CN113344346B - Power utilization abnormity detection method and system based on non-intrusive load decomposition - Google Patents
Power utilization abnormity detection method and system based on non-intrusive load decomposition Download PDFInfo
- Publication number
- CN113344346B CN113344346B CN202110570582.7A CN202110570582A CN113344346B CN 113344346 B CN113344346 B CN 113344346B CN 202110570582 A CN202110570582 A CN 202110570582A CN 113344346 B CN113344346 B CN 113344346B
- Authority
- CN
- China
- Prior art keywords
- user
- power consumption
- utilization
- load
- density
- 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.)
- Active
Links
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 72
- 238000001514 detection method Methods 0.000 title claims abstract description 51
- 230000006399 behavior Effects 0.000 claims abstract description 139
- 230000005611 electricity Effects 0.000 claims abstract description 88
- 230000002159 abnormal effect Effects 0.000 claims abstract description 75
- 238000005516 engineering process Methods 0.000 claims abstract description 21
- 238000005065 mining Methods 0.000 claims abstract description 10
- 238000000034 method Methods 0.000 claims description 36
- 238000012549 training Methods 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 8
- 230000001174 ascending effect Effects 0.000 claims description 3
- 230000008901 benefit Effects 0.000 abstract description 7
- 230000008569 process Effects 0.000 description 8
- 230000009471 action Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000004378 air conditioning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000010438 heat treatment Methods 0.000 description 2
- 239000002356 single layer Substances 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000035622 drinking Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000005057 refrigeration Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- 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)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a power utilization abnormity detection method and system based on non-invasive load decomposition, and relates to the technical field of abnormal power utilization behavior detection of users. The technical scheme is that firstly, the total power consumption of a user is decomposed based on a non-invasive load decomposition technology to obtain the power consumption of each device; mining user behavior characteristics including power utilization time characteristics and load characteristics based on the power utilization power of each device; and finally, detecting abnormal electricity utilization behaviors by using a density peak value clustering algorithm based on the user behavior characteristics. According to the technical scheme, the user behavior characteristics are extracted by using a non-invasive load decomposition technology, clustering is realized based on the user behavior characteristics, so that the detection result of the abnormal electricity utilization behavior of the user is obtained, the accuracy rate of the abnormal electricity utilization detection is improved, and economic benefits are brought to electricity selling companies and energy service providers.
Description
Technical Field
The invention relates to the technical field of user power utilization abnormity detection, in particular to a power utilization abnormity detection method and system based on non-invasive load decomposition.
Background
Abnormal electricity utilization behaviors of users, such as electricity stealing and the like, can cause non-technical loss of power transmission and distribution in the operation of a power grid, so that the benefits of electricity selling companies, energy service providers and the like are damaged. Along with the development of the smart power grid, the popularity of the smart power meter is continuously increased, meanwhile, the electricity stealing means of a user is changed from destroying the traditional power meter or a private power line and the like into attacking the smart power meter through a digital storage technology and a network communication technology, and the electricity consumption at the corresponding moment is reduced or directly zeroed through data tampering. The tradition relies on technical staff to carry out artifical screening or relies on camera, unmanned aerial vehicle control in order to prevent that more manpower and material resources are wasted to the method that the electricity of stealing takes place. Under an advanced measurement system, the load data of the user is collected and analyzed in real time, a new effective way is provided for power utilization abnormity detection, the cost of power utilization abnormity detection can be greatly reduced, and the abnormity detection efficiency and accuracy are improved.
At present, data driving methods for the field of user abnormal electricity utilization behavior detection mainly include methods based on classification, regression and clustering, and several methods have respective advantages and disadvantages, but the methods are more commonly used and are also based on clustering. The clustering-based method is used for extracting the power utilization characteristics of the users and finding out a few abnormal users which do not accord with the power utilization behavior characteristics of most users, so that the abnormal detection is realized.
However, the existing clustering-based abnormal electricity consumption behavior detection excessively focuses on the improvement of the clustering algorithm, the extraction research on the electricity consumption characteristics of the user is less, the adopted characteristics are mostly the electrical characteristics of a time domain, a frequency domain and the like, and the user behavior characteristics which have important effects on the abnormal electricity consumption behavior detection are often ignored or less focused, so that the abnormal electricity consumption behavior detection result of the user is inaccurate.
Disclosure of Invention
Technical problem to be solved
Aiming at the defects of the prior art, the invention provides a power utilization abnormity detection method and system based on non-invasive load decomposition, and solves the problem that the power utilization abnormity detection result is inaccurate because the user behavior characteristics are ignored or less concerned in the prior art.
(II) technical scheme
In order to achieve the purpose, the invention is realized by the following technical scheme:
in a first aspect, the present invention first provides a power utilization anomaly detection method based on non-intrusive load decomposition, where the method includes:
decomposing the total power consumption of a user based on a non-invasive load decomposition technology to obtain the power consumption of each device;
acquiring user behavior characteristics based on the power consumption of each device;
and detecting abnormal electricity utilization behaviors by using a density peak value clustering algorithm based on the user behavior characteristics.
Preferably, the decomposing the total power consumption of the user based on the non-intrusive load decomposition technology to obtain the power consumption of each device includes:
s11, establishing a mapping relation between the decomposed total power consumption and working modes of all electric equipment based on a non-invasive load decomposition model, and obtaining codes of state codes corresponding to the total power consumption of all electric equipment based on the combination of the working modes of all electric equipment;
s12, training a seq2seq model based on RNN based on the coding of the state code;
and S13, carrying out load decomposition by using the trained seq2seq model based on the RNN to obtain the power consumption of each device.
Preferably, the user behavior characteristics include power consumption time characteristics and load characteristics, wherein the power consumption time characteristics include common operation periods, daily operation frequencies, common off periods and daily off frequencies of different devices; the load characteristics comprise the average daily power consumption of equipment of different users, the maximum daily power consumption of the equipment, the maximum load hours of the equipment and the average daily consumption of the equipment.
Preferably, the detecting abnormal electricity consumption behavior by using a density peak clustering algorithm based on the user behavior characteristics includes:
s31, primarily classifying the users by adopting a density clustering algorithm based on the electricity utilization time characteristics to obtain different classes of users;
and S32, aiming at the different classes of users, detecting abnormal electricity utilization behaviors by adopting a KNN-based fast density peak value clustering algorithm according to load characteristics.
Preferably, the determination condition for detecting the abnormal electricity consumption behavior is: when the local density of a certain pointDistance between two adjacent platesIf so, determining that the user is a user with abnormal electricity utilization behavior;
where b represents the device, and b ∈ {1,2, 3., M }, a represents the sample in the dataset, and a ∈ {1,2, 3., N }, γ ∈ {1,2, 3., N } ρ And gamma δ Are empirical parameters.
In a second aspect, the present invention further provides a power consumption anomaly detection system based on non-intrusive load decomposition, where the system includes:
the load decomposition module is used for decomposing the total power consumption of the user based on a non-invasive load decomposition technology to obtain the power consumption of each device;
the user behavior feature mining module is used for acquiring user behavior features based on the power consumption of each device;
and the abnormal electricity utilization behavior detection module is used for realizing the abnormal electricity utilization behavior detection by utilizing a density peak value clustering algorithm based on the user behavior characteristics.
Preferably, the load decomposition module decomposing the total power consumption of the user based on a non-intrusive load decomposition technique to obtain the power consumption of each device includes:
s11, establishing a mapping relation between the decomposed total power consumption and working modes of all electric equipment based on a non-invasive load decomposition model, and obtaining codes of state codes corresponding to the total power consumption of all electric equipment based on the combination of the working modes of all electric equipment;
s12, training a seq2seq model based on RNN based on the coding of the state code;
and S13, performing load decomposition by using the trained seq2seq model based on the RNN to obtain the power consumption of each device.
Preferably, the user behavior characteristics obtained by the user behavior characteristic mining module include power utilization time characteristics and load characteristics, wherein the power utilization time characteristics include common operation periods, daily operation frequencies, common shutdown periods and daily shutdown frequencies of different devices; the load characteristics comprise the average daily power consumption of equipment of different users, the maximum daily power consumption of the equipment, the maximum load hours of the equipment and the average daily consumption of the equipment.
Preferably, the detecting module of abnormal electricity consumption behavior based on the user behavior characteristics and using a density peak clustering algorithm to detect abnormal electricity consumption behavior includes:
s31, primarily classifying the users by adopting a density clustering algorithm based on the electricity utilization time characteristics to obtain different classes of users;
and S32, aiming at the different classes of users, detecting abnormal electricity utilization behaviors by adopting a KNN-based density peak value clustering algorithm according to load characteristics.
Preferably, the condition that the abnormal electricity consumption behavior detection module judges the abnormal electricity consumption behavior is as follows: when the local density of a certain pointDistance between two adjacent devicesIf so, determining that the user is a user with abnormal electricity utilization behavior;
where b represents the device, and b ∈ {1,2, 3., M }, a represents the sample in the dataset, and a ∈ {1,2, 3., N }, γ ∈ {1,2, 3., N } ρ And gamma δ Are empirical parameters.
(III) advantageous effects
The invention provides a power utilization abnormity detection method and system based on non-invasive load decomposition. Compared with the prior art, the method has the following beneficial effects:
1. according to the technical scheme, total power consumption is decomposed on the basis of a non-invasive load decomposition technology to obtain power consumption of each device, user behavior characteristics are mined according to the power consumption of each device, and clustering is performed twice in sequence by using a density peak clustering algorithm and a KNN-based rapid density peak clustering algorithm on the basis of the user behavior characteristics, so that abnormal power consumption behaviors are detected. According to the technical scheme, the user behavior characteristics are extracted by using a non-intrusive load decomposition technology, clustering is realized based on the user behavior characteristics, so that the detection result of the abnormal electricity utilization behavior of the user is obtained, the accuracy rate of the detection of the abnormal electricity utilization behavior is improved, and economic benefits are brought to electricity selling companies and energy service providers;
2. according to the technical scheme, the total power utilization of the electric equipment is decomposed by using a non-invasive load decomposition technology to obtain the power utilization of each equipment, and then the user behavior characteristics including the power utilization time characteristics and the load characteristics are extracted from the power utilization of each equipment, so that the user behavior characteristics which can more accurately represent the power utilization behavior of the user can be obtained;
3. according to the technical scheme, the density peak clustering algorithm and the KNN-based rapid density peak clustering algorithm are utilized to cluster the power utilization time characteristic and the load characteristic of the user behavior characteristic for two times, the power utilization behavior recognition results of the users on different equipment sets are weighted when the final clustering result is obtained, and finally the abnormal power utilization behavior of the users is detected.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the prior art descriptions will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a flow chart of a method for detecting power consumption anomalies based on non-intrusive load decomposition in accordance with an embodiment of the present invention;
FIG. 2 is a block diagram of a Seq2Seq network architecture according to an embodiment of the present invention;
fig. 3 is a flowchart illustrating an embodiment of detecting a power consumption abnormality of a user.
Detailed Description
To make the objects, technical solutions and advantages of the embodiments of the present invention clearer and more complete description of the technical solutions in the embodiments of the present invention, it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
By providing the power consumption abnormity detection method and system based on non-invasive load decomposition, the problem that in the prior art, the detection result of the abnormal power consumption behavior of the user is inaccurate due to neglect or less attention to the behavior characteristics of the user is solved, and the purpose of accurately monitoring the abnormal power consumption behavior of the user is achieved.
In order to solve the technical problems, the general idea of the embodiment of the application is as follows:
in order to obtain an accurate user abnormal electricity consumption behavior detection result, the technical scheme firstly decomposes the total electricity consumption power of the electric equipment based on a non-invasive load decomposition technology to obtain the electricity consumption power of each equipment, then digs user behavior characteristics including electricity consumption time characteristics and load characteristics according to the electricity consumption power of each equipment, finally carries out twice clustering by utilizing a density peak value clustering algorithm and a KNN-based rapid density peak value clustering algorithm based on the user behavior characteristics, and weights the clustering result, thereby realizing the accurate detection of the abnormal electricity consumption behavior.
In order to better understand the technical scheme, the technical scheme is described in detail in the following with reference to the attached drawings of the specification and specific embodiments.
Example 1:
in a first aspect, referring to fig. 1, the present invention first proposes a power utilization anomaly detection method based on non-invasive load decomposition, the method including:
s1, decomposing the total power consumption of a user based on a non-invasive load decomposition technology to obtain the power consumption of each device;
s2, mining user behavior characteristics based on the power consumption of each device;
and S3, detecting abnormal electricity utilization behaviors by using a density peak value clustering algorithm based on the user behavior characteristics.
According to the technical scheme, the total power utilization is decomposed on the basis of a non-intrusive load decomposition technology to obtain the power utilization of each device, then the user behavior characteristics are mined according to the power utilization of each device, and finally clustering is performed twice on the basis of the user behavior characteristics by using a density peak clustering algorithm and a KNN-based rapid density peak clustering algorithm, so that the abnormal power utilization behavior is detected. According to the technical scheme, the user behavior characteristics are extracted by using a non-invasive load decomposition technology, clustering is realized based on the user behavior characteristics, so that the detection result of the abnormal electricity utilization behavior of the user is obtained, the accuracy rate of the detection of the abnormal electricity utilization behavior is improved, and economic benefits are brought to electricity selling companies and energy service providers.
In the method according to the embodiment of the present invention, in order to avoid cost consumption caused by methods such as intrusive load decomposition when decomposing a load, in the present technical solution, the total power consumption of the electrical equipment is decomposed by using a non-intrusive load decomposition technique, and at this time, a preferred processing manner is that the decomposing the total power consumption of the user based on the non-intrusive load decomposition technique to obtain the power consumption of each equipment includes:
s11, establishing a mapping relation between the decomposed total power consumption and working modes of all electric equipment based on a non-invasive load decomposition model, and obtaining codes of state codes corresponding to the total power consumption of all electric equipment based on the combination of the working modes of all electric equipment;
s12, training a seq2seq model based on RNN based on the coding of the state code;
and S13, performing load decomposition by using the trained seq2seq model based on the RNN to obtain the power consumption of each device.
In practice, in order to obtain more accurate mining user behavior characteristics, based on consideration of actual user electricity utilization real conditions, a better processing mode is that the user behavior characteristics include electricity utilization time characteristics and load characteristics, wherein the electricity utilization time characteristics include common operation periods, daily operation frequencies, common shutdown periods and daily shutdown frequencies of different devices; the load characteristics comprise daily average power consumption of equipment of different users, daily maximum power consumption of the equipment, maximum load hours of the equipment and daily average usage of the equipment.
In addition, in order to accurately obtain a determination result of the abnormal electricity consumption behavior of the user, a preferred processing manner is that the detecting of the abnormal electricity consumption behavior by using a density peak value clustering algorithm based on the user behavior characteristics includes:
s31, primarily classifying the users by adopting a density clustering algorithm based on the electricity utilization time characteristics to obtain different classes of users;
and S32, aiming at the different classes of users, detecting abnormal electricity utilization behaviors by adopting a KNN-based fast density peak value clustering algorithm according to load characteristics.
Meanwhile, when the abnormal electricity utilization behavior of the user is judged, two quantities of local density and distance are selected for judgment, and specifically, the abnormal electricity utilization behaviorThe judgment conditions for detection are as follows: when the local density of a certain pointDistance between two adjacent devicesIf so, determining that the user is a user with abnormal electricity utilization behavior;
where b represents the device, and b ∈ {1,2, 3., M }, a represents the sample in the dataset, and a ∈ {1,2, 3., N }, γ ∈ {1,2, 3., N } ρ And gamma δ Are empirical parameters.
The following describes the implementation of an embodiment of the present invention in detail with reference to the explanation of specific steps.
S1, decomposing the total power consumption of a user based on a non-invasive load decomposition technology to obtain the power consumption of each device.
The electric appliance in this embodiment is a household electric appliance, and it is known that in practice, the states of electric appliances of household electric appliances are not only simple to turn on and off, but also can be divided into four types, namely, a switch type, a multi-state type, a continuous change type and a permanent operation type, according to the use, internal structure and operating principle of each electric appliance. Wherein the switch type has only two states of on and off, such as lamp and drinking waterMachines, etc.; the multi-state type device has a limited number of operating states, which include not only on and off states, but also other states, for example, an air conditioner is a multi-state type device, which includes on, off, cooling, heating, and other states; the power of the continuous variation type equipment can be continuously changed, and no obvious conversion signal exists when the state is changed, such as a dimming lamp with continuously adjustable brightness; permanently running devices, such as routers, television set-top boxes, etc., which have relatively low power consumption, maintain constant power consumption for long monitoring periods. Each state of these devices corresponds to a power, for example, for an air conditioner, the power of the air conditioner is a when the air conditioner is turned on 1 The power of the air conditioning equipment during refrigeration is a 2 The power of the air conditioning equipment is a when heating 3 The power is 0 when the air conditioner is turned off.
Because the working modes of the household appliance are limited and the appliance has rated power in each working mode, the traditional non-intrusive load decomposition model can be reduced to the mapping relation of the decomposed power value to each appliance working mode, and the formula is as follows:
[Y t 1 ,Y t 2 ,…,Y t M ]=f(X t )
wherein, X t Represents the total power at time t;
Y t i the power value of the ith electric appliance obtained by decomposition at the t moment is represented;
m represents a total of M electrical devices;
the f-function represents a decomposition mapping function of Non-intrusive load monitoring (NILM).
Consider further, the formulaThe number of the combination is finite, each combination case is expressed by a state code to realize dimension reduction, and the formula can be expressed as follows:
y=[S 1 ,S 2 ,…,S M ],S i ∈{0,1,…,n i -1},i∈{1,2,3,...,M}
wherein y represents a status code; s i Indicating the working mode of the ith electric appliance at a certain moment; n is a radical of an alkyl radical i Indicates the number of operation modes the i-th electrical appliance has.
Assuming that M devices are to be decomposed, the total number N of state codes can be expressed as:
wherein N represents the total number of status codes of the M electrical devices.
In conjunction with the above equations, the conventional non-intrusive load detection problem can be transformed into the following equation:
y'=f(X t ),y'∈{0,1,…,N-1}
wherein X t Represents the total power of all appliances; y' represents the code of the state code corresponding to the total power at the current moment; n represents the number of state codes of the electric appliance; f denotes the NILM decomposition function.
Adopting a Seq2Seq model based on RNN, referring to FIG. 2, which is a structural diagram of a Seq2Seq network structure, an encoder (encoder) selects a bidirectional single-layer LSTM, a decoder (decoder) selects a single-layer LSTM, and a hidden state { h) corresponding to each moment of an input time sequence obtained in the encoding process is utilized 1 ,h 2 ,…,h l L represents the length of the input sequence, and an Attention mechanism is introduced to calculate to obtain a dynamic semantic vector C t As part of the input to participate in the operation of the decoding process.
Then, the seq2seq model based on the RNN is trained, and the model training process is judged by taking the model loss function as a judgment standard, so as to obtain a model meeting the expected requirements. Specifically, the model loss function may adopt a cross entropy loss function, and the smaller the value of the cross entropy loss function is, the better the model training effect is represented. When the value of the cross entropy loss function no longer changes, it can be determined that training of the model can be stopped.
Network parameters are adjusted in a back propagation mode, so that the model is improved in training. In order to prevent the risk of over-fitting the model, methods such as Dropout and L2 regularization are introduced in the model training process.
And finally, performing load decomposition by using the trained seq2seq model based on the RNN to obtain power information of each electric device.
And S2, mining user behavior characteristics based on the power consumption of each device.
And mining the user behavior characteristics based on the result of the non-invasive load decomposition, namely the power value of each electric device. Wherein the user behavior characteristics include four power consumption time characteristics and four load characteristics, for a total of eight specific characteristics, specifically,
acquiring electricity consumption time characteristics: and obtaining the electricity utilization time characteristic by considering the behavior habits of users based on the power value of each device obtained by non-invasive load decomposition. The electricity consumption time characteristics specifically include common operation periods, daily operation frequencies, common off periods, and daily off frequencies of different devices.
Acquiring load characteristics: and obtaining load characteristics by combining the power consumption of different electrical equipment and the power consumption based on the power value of each equipment obtained by non-intrusive load decomposition. The load characteristics specifically include daily average power consumption of the devices of different users, daily maximum power consumption of the devices, maximum load hours of the devices, and daily average usage of the devices.
And S3, detecting abnormal electricity utilization behaviors by using a density peak value clustering algorithm based on the user behavior characteristics.
Referring to fig. 3, firstly, according to four characteristics of the electricity consumption time characteristics in the user behavior characteristics, a density clustering algorithm is adopted to preliminarily classify the users to obtain users of different categories, and then a KNN-based fast density peak clustering algorithm is adopted to realize abnormal electricity consumption behavior detection according to the four characteristics of the load characteristics in the user behavior characteristics for the users of different categories. The specific flow of the algorithm is as follows:
calculating any sample x in different equipment data sets of various users a With other samples x b Euclidean distance d (x) therebetween a ,x b ) And arranging the calculation results in ascending order, and recording the sample corresponding to the kth distance as N k (x a ) Then x a The k nearest neighbors of (c) are:
N(x a )={x b ∈X|d(x a ,x b )≤d(x a ,N k (x a ))}
Wherein K is defined by the parameter p, K = p N P is the percentage of the number of samples N, the local densityThe larger the value of (A), the larger is x a The greater the density of (a).
The distance between the KNN-based point and the high-density point is given on the basis of the formulaIs defined as follows:
calculating the local density of each sample and the distance between the local density and the high-density point according to a formula
Based on samples obtained by the preceding stepAnd determining the users with abnormal electricity utilization behaviors.
The technical scheme is that clustering is carried out on different electric equipment sets, so that the performance of users on different electric equipment sets needs to be comprehensively considered in a weighting mode when abnormal values are detected, clustering can better accord with the actual condition, clustering results are more accurate, namely, for the electric equipment sets obtained through decomposition, the clustering results are respectively calculated on different electric equipment setsAnd then, carrying out average weighting on the values on the electric equipment sets, and judging the outlier by using the weighted values. Due to calculation ofThe value of (2) is corresponding to each user using the electric equipment, and finally, the user with abnormal electricity utilization behavior can be found according to the outlier. Specifically, the judgment conditions of the outliers are as follows:
when the local density of a certain pointDistance between two adjacent devicesThen the point is an outlier;
where b represents the device, and b ∈ {1,2, 3., M }, a and represents the sample in the dataset, and a ∈ {1,2, 3., N }, γ ∈ {1,2, 3., N } ρ And gamma δ Are empirical parameters.
And finally, judging the user corresponding to the obtained outlier as a user with abnormal electricity utilization behavior.
Therefore, the whole process of the power utilization abnormity detection method based on non-invasive load decomposition is completed.
Example 2:
in a second aspect, the present invention further provides a power consumption anomaly detection system based on non-intrusive load decomposition, including:
the load decomposition module is used for decomposing the total power consumption of the user based on a non-invasive load decomposition technology to obtain the power consumption of each device;
the user behavior feature mining module is used for acquiring user behavior features based on the power consumption of each device;
and the abnormal electricity utilization behavior detection module is used for realizing the abnormal electricity utilization behavior detection by utilizing a density peak value clustering algorithm based on the user behavior characteristics.
Preferably, the load decomposition module decomposing the total power consumption of the user based on a non-intrusive load decomposition technique to obtain the power consumption of each device includes:
s11, establishing a mapping relation between the decomposed total power consumption and working modes of all electric equipment based on a non-invasive load decomposition model, and obtaining codes of state codes corresponding to the total power consumption of all electric equipment based on the combination of the working modes of all electric equipment;
s12, training a seq2seq model based on RNN based on the coding of the state code;
and S13, performing load decomposition by using the trained seq2seq model based on the RNN to obtain the power consumption of each device.
Preferably, the user behavior characteristics obtained by the user behavior characteristic mining module include power utilization time characteristics and load characteristics, wherein the power utilization time characteristics include common operation periods, daily operation frequencies, common shutdown periods and daily shutdown frequencies of different devices; the load characteristics comprise daily average power consumption of equipment of different users, daily maximum power consumption of the equipment, maximum load hours of the equipment and daily average usage of the equipment.
Preferably, the detecting module of abnormal electricity consumption behavior based on the user behavior characteristics and using a density peak clustering algorithm to detect abnormal electricity consumption behavior includes:
s31, primarily classifying the users by adopting a density clustering algorithm based on the electricity utilization time characteristics to obtain different classes of users;
and S32, aiming at the different classes of users, detecting abnormal power utilization behaviors by adopting a KNN-based density peak clustering algorithm according to load characteristics.
Preferably, the condition that the abnormal electricity consumption behavior detection module judges the abnormal electricity consumption behavior is as follows: when the local density of a certain pointDistance between two adjacent platesIf so, determining that the user is a user with abnormal electricity utilization behavior;
where b represents the device, and b ∈ {1,2, 3., M }, a represents the sample in the dataset, and a ∈ {1,2, 3., N }, γ ∈ {1,2, 3., N } ρ And γ δ are empirical parameters.
It can be understood that the power consumption anomaly detection system based on non-invasive load decomposition provided in the embodiment of the present invention corresponds to the above power consumption anomaly detection method based on non-invasive load decomposition, and the explanations, examples, and beneficial effects of the relevant contents thereof may refer to the corresponding contents in the power consumption anomaly detection method based on non-invasive load decomposition, and are not repeated herein.
In summary, compared with the prior art, the method has the following beneficial effects:
1. according to the technical scheme, total power consumption is decomposed on the basis of a non-intrusive load decomposition technology to obtain power consumption of each device, user behavior characteristics are mined according to the power consumption of each device, and clustering is performed twice in sequence by using a density peak clustering algorithm and a KNN-based rapid density peak clustering algorithm on the basis of the user behavior characteristics, so that abnormal power consumption behavior detection is achieved. According to the technical scheme, the user behavior characteristics are extracted by using a non-intrusive load decomposition technology, clustering is realized based on the user behavior characteristics, so that the detection result of the abnormal electricity utilization behavior of the user is obtained, the accuracy rate of the detection of the abnormal electricity utilization behavior is improved, and economic benefits are brought to electricity selling companies and energy service providers;
2. according to the technical scheme, the total power utilization of the electric equipment is decomposed by using a non-invasive load decomposition technology to obtain the power utilization of each equipment, and then user behavior characteristics including power utilization time characteristics and load characteristics are extracted from the power utilization of each equipment, so that the user behavior characteristics which can more accurately represent the power utilization behavior of the user can be obtained;
3. according to the technical scheme, the density peak clustering algorithm and the KNN-based rapid density peak clustering algorithm are utilized to cluster the power utilization time characteristic and the load characteristic of the user behavior characteristic for two times, the power utilization behavior recognition results of the users on different equipment sets are weighted when the final clustering result is obtained, and finally the abnormal power utilization behavior of the users is detected.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrases "comprising a," "8230," "8230," or "comprising" does not exclude the presence of additional like elements in a process, method, article, or apparatus that comprises the element.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (4)
1. A method for detecting power consumption anomaly based on non-intrusive load decomposition, the method comprising:
decomposing the total power consumption of a user based on a non-invasive load decomposition technology to obtain the power consumption of each device;
acquiring user behavior characteristics based on the power consumption of each device;
detecting abnormal electricity utilization behaviors by using a density peak value clustering algorithm based on the user behavior characteristics;
the user behavior characteristics comprise power utilization time characteristics and load characteristics, wherein the power utilization time characteristics comprise common operation periods, daily operation frequencies, common closing periods and daily closing frequencies of different devices; the load characteristics comprise daily average power consumption of equipment of different users, daily maximum power consumption of the equipment, maximum load hours of the equipment and daily average consumption of the equipment;
the method for realizing abnormal electricity utilization behavior detection by using a density peak value clustering algorithm based on the user behavior characteristics comprises the following steps:
s31, primarily classifying the users by adopting a density clustering algorithm based on the electricity utilization time characteristics to obtain different classes of users;
s32, aiming at the different classes of users, detecting abnormal electricity utilization behaviors by adopting a KNN-based fast density peak value clustering algorithm according to load characteristics;
the S31 and S32 include: calculating any sample x in different equipment data sets of various users a With other samples x b Euclidean distance d (x) between a ,x b ) And arranging the calculation results in ascending order, and recording the sample corresponding to the kth distance as N k (x a ) Then x a The k nearest neighbors of (c) are:
N(x a )={x b ∈X|d(x a ,x b )≤d(x a ,N k (x a ))}
Where K is determined by the parameter p, K = pN, p being the percentage of the number of samples N;
the distance between the KNN-based and the high-density point is given on the basis of the formulaIs defined as follows:
calculating the local density of each sample and the distance between the local density and the high-density point according to a formula
Based on samples obtained by the preceding calculationDetermining a user with abnormal electricity utilization behaviors;
for the electric equipment sets obtained through the decomposition, the calculation is respectively carried out on different electric equipment setsThe values on the power utilization equipment sets are weighted averagely, then outliers are judged by using the weighted values, and users with abnormal power utilization behaviors are detected according to the outliers;
the judgment conditions for detecting the abnormal electricity utilization behavior are as follows: when the local density of a certain pointDistance between two adjacent platesIf so, determining that the user is a user with abnormal electricity utilization behavior;
where b represents the device, b is in {1,2, 3.., M }, a represents the sample in the dataset, a is in {1,2, 3.., N }, γ is in the dataset ρ And gamma δ Are empirical parameters.
2. The method of claim 1, wherein the decomposing the total power usage of the user based on the non-intrusive load decomposition technique to obtain the power usage of each device comprises:
s11, establishing a mapping relation between the decomposed total power consumption and working modes of all electric equipment based on a non-invasive load decomposition model, and obtaining codes of state codes corresponding to the total power consumption of all electric equipment based on the combination of the working modes of all electric equipment;
s12, training a seq2seq model based on RNN based on the coding of the state code;
and S13, carrying out load decomposition by using the trained seq2seq model based on the RNN to obtain the power consumption of each device.
3. A system for detecting power usage anomalies based on non-intrusive load splitting, the system comprising:
the load decomposition module is used for decomposing the total power consumption of the user based on a non-invasive load decomposition technology to obtain the power consumption of each device;
the user behavior feature mining module is used for acquiring user behavior features based on the power consumption of each device;
the abnormal electricity utilization behavior detection module is used for realizing abnormal electricity utilization behavior detection by utilizing a density peak value clustering algorithm based on the user behavior characteristics;
the user behavior characteristics comprise power utilization time characteristics and load characteristics, wherein the power utilization time characteristics comprise common operation periods, daily operation frequencies, common closing periods and daily closing frequencies of different devices; the load characteristics comprise daily average power consumption of equipment of different users, daily maximum power consumption of the equipment, maximum load hours of the equipment and daily average consumption of the equipment;
the method for realizing abnormal electricity utilization behavior detection by using a density peak value clustering algorithm based on the user behavior characteristics comprises the following steps:
s31, primarily classifying the users by adopting a density clustering algorithm based on the electricity utilization time characteristics to obtain different classes of users;
s32, aiming at the different classes of users, detecting abnormal electricity utilization behaviors by adopting a KNN-based fast density peak value clustering algorithm according to load characteristics;
the S31 and S32 include: calculating any sample x in different equipment data sets of various users a With other samples x b Euclidean distance d (x) therebetween a ,x b ) And arranging the calculation results in ascending order, and recording the sample corresponding to the kth distance as N k (x a ) Then x a The k nearest neighbors of (c) are:
N(x a )={x b ∈X|d(x a ,x b )≤d(x a ,N k (x a ))}
Where K is determined by the parameter p, K = pN, p being the percentage of the number of samples N;
the distance between the KNN-based point and the high-density point is given on the basis of the formulaIs defined as:
calculating the local density of each sample and the distance between the local density and the high-density point according to a formula
Based on samples obtained by the preceding calculationDetermining a user with abnormal electricity utilization behavior; for the electric equipment sets obtained through the decomposition, the calculation is respectively carried out on different electric equipment setsThe values on the power utilization equipment sets are weighted averagely, then outliers are judged by using the weighted values, and users with abnormal power utilization behaviors are detected according to the outliers;
the judgment conditions for detecting the abnormal electricity utilization behavior are as follows: when the local density of a certain pointDistance between two adjacent platesIf so, determining that the user is a user with abnormal electricity utilization behavior;
where b represents the device, b is in {1,2, 3.., M }, a represents the sample in the dataset, a is in {1,2, 3.., N }, γ is in the dataset ρ And gamma δ Are empirical parameters.
4. The system of claim 3, wherein the load decomposition module decomposing the total power usage of the user to obtain the power usage of each device based on a non-intrusive load decomposition technique comprises:
s11, establishing a mapping relation between decomposed total power utilization and working modes of all electric equipment based on a non-invasive load decomposition model, and obtaining codes of state codes corresponding to the total power utilization of all electric equipment based on the combination of the working modes of all electric equipment;
s12, training a seq2seq model based on RNN based on the coding of the state code;
and S13, performing load decomposition by using the trained seq2seq model based on the RNN to obtain the power consumption of each device.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110570582.7A CN113344346B (en) | 2021-05-25 | 2021-05-25 | Power utilization abnormity detection method and system based on non-intrusive load decomposition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110570582.7A CN113344346B (en) | 2021-05-25 | 2021-05-25 | Power utilization abnormity detection method and system based on non-intrusive load decomposition |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113344346A CN113344346A (en) | 2021-09-03 |
CN113344346B true CN113344346B (en) | 2022-10-18 |
Family
ID=77471245
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110570582.7A Active CN113344346B (en) | 2021-05-25 | 2021-05-25 | Power utilization abnormity detection method and system based on non-intrusive load decomposition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113344346B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115112989B (en) * | 2022-08-29 | 2022-11-22 | 四川大学 | Non-invasive load monitoring method based on low-frequency data |
CN115912359B (en) * | 2023-02-23 | 2023-07-25 | 豪派(陕西)电子科技有限公司 | Digital potential safety hazard identification, investigation and treatment method based on big data |
CN116777124B (en) * | 2023-08-24 | 2023-11-07 | 国网山东省电力公司临沂供电公司 | Power stealing monitoring method based on user power consumption behavior |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20140065897A (en) * | 2012-11-22 | 2014-05-30 | 삼성전자주식회사 | Non-intrusive load monitoring apparatus and method |
CN108197751A (en) * | 2018-01-23 | 2018-06-22 | 国网山东省电力公司电力科学研究院 | Seq2seq network Short-Term Load Forecasting Methods based on multilayer Bi-GRU |
CN109490679A (en) * | 2018-12-31 | 2019-03-19 | 天津求实智源科技有限公司 | Intelligent stealing auditing system and method based on non-intrusion type load monitoring |
CN110445126A (en) * | 2019-06-25 | 2019-11-12 | 中国电力科学研究院有限公司 | A kind of non-intrusion type load decomposition method and system |
CN111428816A (en) * | 2020-04-17 | 2020-07-17 | 贵州电网有限责任公司 | Non-invasive load decomposition method |
CN112327046A (en) * | 2020-11-09 | 2021-02-05 | 北华航天工业学院 | Non-invasive load monitoring method based on fuzzy clustering and support vector regression |
CN112348096A (en) * | 2020-11-11 | 2021-02-09 | 合肥工业大学 | Non-invasive load decomposition method and system |
-
2021
- 2021-05-25 CN CN202110570582.7A patent/CN113344346B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20140065897A (en) * | 2012-11-22 | 2014-05-30 | 삼성전자주식회사 | Non-intrusive load monitoring apparatus and method |
CN108197751A (en) * | 2018-01-23 | 2018-06-22 | 国网山东省电力公司电力科学研究院 | Seq2seq network Short-Term Load Forecasting Methods based on multilayer Bi-GRU |
CN109490679A (en) * | 2018-12-31 | 2019-03-19 | 天津求实智源科技有限公司 | Intelligent stealing auditing system and method based on non-intrusion type load monitoring |
CN110445126A (en) * | 2019-06-25 | 2019-11-12 | 中国电力科学研究院有限公司 | A kind of non-intrusion type load decomposition method and system |
CN111428816A (en) * | 2020-04-17 | 2020-07-17 | 贵州电网有限责任公司 | Non-invasive load decomposition method |
CN112327046A (en) * | 2020-11-09 | 2021-02-05 | 北华航天工业学院 | Non-invasive load monitoring method based on fuzzy clustering and support vector regression |
CN112348096A (en) * | 2020-11-11 | 2021-02-09 | 合肥工业大学 | Non-invasive load decomposition method and system |
Non-Patent Citations (6)
Title |
---|
Deep Neural Network in Sequence to Short Sequence Form for Non-intrusive Load Monitoring;Jiage Liang 等;《2019 IEEE 3rd Conference on Energy Internet and Energy System Integration (EI2)》;20200409;第565-570页 * |
Seq2seq Deep Learning Method for Summary Generation by LSTM with Two-way Encoder and Beam Search Decoder;Gábor Szűcs 等;《2019 IEEE 17th International Symposium on Intelligent Systems and Informatics (SISY)》;20200609;第000221-000224页 * |
低压配电网非侵入式电力负荷异常数据辨识方法;杨蒙;《电工技术》;20200925(第18期);全文 * |
基于决策树的非入侵式负荷分解算法的研究;钟韬等;《计算机应用研究》;20200630;全文 * |
基于深度序列翻译模型的非侵入式负荷分解方法;任文龙 等;《电网技术》;20200131;第44卷(第1期);第28-33页 * |
基于设备运行状态挖掘的非侵入式负荷分解方法;庄卫金等;《电力建设》;20200801(第08期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN113344346A (en) | 2021-09-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113344346B (en) | Power utilization abnormity detection method and system based on non-intrusive load decomposition | |
US20170351288A1 (en) | Non-invasive online real-time electric load identification method and identification system | |
Basu et al. | Time series distance-based methods for non-intrusive load monitoring in residential buildings | |
CN107257351B (en) | OF flow anomaly detection system based on gray L and detection method thereof | |
Aquino et al. | Characterization of electric load with information theory quantifiers | |
CN111027408A (en) | Load identification method based on support vector machine and V-I curve characteristics | |
CN112434799B (en) | Non-invasive load identification method based on full convolution neural network | |
CN113011481B (en) | Electric energy meter function abnormality assessment method and system based on decision tree algorithm | |
CN113036759B (en) | Fine granularity identification method and identification system for power consumer load | |
CN116956198B (en) | Intelligent electricity consumption data analysis method and system based on Internet of things | |
Himeur et al. | On the applicability of 2d local binary patterns for identifying electrical appliances in non-intrusive load monitoring | |
CN111563827A (en) | Load decomposition method based on electrical appliance physical characteristics and residential electricity consumption behaviors | |
Fatouh et al. | New semi-supervised and active learning combination technique for non-intrusive load monitoring | |
Yin et al. | Non-intrusive load monitoring by load trajectory and multi feature based on DCNN | |
CN113193654A (en) | Event-driven non-intrusive power load monitoring method based on transient and steady state combination characteristics | |
CN114236234A (en) | Electrical appliance characteristic identification method based on fundamental wave and harmonic wave mixed criterion | |
CN116861316B (en) | Electrical appliance monitoring method and device | |
CN109946543A (en) | V-I figure sample data set screening technique for non-intrusion type load monitoring | |
CN111506636A (en) | System and method for analyzing residential electricity consumption behavior based on autoregressive and neighbor algorithm | |
Gurbuz et al. | A brief review of non-intrusive load monitoring and its impact on social life | |
CN115112989B (en) | Non-invasive load monitoring method based on low-frequency data | |
CN111881793A (en) | Non-invasive load monitoring method and system based on capsule network | |
CN116680641A (en) | User electricity consumption abnormality detection method based on machine learning algorithm | |
CN115828091A (en) | Non-invasive load identification method and system based on end-cloud cooperation | |
CN114884075A (en) | Event type non-intrusive load monitoring method based on supervised learning |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |