuit t Determining Δ S t >S on1 Time t;
based on a pre-constructed event detection window, calculating and determining the total real power change quantity deltaS when t = t + TR t+TR <S on1 (ii) a R is the step length of the event detection window, delta S t =S t+1 -S t ;
If S t+TR -S t <S on2 And judging that a load state transition event occurs in the time period of t-t + TR.
And collecting total voltage and total current data of T time period periods before and after the load state transition event occurs, and acquiring training monitoring data samples and verification detection data samples.
S is on1 For a predefined load state transition event initiation threshold, S on2 Is a predefined load state transition event end threshold.
And S02, coding the training monitoring data samples and the verification detection data samples based on a pre-constructed power strategy, and acquiring a training V-I track image and a verification V-I track image of each single load in the power circuit.
In practical operation, the processes of generating the training V-I track image and verifying the training V-I track image in S02 may be the same as the sub-steps of generating the real-time V-I track image in the above embodiment.
And S03, inputting the training V-I track image to the conditional countermeasure generation network to reconstruct and generate a feature reconstruction image of the training V-I track aiming at each single load.
And S04, respectively inputting the verification V-I track image and the feature reconstruction image corresponding to each single load into a pre-constructed discriminator, and judging whether the feature reconstruction image is matched with the verification V-I track image.
And S05, adjusting training parameters of the conditional countermeasure generation network, and alternately generating a feature reconstruction image and inputting the feature reconstruction image into a discrimination network so as to match the feature reconstruction image finally generated by the conditional countermeasure monitoring network with the verification V-I track image and obtain the trained conditional countermeasure monitoring network.
In some embodiments, in S05, adjusting the training parameters of the conditional countermeasure monitoring network may be implemented as:
calculating the minimum value of the training loss of the conditional countermeasure monitoring network based on a loss function constructed in advance;
adjusting the parameters of the conditional generation countermeasure network by the minimum value weighting calculation;
in a particular embodiment, the loss function may include a feature matching loss function, a reconstruction loss function, an additional encoder loss function, a center constrained loss function, and/or a contrast loss function, among others.
Specifically, the feature matching loss function is used for counterlearning, is used for reducing instability of conditional generation counternetwork training, aligns the conditional generation counternetwork coding generation feature distribution and the generated V-I track feature reconstruction image with a real V-I track image, and can effectively distinguish feature representations of known equipment and unknown equipment on the basis that the generated V-I track is enough to deceive a discriminator. In particular, the generator is updated according to the internal representation of the discriminator. Formally, let f be a function of a given input V-I trajectory x output discriminator intermediate layer drawn from the input data distribution, and feature matching calculates L between the feature representation of the original V-I trajectory image and the generated V-I trajectory image, respectively 2 Distance.
The feature matching loss function, namely the antagonism loss function expression is as follows:
and f (x) is the output of the discriminator intermediate layer according to the input V-I track x.
In another embodiment, the problem that the reconstruction loss between the input and the reconstructed generated V-I track image is measured to obtain a credible reconstruction result without optimizing the use of the context information of the input V-I data can be solved, the generation process based on the V-I track image contains abundant structural information, and in some embodiments, the structural similarity loss is used as the reconstruction loss of the generator, takes the brightness, the contrast and the structural information into consideration, is less sensitive to the position offset of the input V-I track and the reconstruction thereof, and therefore the network is easier to converge. Therefore, the model trained by the structural similarity loss tends to focus more on global information than local features in the V-I trajectory reconstruction process.
The reconstruction loss function, namely the structural similarity loss function expression is as follows:
the above-mentionedMu is the average intensity of the training V-I track image, delta is the standard deviation of the training V-I track image,reconstructing the covariance of the image for the training V-I trajectory and features; c1, c2 are constants, and in some embodiments, the constants c1 and c2 are set to 0.01 and 0.03, respectively.
Based on the two loss functions described above, the generator can be forced to produce an image that is both real and linked to contextual information.
Further, in other embodiments, additional encoder losses L may also be utilized 1 To minimize sampled vector features from capsule network output at input z and coding features to reconstruct V-I trajectory imagesBased on which the conditional generation countermeasure network learns how to encode the V-I trajectory features of the known load samples, both the generator and the additional encoder network are optimized only for the data samples of the known load.
The additional encoder penalty function expression is:
the z is a sampling vector characteristic output by the capsule network when the V-I track image is trained,the encoding characteristics of the image are reconstructed for the characteristics.
In other embodiments, in order to encode the V-I trajectory features of each type of known power circuit load, each type of load features is formed into a compact cluster, making it easier for the model to identify unknown load features. It is also possible to push the probability capsules C towards the target load clusters P using a central constraint penalty in the generator's potential space y The center of (a), concentrates the density of all known load samples in the target area.
The central constraint loss function expression is as follows:
L KL =d(C,sg[P y ]);
the function sg · represents the stopping gradient operator, which is defined as an identity in the forward calculation and has zero partial derivative, limiting its parameters to an un-updated constant.
The above-mentionedC is a probability capsule, and P is Gaussian distribution of the target load cluster; the probability capsule isThe Gaussian distribution of the V-I trajectory is input.
In one embodiment, a contrast Loss function is also constructed using a margin Loss function of margin Loss and margin m k The boundary pushes all target loads not belonging to y far away from the distribution C by considering P ≠y Is P y Avoids the condition generation from collapsing against the previous target load of the network, facilitating the separation between the load and all other loads (possibly unknown corresponding loads).
The contrast loss function expression is:
said [ ·] + Is a function that returns a positive number of parameters.
Based on the above five loss functions, the formula for updating the parameters of the network by performing weighted combination calculation on the minimum value of the five loss functions is as follows:
L=αL KL +βL rec +γL contr +σL enc +λL adv 。
the α, β, γ, σ and λ are all constants, and in one embodiment, α is preferably 1, β is preferably 0.01, γ is preferably 1, σ is preferably 0.01, and λ is preferably 10.
In the embodiment, a conditional autoencoder and a capsule network are used as generators for generating a countermeasure network by using conditions, in the model training process, capsule characteristics of the same type of load are matched with a predetermined Gaussian distribution, a Gaussian distribution is defined for each type of load, specifically, a variational autoencoder framework is used, and a group of Gaussian priors are used as approximations of posterior distribution, so that the compactness of the same type of characteristics near the center of the Gaussian distribution can be controlled, and the capability of a classifier for detecting unknown loads can be controlled. The V-I track generated by the generator respectively passes through the discriminator and an additional coding network, the additional coding network maps the generated V-I track to the hidden layer characterization to minimize the distance between the hidden layer characterization and the V-I track in the generator, and the model parameters are adjusted by constructing a plurality of loss functions and calculating the minimum value, so that the generator can better learn the distribution characteristics of the known load, the capability of monitoring the unknown load is improved, the monitoring flexibility is high, and the error is small.
Fig. 3 is a schematic diagram of a model of a trained condition-generating countermeasure network according to an embodiment of the present invention, and in the embodiment shown in fig. 3, the condition-generating countermeasure network includes a condition self-encoder and a capsule network, and further includes an additional encoder. Each load has a non-independent gaussian distribution, and the conditional autoencoder approximates the gaussian priors of the loads to a posterior probability. The additional encoder is used to map the generated V-I track to hidden layer representations to minimize and conditionally derive the distance between the hidden layer representations of the V-I track in the encoder. The capsule network is used to achieve compactness of the same class of features near the center of the gaussian distribution, thereby controlling the classifier's ability to detect unknown loads.
In order to better explain the technical solution proposed by the present invention, the following detailed description is made with reference to a specific embodiment.
The present embodiment is the monitoring of abnormal load behavior of a power circuit of a home using a non-intrusive load monitoring device. In the household power circuit, each electrical consumer/equipment/appliance acts as a single load. The load state switching process is accompanied by actual power change, and the motor type load is often accompanied by power and current effective value change when starting, the load state is changed, and the process is regarded as a load state switching event. The variation of the effective value of the power or the current can be compared with a preset threshold value, and if the variation is larger than the threshold value, an event is judged to occur. The voltage and current values of the load causing the event can be obtained from the changes of the voltage and current before and after the event.
Firstly, training a condition generating network; the network is trained using only the V-I trajectories of known load.
The method comprises the following steps:
a1, obtaining original monitoring data of non-invasive load monitoring equipment, and denoising the original monitoring data;
and A2, calculating the time period of the load state transition event based on a predefined event detection window.
As shown in fig. 4, fig. 4 is a schematic logic flow diagram of detecting a load state switching event according to this embodiment.
In this embodiment, the detecting the load state switching event is specifically implemented as:
defining the step length of the time detection window as R, S t Represents the total real power at t seconds, Δ S t =S t+1 -S t And represents the total real power variation. When Δ S t >S on1 At that time, the event detection window starts to move and Δ S is calculated t+1 ,ΔS t+2 …, up to Δ S t+TR <S o1n . If S is t+TR -S t <S o2n And then, the state change of the load is shown within t-t + TR seconds, namely, an unknown load state switching event is detected. Wherein the load state switch event start time t on Is t seconds, the event end time is t off T + TR, which indicates the duration of the event.
The above process of detecting a load state transition event can be represented by the following formula:
ΔS t |≥S on1 &&|ΔS t+1 |≥S on1 &&...&&|ΔS t+TR-1 |≥S on1
&&|ΔS t+TR |<S on1 &&|ΔS t+TR+1 |<S on1 &&|S t+TR -S t |≥S on2 。
a3, extracting steady-state voltage and current waveforms of T periods before and after a load state switching event, and acquiring the voltage and current values of a single load based on a spectrum analysis method;
specifically, a steady-state voltage current waveform v of T periods before and after a load state transition event is extracted on ,v off ,i on ,i off Calculating a phase angle of a base voltage by using a fast Fourier transform equal-frequency spectrum analysis method, and then taking a sampling point with a zero phase angle as an initial sampling point to ensure a current waveform i off And i on The subtraction can be done directly in the time domain. Of a single loadVoltage v = (v) off +v on ) 2 and current i = i off -i on 。
V on Refers to the total voltage, V, after a load state switching event off Refers to the total voltage before the load state switching event, I on Refers to the total current, I, after a load state switching event off Refers to the total current before the load state switching event.
And A4, carrying out color coding on the voltage and current values of the single load by utilizing a Fryze power strategy to obtain a colorful load V-I track image of the load. And dividing the V-I track image into a training V-I track image and a checking V-I track image, and training the condition generation countermeasure network until the correct recognition rate of the condition generation countermeasure network is 95%.
The correct recognition rate of the conditional generation countermeasure network is 95%, a threshold tau for recognizing abnormal loads is set, and the threshold tau is a distribution threshold of Gaussian distribution of the countermeasure network reconstruction V-I track image generated by the same loads through the conditional generation countermeasure network.
In this embodiment, the training condition generating confrontation network correct recognition rate is 95% according to the actual requirement of this embodiment, and in other embodiments, the training condition generating confrontation network correct recognition rate is determined according to the actual requirement, which is not limited herein.
Then, the power circuit is monitored by using a training conditional countermeasure generation network, and whether an abnormal load occurs is judged.
The method specifically comprises the following steps:
b1, acquiring real-time monitoring data of the power circuit, and denoising, wherein the real-time monitoring data are total voltage and total current of the power circuit;
b2, regarding the de-noised monitoring data, taking the monitoring data with load state conversion as effective monitoring data; in this embodiment, the flow of the monitoring data for determining the load state transition is the same as that in the above A2, and the event detection windows with the same step size are used.
B3, extracting steady-state voltage and current waveforms of T periods before and after the load state switching event, and acquiring the voltage and current values of a single load based on a spectrum analysis method(ii) a The procedure of the step is the same as the procedure A3, a base voltage phase angle is calculated by utilizing a fast Fourier transform equal frequency spectrum analysis method, and then a sampling point with the phase angle being zero is taken as an initial sampling point to ensure that the current waveform i off And i on The subtraction can be done directly in the time domain.
And B4, carrying out color coding on the voltage and current values of the single load by utilizing a Fryze power strategy to obtain a load V-I track image of the load color.
And B5, inputting the V-I track image into a training condition generation countermeasure network, and judging whether each power circuit has an abnormal load or not based on the generated feature reconstruction image.
Inputting the V-I track to be tested into a training condition generation countermeasure network, and if the capsule characteristics of the V-I track of the load, namely the minimum distance between the characteristic reconstruction image and the Gaussian distribution of each known load is greater than or equal to a threshold value, judging that abnormal load occurs at the moment. And if the minimum distance between the load type label and the Gaussian distribution of the known load is smaller than the threshold value, the load type label is the load type label corresponding to the minimum distance at the moment.
Furthermore, the present invention provides an electronic device, which includes a memory and a processor, where the memory stores a computer program, and the processor executes the computer program stored in the memory to implement the steps of the method for monitoring non-invasive abnormal load behavior according to any of the above embodiments.
In practical application, a man-machine interaction device can be arranged, so that a user can check the current monitoring result in real time.
The present invention also provides a computer readable storage medium, having a computer program stored thereon, which, when being executed by a processor, implements the steps of the method for non-intrusive abnormal load behavior monitoring according to any one of the above embodiments.
The invention provides a non-invasive abnormal load behavior monitoring method, electronic equipment and a storage medium, wherein a countermeasure network is generated through training conditions, so that abnormal load behavior monitoring and known load correct identification are realized.
The generator of the conditional generation countermeasure network provided by the embodiment of the invention comprises an auto-encoder and a capsule network, the capsule characteristics of the same load V-I track passing through the encoder and the capsule network correspond to a predefined Gaussian distribution, each load has an independent Gaussian distribution, and the conditional auto-encoder approximates the Gaussian prior of the load to the posterior probability. An additional encoding network maps the generator-generated V-I trajectory to hidden layer representations to minimize the distance between the hidden layer representations of the V-I trajectory in the generator. The method has the advantages that the multiple loss functions are constructed to minimize losses, the adjustment and the updating of the parameters of the condition generation countermeasure network are achieved, the compactness of the same load characteristic near a Gaussian distribution center is achieved, the condition generation countermeasure network learns how to encode the V-I track characteristic of the load and learns the data distribution of the known load, the real-time monitoring of unknown loads or abnormal loads in power users can be achieved, and the method is high in expandability, strong in flexibility and low in error.
The non-invasive abnormal load behavior monitoring method, the electronic device and the storage medium provided by the embodiments of the invention have high monitoring accuracy of the abnormal load behavior, and can be applied to power circuit systems such as families, and the like, so that the real-time power consumption information of each electric appliance in a user can be obtained, including the working state, the power consumption and the accumulated electric quantity of the electric appliance, and even fault information and the like. The method is beneficial to formulation of an energy efficiency policy, and avoids the occurrence of an event that the electric appliance fails to work normally and even direct economic loss is brought to power consumers due to aging of a circuit hardware part in the load. The method provided by the invention has the advantages of low installation cost, flexible and convenient application, high expansibility and good application prospect.
In the description of the present invention, it is to be understood that the terms "first", "second" and the like are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implying any number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.
In the description herein, the description of the terms "one embodiment," "some embodiments," "an embodiment," "an example," "a specific example" or "some examples" or the like, means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Although embodiments of the present invention have been shown and described above, it should be understood that the above embodiments are illustrative and not restrictive, and that those skilled in the art may make changes, modifications, substitutions and alterations to the above embodiments without departing from the scope of the present invention.
Claims (10)
1. A non-invasive abnormal load behavior monitoring method is characterized by comprising the following steps:
s1, acquiring real-time monitoring data of non-invasive load monitoring equipment and denoising; the monitoring data includes: total voltage and total current data of a preselected power circuit monitored by the non-intrusive load monitoring device;
s2, regarding the de-noised monitoring data, taking the monitoring data with load state conversion as effective monitoring data;
s3, carrying out color coding processing on effective monitoring data based on a pre-constructed power strategy to obtain a V-I track image of each single load in the power circuit;
s4, inputting the V-I track image into a training condition to generate a countermeasure network, and judging whether each power circuit has abnormal load or not based on the generated feature reconstruction image;
wherein the conditional generation countermeasure network comprises: the system comprises a conditional self-encoder, a capsule network and a classifier, wherein the conditional self-encoder is used for converting Gaussian prior probability of loads in a V-I track image into Gaussian posterior probability, and the capsule network is used for realizing compactness of the same type of characteristics near a Gaussian distribution center so that the classifier detects the loads.
2. The monitoring method of claim 1, wherein:
before S1, the method further comprises: s0, training the conditional generation countermeasure network:
the S0 comprises:
s01, acquiring training monitoring data samples and checking monitoring data samples for training the condition generation countermeasure network; the training monitoring data sample and the verification monitoring data sample are historical monitoring total voltage and total current data of the same power circuit;
s02, coding the training monitoring data samples and the verification detection data samples based on a pre-constructed power strategy to obtain a training V-I track image and a verification V-I track image of each single load in the power circuit;
s03, inputting the training V-I track image into the conditional countermeasure generation network to reconstruct and generate a feature reconstruction image of the training V-I track aiming at each single load;
s04, inputting the verification V-I track image and the feature reconstruction image corresponding to each single load into a pre-constructed discriminator respectively, and judging whether the feature reconstruction image is matched with the verification V-I track image or not;
s05, adjusting training parameters of the conditional countermeasure generation network, and alternately generating a feature reconstruction image and inputting the feature reconstruction image into a discrimination network so as to enable the feature reconstruction image generated finally by the conditional countermeasure monitoring network to be matched with the verification V-I track image and obtain the trained conditional countermeasure monitoring network.
3. The monitoring method of claim 2, wherein S01 comprises:
acquiring historical monitoring data of at least one load state conversion event in a power circuit monitored by non-invasive load monitoring equipment; the load state transition event is a circuit load transition process caused when a single load in the preselected power circuit is turned on and/or turned off;
calculating a time period of occurrence of a load state transition event based on a predefined event detection window; the method comprises the following specific steps:
calculating the total real power S of the pre-selected power circuit t Determining Δ S t >S on1 Time t;
based on a pre-constructed event detection window, calculating and determining the total real power change quantity deltaS when t = t + TR t+TR <S on1 (ii) a R is the step length of the event detection window, delta S t =S t+1 -S t ;
If S t+TR -S t <S on2 Judging that a load state transition event occurs in the time period of t-t + TR;
acquiring total voltage and total current data of T time period periods before and after the load state transition event occurs, and acquiring a training monitoring data sample and a verification monitoring data sample;
said S on1 For a predefined load state transition event initiation threshold, S on2 Is a predefined load state transition event end threshold.
4. The monitoring method of claim 1, wherein S3 comprises:
s30, sampling the effective monitoring data based on a pre-constructed spectrum analysis method to obtain the voltage and current values of each single load in the power circuit;
s31, determining the active component current i of the current i (t) of each single load based on a pre-constructed Fryze power strategy a (t) and the reactive component current i f (t);
Based on the active power distributionMagnitude current i a (t) and the reactive component current i f (t) calculating and obtaining a power factor matrixThe power factor is the ratio of the power of the active component current to the power of the reactive component current;
k is the total number of sampling points, P apparent To be real power, V rms 、I rms Effective values of load voltage and current respectively;
s32, constructing a hue matrix of the V-I track based on a pre-constructed HSV color space aiming at each single loadAnd a voltage period matrix V;
5. The monitoring method of claim 4,
the S32 specifically includes;
s321, based on the HSV color space, obtaining the V by using hue attribute hue-direction of motion H of the I trajectory j ;
Based on the direction of movement H j Storing the hue of the jth sampling point into a2 Nx 2N matrix to obtain a hue matrix
The direction of motion H j The calculation expression is:
the arg is an arc tangent function of four quadrants;
s322, based on the pre-constructed binary image W m (1,2,. Once, M), averaging M cycles of a single load voltage to obtain a voltage cycle matrix V;
the expression of the voltage period matrix V is:
6. the monitoring method according to claim 2, characterized in that the condition is adjusted against a training parameter of the monitoring network, in particular,
calculating the minimum value of the training loss of the conditional countermeasure monitoring network based on a loss function constructed in advance;
adjusting the minimum value weighting calculation to generate parameters of a countermeasure network;
the loss functions include feature matching loss functions, reconstruction loss functions, additional encoder loss functions, center constrained loss functions, and/or contrast loss functions.
7. The monitoring method of claim 6,
the feature matching loss function expression is as follows:
f (x) is given input V-I track x, the output of the discriminator intermediate layer;
the reconstruction loss function expression is as follows:
the above-mentionedMu is the average intensity of the training V-I track image, delta is the standard deviation of the training V-I track image,reconstructing the covariance of the image for the training V-I trajectory and features; c1 and c2 are constants;
the additional encoder loss function expression is:
the z is a sampling vector characteristic output by the capsule network when the V-I track image is trained,reconstructing the coding features of the image for the features;
the central constraint loss function expression is as follows:
L KL =d(C,sg[P y ]);
the above-mentionedC is a probability capsule, and P is Gaussian distribution of the target load cluster;
the contrast loss function expression is:
said [ ·] + Is a function of the positive number of the return parameter;
the formula for calculating the minimum value weight is as follows:
L=αL KL +βL rec +γL contr +σL enc +λL adv and the alpha, the beta, the gamma, the sigma and the lambda are constants.
8. The monitoring method of claim 1,
the S4 specifically includes:
inputting the real-time V-I track of the power circuit into the trained condition generation countermeasure network to generate a real-time feature reconstruction image;
calculating the minimum distance between the real-time feature reconstruction image and a final training condition to generate a historical feature reconstruction image of the countermeasure network;
and if the minimum distance between the real-time characteristic reconstruction image and the historical characteristic reconstruction image is larger than the threshold tau meeting the preset requirement, judging that an abnormal load occurs in the pre-selection power circuit.
9. An electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program stored in the memory to implement the steps of the method for non-invasive abnormal load behavior monitoring according to any one of the claims 1 to 8.
10. A computer-readable storage medium, having stored thereon a computer program which, when being executed by a processor, carries out the steps of the method for non-intrusive abnormal load behavior monitoring method as defined in any one of the preceding claims 1 to 8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211066557.6A CN115423128A (en) | 2022-08-31 | 2022-08-31 | Non-invasive abnormal load behavior monitoring method, electronic equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211066557.6A CN115423128A (en) | 2022-08-31 | 2022-08-31 | Non-invasive abnormal load behavior monitoring method, electronic equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115423128A true CN115423128A (en) | 2022-12-02 |
Family
ID=84202049
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211066557.6A Pending CN115423128A (en) | 2022-08-31 | 2022-08-31 | Non-invasive abnormal load behavior monitoring method, electronic equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115423128A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116774135A (en) * | 2023-08-22 | 2023-09-19 | 山东国研自动化有限公司 | Remote meter reading abnormity monitoring method and system |
-
2022
- 2022-08-31 CN CN202211066557.6A patent/CN115423128A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116774135A (en) * | 2023-08-22 | 2023-09-19 | 山东国研自动化有限公司 | Remote meter reading abnormity monitoring method and system |
CN116774135B (en) * | 2023-08-22 | 2023-11-17 | 山东国研自动化有限公司 | Remote meter reading abnormity monitoring method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Liu et al. | Energy disaggregation of appliances consumptions using ham approach | |
CN109387712B (en) | Non-invasive load detection and decomposition method based on state matrix decision tree | |
CN108021736A (en) | A kind of load switching action monitoring method based on sliding window Remanent Model | |
CN109598451B (en) | Non-invasive load identification method based on PCA (principal component analysis) and LSTM (least Square TM) neural network | |
CN107546855B (en) | Non-invasive decomposition method for residential electricity load | |
CN111864896B (en) | Power load monitoring method and system | |
CN112434799B (en) | Non-invasive load identification method based on full convolution neural network | |
US20120059607A1 (en) | Method and device for filtering electrical consumption curves and allocating consumption to classes of appliances | |
CN111027408A (en) | Load identification method based on support vector machine and V-I curve characteristics | |
CN105548949A (en) | Method and system for fault remote determination of intelligent ammeter | |
CN110119545B (en) | Non-invasive power load identification method based on stack type self-encoder | |
CN110288039A (en) | Based on user power utilization load characteristic stealing detection method | |
CN115423128A (en) | Non-invasive abnormal load behavior monitoring method, electronic equipment and storage medium | |
CN108918932A (en) | Power signal adaptive filter method in load decomposition | |
CN112327070A (en) | Non-invasive load real-time monitoring method and storage medium | |
CN113807225A (en) | Load identification method based on feature fusion | |
CN111722028A (en) | Load identification method based on high-frequency data | |
Liebgott et al. | Active learning with cross-dataset validation in event-based non-intrusive load monitoring | |
CN114236234A (en) | Electrical appliance characteristic identification method based on fundamental wave and harmonic wave mixed criterion | |
Li et al. | Non-intrusive load monitoring based on convolutional neural network mixed residual unit | |
Luo et al. | A generalizable method for practical non-intrusive load monitoring via metric-based meta-learning | |
CN109377052B (en) | Power load matching method, device, equipment and readable storage medium | |
Li et al. | Incipient Fault Detection in Power Distribution System: A Time–Frequency Embedded Deep-Learning-Based Approach | |
Kee et al. | Non-intrusive load monitoring (nilm)–a recent review with cloud computing | |
CN112329809A (en) | High-voltage circuit breaker fault diagnosis method based on decision tree algorithm |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |