CN111932051A - Malicious behavior detection method based on non-invasive power terminal time sequence monitoring - Google Patents
Malicious behavior detection method based on non-invasive power terminal time sequence monitoring Download PDFInfo
- Publication number
- CN111932051A CN111932051A CN202010503563.8A CN202010503563A CN111932051A CN 111932051 A CN111932051 A CN 111932051A CN 202010503563 A CN202010503563 A CN 202010503563A CN 111932051 A CN111932051 A CN 111932051A
- Authority
- CN
- China
- Prior art keywords
- power
- power terminal
- detected
- time
- terminal
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 37
- 238000012544 monitoring process Methods 0.000 title claims abstract description 18
- 230000005611 electricity Effects 0.000 claims abstract description 42
- 238000000034 method Methods 0.000 claims abstract description 34
- 238000012549 training Methods 0.000 claims abstract description 16
- 230000002159 abnormal effect Effects 0.000 claims abstract description 15
- 238000000605 extraction Methods 0.000 claims abstract description 12
- 238000005265 energy consumption Methods 0.000 claims description 14
- 230000008859 change Effects 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 8
- 206010000117 Abnormal behaviour Diseases 0.000 claims description 7
- 230000005856 abnormality Effects 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 4
- 230000006399 behavior Effects 0.000 abstract description 37
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 4
- 238000004891 communication Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 230000009466 transformation Effects 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
- G06F18/24—Classification techniques
-
- 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)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Tourism & Hospitality (AREA)
- Primary Health Care (AREA)
- General Health & Medical Sciences (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Remote Monitoring And Control Of Power-Distribution Networks (AREA)
Abstract
The invention discloses a malicious behavior detection method based on non-invasive power terminal time sequence monitoring, which comprises the following steps of: s1, edge computing equipment constructs a database of historical electrical quantities of each power terminal device; s2, establishing a power consumption behavior data set for each power terminal device; s3, performing event detection and feature extraction on the electric quantity time sequence series of each power terminal to obtain a sample set, and training a load classifier; s4, carrying out event detection and feature extraction on the time sequence number series of the electric power total outlet electric quantity, inputting the event detection and feature extraction into the trained classifier model, and determining the type of equipment to which the electric power terminal to be detected belongs; s5, counting power utilization information of the power terminal to be tested; s6, detecting the electricity utilization behavior of the power terminal to be detected, and judging whether the electricity utilization equipment is abnormal or not according to the electricity utilization behavior; and S7, the edge computing equipment records and reports the abnormity. The method and the device realize the online detection of the malicious behaviors of the power terminal equipment by comparing the power consumption behaviors formed by the power terminal with historical power information based on the non-invasive power terminal time sequence monitoring.
Description
Technical Field
The invention relates to a method for detecting malicious behaviors of power terminal equipment under an edge computing system, in particular to a method for detecting malicious behaviors based on non-invasive power terminal time sequence monitoring.
Background
In recent years, with the continuous deployment of smart grid systems, smart grid devices such as smart meters play an important role in collecting data related to power usage and smart grid conditions. With the continuous emergence of new equipment and new technology in the fields of primary equipment intellectualization, wireless communication and the like, and the method is gradually and widely applied to the construction of a power communication network, so that the smart grid has the characteristics of complex access environment, flexible and various access modes, a large number of intelligent access terminals and the like, and the security risk of the smart grid is increased. Therefore, technical precautions against unsafe behavior of the smart grid are needed.
The electric power internet of things edge computing technology utilizes intelligent equipment close to a user side of an intelligent power grid and adopts an open platform integrating network, computing, storing and application core capabilities, so that services are provided nearby. Under the background, based on the calculation resource support of the edge side, the non-invasive power terminal time sequence monitoring and analysis can be carried out by adopting a complex calculation method, the malicious behavior detection of the intelligent power grid equipment level is realized, and the method has important significance for improving the safety of the intelligent power grid equipment level.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a malicious behavior detection method based on non-invasive power terminal time sequence monitoring.
The purpose of the invention is realized by the following technical scheme: a malicious behavior detection method based on non-intrusive power terminal time sequence monitoring comprises the following steps:
s1, collecting power terminal data by edge computing equipment, and constructing a historical database for a power terminal needing to be detected; the database comprises a plurality of types of the electric terminal equipment to be detected and electric quantity data thereof;
s2, extracting a time sequence of electric quantity of each power terminal in a historical database to obtain historical power consumption information of the corresponding power terminal, wherein the historical power consumption information of the power terminal comprises steady-state power, starting time, running duration and electric energy consumption, and constructing a power consumption behavior data set of the power terminal according to the historical power consumption information of the power terminal;
s3, performing event detection and feature extraction according to the historical electrical quantity time sequence number of each power terminal in the database to obtain a sample set, and training a load classifier;
s4, carrying out event detection and feature extraction on the electric quantity time sequence number series of the electric power total outlet, inputting the electric quantity time sequence number series into a trained classifier model, and determining the specific electric power terminal type to which the electric power terminal to be detected belongs;
s5, counting power consumption information of the power terminal to be measured, wherein the power consumption information specifically comprises steady-state power, starting time, running time and electric energy consumption;
s6, detecting the electricity utilization behavior of the electric power terminal to be detected according to the electricity utilization behavior data set obtained in the step S2 and the electricity utilization information of the electric power terminal to be detected obtained in the step S5, and judging whether the electric power terminal equipment is abnormal or not according to the electricity utilization behavior data set;
and S7, the edge computing equipment records the abnormal behavior of the power terminal and reports the abnormality.
Wherein the step S1 includes the following substeps:
s11, historical electric quantity data of each power terminal to be detected are collected by edge computing equipment, the historical electric quantity data comprise current and power data with time stamps, and the electric quantity data of a power main outlet comprise real-time current and power data; the data sampling frequency meets the Nyquist sampling theorem;
s12, the edge computing equipment constructs a database of historical electric quantity data and total electric outlet electric quantity data of each power terminal to be detected.
Further, the step S2 includes the following sub-steps:
s21, counting and calculating the steady-state power of each power terminal to be detected, wherein the steady-state power is a power value p when the equipment normally operates;
s22, counting and calculating the starting time of each power terminal to be detected, wherein the starting time is the power-on time t of the equipmentstartTo the moment t when the steady-state power is reachedpDifference t ofp-tstartIs denoted by ton;
S23, counting and calculating the operation time of each power terminal to be detected, wherein the operation time is the power-on time t of the equipmentstartTo the power-off time tstopDifference t ofstop-tstartIs denoted by trun;
S24, counting and calculating the power consumption of each power terminal to be detected, wherein the power consumption is the power-on time t of the equipmentstartTo the power-off time tstopIntegral of power P to operating time ^ integraltP, is marked as w;
and S25, constructing a power consumption behavior data set of the power terminal, wherein the data set comprises the steady-state power, the starting time, the running duration and the power consumption.
Further, the step S3 includes the following sub-steps:
s31, taking a sliding window for the specific power terminal power time sequence, detecting an event, and carrying out event detection at a power point PiTaking the window power sequence S ═ Pi-N…Pi…Pi+N]N is PiWindow length on both sides, 2N +1 for slidingThe overall length of the window; calculating the variance of SvarCalculating the average power S as Smean. If S isvar>>αSmeanIf alpha is a threshold control coefficient, judging that the power is suddenly changed;
s32, extracting a power terminal window current time sequence array sample from the power abrupt change point, performing Fourier series expansion, taking each current harmonic amplitude as a load characteristic, and recording as x ═ x (x ═ x-1,x2,…,xn) Wherein n is the number of odd harmonics with the largest amplitude;
s33, load characteristic sample x is usedj=(xj1,xj2,…,xjn) As input, the load classifier M, the sample label y, is trainedjWhere j is the number of detected power terminals.
The step S33 includes:
s331, forming a training set X ═ X according to the load characteristic samples1,x2,…,xmAnd the corresponding power terminal type label sequence Y is { Y ═ Y1,y2,…,ymIs the desired output; initializing sample weight D1={d11,d12,…,d1mTherein ofi is 1,2, …, m is the number of load characteristic samples in the training set;
s332, selecting the weak classifier H with the lowest current error as the kth basic classifier HkAccording to weak classifier HkClassification result of (G)t={y′1,y′2,…,y′mAnd the desired output Y ═ Y1,y2,…,ymComparing to obtain misclassified samples, and calculating HtClassification error ofWherein K is 1,2, …, and K is the number of weak classifiers;
s333, calculating the kth weak classifier HkHas a weight coefficient ofSample weight d (k) ═ d for the kth weak classifierk1,dk2,…,dkmH, updating the corresponding (k + 1) th weak classifier Hk+1The sample set weight coefficients of (a) are:
s334, executing steps S332 and S333 in a circulating mode, and obtaining the final strong classifier when the training of the K weak classifiers is finishedNamely the classifier M.
Further, the step S4 includes the following sub-steps:
s41, taking a sliding window for the total outlet power time sequence, and detecting an event at a power pointTaking window power sequenceN isThe length of the window at two sides, 2N +1 is the total length of the sliding window; to pairCalculate its variance asAverage power ofIf it is notWhereinIf the power is the threshold control coefficient, judging that the power is suddenly changed;
s42, extracting a current time sequence array sample of a power main outlet window for the power abrupt change point, performing Fourier series expansion, taking the amplitude of each current harmonic as a load characteristic, and recording as x ═ x1,x2,…,xn) Wherein n is the number of odd harmonics with the largest amplitude;
s43, load characteristic sample x ═ x1,x2,…,xn) As input, y is output by the trained load classifier MjAnd obtaining the type of the specific electric power terminal equipment.
Further, the step S5 includes the following sub-steps:
s51, calculating the steady-state power of the specific power terminal according to the event detected by the total outlet power and the obtained specific power terminal, and recording the steady-state power as p';
s52, calculating the starting time of the specific power terminal according to the event detected at the total outlet of the step S4 and the obtained specific power terminal, and recording the starting time as ton′;
S53, calculating the operation duration of the specific power terminal according to the events detected at the total outlet of the step S4 and the obtained specific power terminal, and recording the operation duration as trun′;
And S54, calculating the electric energy consumption of the specific electric power terminal according to the event detected at the total outlet of the step S4 and the obtained specific electric power terminal, and recording the electric energy consumption as w'.
Further, the step S6 includes the following sub-steps:
s61, according to the historical electricity utilization information of each power terminal in the step S2 and the electricity utilization information of the specific power terminal detected by the total exit in the step S5, carrying out steady-state power detection, if so, carrying out steady-state power detectionThe electrical equipment operates with power abnormality, whereinpIs a threshold control coefficient;
s62, according to the historical electricity utilization information of each power terminal in the step S2 and the electricity utilization information of the specific power terminal detected by the total exit in the step S5, the startup time is detected, if so, the startup time is detectedThe powered device startup process takes an exception, whereinIs a threshold control coefficient;
s63, according to the historical electricity utilization information of each power terminal in the step S2 and the electricity utilization information of the specific power terminal detected by the total outlet in the step S5, detecting the running time, if so, detecting the running timeThe running time of the electric equipment is abnormal, wherein trun,minThe shortest time length t of single operation of the power terminal in the historical databaserun,maxThe maximum operation time of the power terminal in the historical database is the maximum time of single operation;
s64, detecting the power consumption according to the historical power consumption information of each power terminal in the step S2 and the power consumption information of the specific power terminal detected by the total outlet in the step S5, if the power consumption information is detected, detecting the power consumption of the specific power terminal according to the historical power consumption information of each power terminal in the step S2The power consumption of the electric device is abnormal, whereinwIs a threshold control coefficient.
The invention has the beneficial effects that: (1) according to the method, the edge computing equipment is used for collecting power data, non-invasive power terminal time sequence monitoring is carried out, malicious behavior detection of the power terminal equipment is realized, and the safety level of the intelligent power grid power terminal equipment based on the edge computing is improved; (2) the invention utilizes the non-invasive power terminal time sequence monitoring, is superior to the traditional invasive monitoring system, can measure the power data of the power terminal without directly installing any instrument on specific power terminal equipment, and can reduce the cost of a sensor and equipment; (3) the invention utilizes the edge computing equipment to collect and compute data at the power data generation source end, and has the characteristics of timeliness and instantaneity. Meanwhile, compared with the method that the data of the source power terminal is transmitted to the power load management terminal and the master station system for centralized processing, the method greatly saves the communication bandwidth; (4) the method and the device make full use of the edge computing power, and effectively perform online detection on the malicious behaviors of the power equipment by analyzing and comparing the historical power information.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
fig. 2 is a flowchart of a method for detecting malicious behavior of power terminal equipment based on non-intrusive time-series monitoring based on edge calculation in the embodiment.
Detailed Description
The technical solutions of the present invention are further described in detail below with reference to the accompanying drawings, but the scope of the present invention is not limited to the following.
As shown in fig. 1, a malicious behavior detection method based on non-intrusive power terminal timing monitoring includes the following steps:
s1, collecting power terminal data by edge computing equipment, and constructing a historical database for a power terminal needing to be detected; the database comprises a plurality of types of the electric terminal equipment to be detected and electric quantity data thereof;
s2, extracting a time sequence of electric quantity of each power terminal in a historical database to obtain historical power consumption information of the corresponding power terminal, wherein the historical power consumption information of the power terminal comprises steady-state power, starting time, running duration and electric energy consumption, and constructing a power consumption behavior data set of the power terminal according to the historical power consumption information of the power terminal;
s3, performing event detection and feature extraction according to the historical electrical quantity time sequence number of each power terminal in the database to obtain a sample set, and training a load classifier;
s4, carrying out event detection and feature extraction on the electric quantity time sequence number series of the electric power total outlet, inputting the electric quantity time sequence number series into a trained classifier model, and determining the specific electric power terminal type to which the electric power terminal to be detected belongs;
s5, counting power consumption information of the power terminal to be measured, wherein the power consumption information specifically comprises steady-state power, starting time, running time and electric energy consumption;
s6, detecting the electricity utilization behavior of the electric power terminal to be detected according to the electricity utilization behavior data set obtained in the step S2 and the electricity utilization information of the electric power terminal to be detected obtained in the step S5, and judging whether the electric power terminal equipment is abnormal or not according to the electricity utilization behavior data set;
and S7, the edge computing equipment records the abnormal behavior of the power terminal and reports the abnormality.
Wherein the step S1 includes the following substeps:
s11, historical electric quantity data of each power terminal to be detected are collected by edge computing equipment, the historical electric quantity data comprise current and power data with time stamps, and the electric quantity data of a power main outlet comprise real-time current and power data; the data sampling frequency meets the Nyquist sampling theorem;
s12, the edge computing equipment constructs a database of historical electric quantity data and total electric outlet electric quantity data of each power terminal to be detected.
Further, the step S2 includes the following sub-steps:
s21, counting and calculating the steady-state power of each power terminal to be detected, wherein the steady-state power is a power value p when the equipment normally operates;
s22, counting and calculating the starting time of each power terminal to be detected, wherein the starting time is the power-on time t of the equipmentstartTo the moment t when the steady-state power is reachedpDifference t ofp-tstartIs denoted by ton;
S23, counting and calculating the operation time of each power terminal to be detected, wherein the operation time is the power-on time t of the equipmentstartTo the power-off time tstopDifference t ofstop-tstartIs denoted by trun;
S24, counting and calculating the power consumption of each power terminal to be detected, wherein the power consumption is calculatedThe power consumption is the time t from the power-on of the equipmentstartTo the power-off time tstopIntegral of power P to operating time ^ integraltP, is marked as w;
and S25, constructing a power consumption behavior data set of the power terminal, wherein the data set comprises the steady-state power, the starting time, the running duration and the power consumption.
Further, the step S3 includes the following sub-steps:
s31, taking a sliding window for the specific power terminal power time sequence, detecting an event, and carrying out event detection at a power point PiTaking the window power sequence S ═ Pi-N…Pi…Pi+N]N is PiThe length of the window at two sides, 2N +1 is the total length of the sliding window; calculating the variance of SvarCalculating the average power S as Smean. If S isvar>>αSmeanIf alpha is a threshold control coefficient, judging that the power is suddenly changed;
s32, extracting a power terminal window current time sequence array sample from the power abrupt change point, performing Fourier series expansion, taking each current harmonic amplitude as a load characteristic, and recording as x ═ x (x ═ x-1,x2,…,xn) Wherein n is the number of odd harmonics with the largest amplitude;
s33, load characteristic sample x is usedj=(xj1,xj2,…,xjn) As input, the load classifier M, the sample label y, is trainedjWhere j is the number of detected power terminals.
The step S33 includes:
s331, forming a training set X ═ X according to the load characteristic samples1,x2,…,xmAnd the corresponding power terminal type label sequence Y is { Y ═ Y1,y2,…,ymIs the desired output; initializing sample weight D1={d11,d12,…,d1mTherein ofi is 1,2, …, m is the number of load characteristic samples in the training set;
s332, selecting the weak classifier H with the lowest current error as the kth basic classifier HkAccording to weak classifier HkClassification result of (G)t={y′1,y′2,…,y′mAnd the desired output Y ═ Y1,y2,…,ymComparing to obtain misclassified samples, and calculating HtClassification error ofWherein K is 1,2, …, and K is the number of weak classifiers;
s333, calculating the kth weak classifier HkHas a weight coefficient ofSample weight d (k) ═ d for the kth weak classifierk1,dk2,…,dkmH, updating the corresponding (k + 1) th weak classifier Hk+1The sample set weight coefficients of (a) are:
s334, executing steps S332 and S333 in a circulating mode, and obtaining the final strong classifier when the training of the K weak classifiers is finishedNamely the classifier M.
Further, the step S4 includes the following sub-steps:
s41, taking a sliding window for the total outlet power time sequence, and detecting an event at a power pointTaking window power sequenceN isThe length of the window at two sides, 2N +1 is the total length of the sliding window; to pairCalculate its variance asAverage power ofIf it is notWhereinIf the power is the threshold control coefficient, judging that the power is suddenly changed;
s42, extracting a current time sequence array sample of a power main outlet window for the power abrupt change point, performing Fourier series expansion, taking the amplitude of each current harmonic as a load characteristic, and recording as x ═ x1,x2,…,xn) Wherein n is the number of odd harmonics with the largest amplitude;
s43, load characteristic sample x ═ x1,x2,…,xn) As input, y is output by the trained load classifier MjAnd obtaining the type of the specific electric power terminal equipment.
Further, the step S5 includes the following sub-steps:
s51, calculating the steady-state power of the specific power terminal according to the event detected by the total outlet power and the obtained specific power terminal, and recording the steady-state power as p';
s52, calculating the starting time of the specific power terminal according to the event detected at the total outlet of the step S4 and the obtained specific power terminal, and recording the starting time as ton′;
S53, according to the aboveStep S4, calculating the operation time length of the specific power terminal and recording the operation time length as t according to the event detected at the total outlet of the step S4 and the obtained specific power terminalrun′;
And S54, calculating the electric energy consumption of the specific electric power terminal according to the event detected at the total outlet of the step S4 and the obtained specific electric power terminal, and recording the electric energy consumption as w'.
Further, the step S6 includes the following sub-steps:
s61, according to the historical electricity utilization information of each power terminal in the step S2 and the electricity utilization information of the specific power terminal detected by the total exit in the step S5, carrying out steady-state power detection, if so, carrying out steady-state power detectionThe electrical equipment operates with power abnormality, whereinpIs a threshold control coefficient;
s62, according to the historical electricity utilization information of each power terminal in the step S2 and the electricity utilization information of the specific power terminal detected by the total exit in the step S5, the startup time is detected, if so, the startup time is detectedThe powered device startup process takes an exception, whereinIs a threshold control coefficient;
s63, according to the historical electricity utilization information of each power terminal in the step S2 and the electricity utilization information of the specific power terminal detected by the total outlet in the step S5, detecting the running time, if so, detecting the running timeThe running time of the electric equipment is abnormal, wherein trun,minThe shortest time length t of single operation of the power terminal in the historical databaserun,maxThe maximum operation time of the power terminal in the historical database is the maximum time of single operation;
s64, according to the historical electricity utilization information of each power terminal in the step S2 and the electricity utilization information of the specific power terminal detected by the total exit in the step S5Information, detecting the power consumption, ifThe power consumption of the electric device is abnormal, whereinwIs a threshold control coefficient.
As shown in fig. 2, in the embodiment of the application, according to power terminal data collected by an edge computing side of a smart grid, detection of malicious behavior of electric equipment on the edge computing side is implemented by using non-intrusive power terminal timing monitoring, which specifically includes the following processes:
1. the method comprises the steps that edge computing equipment collects historical power data of a power terminal to be detected and builds a historical current and power database for the power terminal equipment; meanwhile, collecting power data of a power main outlet, and constructing a current and power database with a timestamp on the power main outlet;
2. obtaining historical power utilization information of the power terminal according to the time sequence number series of each specific power terminal in the historical database of the power terminal equipment, wherein the historical power utilization information specifically comprises steady-state power p and starting time tonLength of operation trunThe electric energy consumption w is obtained, and an electric power terminal electricity consumption behavior data set is constructed according to the historical electricity consumption information of the electric power terminal;
3. according to the electrical quantity time sequence series of each power terminal in the database, performing event detection and feature extraction to obtain a sample set, and training a load classifier;
a) taking a sliding window for the power time sequence of each specific power terminal, and detecting events at a power point PiTaking the window power sequence S ═ Pi-N…Pi…Pi+N]N is PiThe length of the window at two sides, 2N +1 is the total length of the sliding window; calculating the variance of SvarCalculating the average power S as Smean. When S isvra>>αSmeanAnd if alpha is a threshold control coefficient and the power is judged to be suddenly changed at the moment, the event occurs in the power terminal.
b) Extracting window current samples of the power terminal from the power abrupt change points, performing Fourier series expansion, taking the harmonic amplitude of each current as a load characteristic, and recordingIs x ═ x1,x2,…,xn) Where n is the number of odd harmonics with the largest amplitude, where n is 7.
c) When all the load characteristics of the power terminal to be detected are extracted, each load characteristic sample X is set as { X ═ X }1,x2,…,xmAs input, the corresponding power terminal type tag sequence Y ═ Y1,y2,…,ymIs the desired output. Initializing sample weight D1={d11,d12,…,d1mTherein ofi is 1,2, …, m is the number of load characteristic samples in the training set;
d) selecting the weak classifier H with the lowest current error as the kth basic classifier HkAccording to weak classifier HkClassification result of (G)t={y′1,y′2,…,y′mAnd the desired output Y ═ Y1,y2,…,ymComparing to obtain misclassified samples, and calculating HtClassification error ofWherein K is 1,2, …, and K is the number of weak classifiers;
e) computing the kth weak classifier HkHas a weight coefficient ofSample weight d (k) ═ d for the kth weak classifierk1,dk2,…,dkmH, updating the corresponding (k + 1) th weak classifier Hk+1The sample set weight coefficient ofWhere Z iskIs a normalization factor that is a function of the normalization factor,
f) circularly executing the step d) and the step e) when K weak classifiers are usedTraining is completed to obtain the final strong classifierNamely a classifier M;
4. event detection and feature extraction are carried out on the time sequence series of the total power outlet, the time sequence series are input into a trained classifier model M, and the specific equipment type of the power terminal to be detected is determined;
a) taking a sliding window for the total outlet power time sequence, detecting the event, and carrying out the detection at the power pointTaking window power sequenceN isThe length of the window at two sides, 2N +1 is the total length of the sliding window; to pairCalculate its variance asAverage power ofWhen in useWhereinAnd if the total power outlet power is a threshold control coefficient, judging that the total power outlet power suddenly changes, and detecting that the behavior event of the power terminal occurs.
b) And extracting a current sample of a total power outlet window of the power catastrophe point, performing Fourier series expansion, taking the amplitude of each current harmonic as a load characteristic, and recording as x ═ x (x)1,x2,…,xn) Wherein n isThe number of odd harmonics with the largest amplitude is taken, where n is 7.
c) Obtaining a load characteristic sample x ═ x (x) as described above1,x2,…,x7) As input, y is output by the trained load classifier MjAnd obtaining the specific electric power terminal to which the power abrupt change point corresponds to the event.
5. Counting the detected current-period power utilization information of the power terminal, specifically including the steady-state power p' and the starting time ton', length of operation trun', power consumption w';
6. detecting the current time period electricity consumption behavior of the electric power terminal to be detected according to the electricity consumption behavior data set in the step 2 and the current time period electricity consumption information of the electric power terminal to be detected in the step 5, and judging whether the electric power terminal equipment is abnormal or not according to the current time period electricity consumption behavior;
7. and (3) the edge computing device records the abnormal behavior monitored in the step (6) and reports the abnormal behavior, and under a general condition, the monitored abnormal behavior can be reported to a monitoring center communicated with the edge computing device, and the abnormal behavior can also be reported to a handheld mobile terminal communicated with the edge computing device.
In the embodiment of the present application, the power terminal electricity consumption information in step S2 includes, but is not limited to, steady state power, start time, operation duration, and power consumption. The power utilization information of the power terminal can be represented by instantaneous peak power, frequency spectrum jitter, frequency variation and the use sequence of the power terminal. The feature extraction method in step S3 and step S4 adopts fourier series expansion of current to extract odd harmonics as features, and may also adopt wavelet transformation, variable point detection, and edge detection methods to extract feature tuples. Step S3 may also use, for example, a k-nearest neighbor algorithm, an SVM algorithm, and a decision tree algorithm, or a convolutional neural network algorithm, a feed-forward neural network algorithm, and a radial basis function neural network algorithm to construct a corresponding neural network, in addition to the machine learning algorithm to construct the load classification model, and train the neural network using the training set to obtain a corresponding mature model.
In conclusion, the method adopts non-invasive power terminal time sequence monitoring, the collected power terminal data are decomposed on line, and the malicious behavior of the power terminal equipment is detected on line by comparing the collected power terminal data with the power consumption behavior formed by historical power information, so that the safety guarantee of the power terminal equipment level of the smart grid under the edge calculation is realized; compared with the traditional invasive monitoring method, the non-invasive introduction greatly reduces the use of the sensor and effectively reduces the detection cost; the event detection algorithm based on the sliding window is designed, so that the change of the power utilization behavior of the power terminal equipment can be accurately judged, and meanwhile, the influence of system noise and short-time peaks can be reduced; the load identification is carried out by adopting a machine learning algorithm, so that the accuracy of the load identification is effectively improved; the electric power information of the electric power terminal is collected and processed by utilizing the computing performance of the edge computing device, so that the malicious behavior of the electric power terminal equipment is detected on line, the characteristics of real-time performance and timeliness are met, and compared with the method that the data of the source electric power terminal is transmitted to the electric power load management terminal and the main station system for centralized processing, the communication bandwidth is greatly saved and the data delay is effectively avoided.
The foregoing is a preferred embodiment of the present invention, it is to be understood that the invention is not limited to the form disclosed herein, but is not to be construed as excluding other embodiments, and is capable of other combinations, modifications, and environments and is capable of changes within the scope of the inventive concept as expressed herein, commensurate with the above teachings, or the skill or knowledge of the relevant art. And that modifications and variations may be effected by those skilled in the art without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (8)
1. A malicious behavior detection method based on non-invasive power terminal time sequence monitoring is characterized in that: the method comprises the following steps:
s1, collecting power terminal data by edge computing equipment, and constructing a historical database for a power terminal needing to be detected; the database comprises a plurality of types of the electric terminal equipment to be detected and electric quantity data thereof;
s2, extracting a time sequence of electric quantity of each power terminal in a historical database to obtain historical power consumption information of the corresponding power terminal, wherein the historical power consumption information of the power terminal comprises steady-state power, starting time, running duration and electric energy consumption, and constructing a power consumption behavior data set of the power terminal according to the historical power consumption information of the power terminal;
s3, performing event detection and feature extraction according to the historical electrical quantity time sequence number of each power terminal in the database to obtain a sample set, and training a load classifier;
s4, carrying out event detection and feature extraction on the electric quantity time sequence number series of the electric power total outlet, inputting the electric quantity time sequence number series into a trained classifier model, and determining the specific electric power terminal type to which the electric power terminal to be detected belongs;
s5, counting power consumption information of the power terminal to be measured, wherein the power consumption information specifically comprises steady-state power, starting time, running time and electric energy consumption;
s6, detecting the electricity utilization behavior of the electric power terminal to be detected according to the electricity utilization behavior data set obtained in the step S2 and the electricity utilization information of the electric power terminal to be detected obtained in the step S5, and judging whether the electric power terminal equipment is abnormal or not according to the electricity utilization behavior data set;
and S7, the edge computing equipment records the abnormal behavior of the power terminal and reports the abnormality.
2. The method of claim 1, wherein the method comprises the following steps: the step S1 includes the following sub-steps:
s11, historical electric quantity data of each power terminal to be detected are collected by edge computing equipment, the historical electric quantity data comprise current and power data with time stamps, and the electric quantity data of a power main outlet comprise real-time current and power data; the data sampling frequency meets the Nyquist sampling theorem;
s12, the edge computing equipment constructs a database of historical electric quantity data and total electric outlet electric quantity data of each power terminal to be detected.
3. The method of claim 1, wherein the method comprises the following steps: the step S2 includes the following sub-steps:
s21, counting and calculating the steady-state power of each power terminal to be detected, wherein the steady-state power is a power value p when the equipment normally operates;
s22, counting and calculating the starting time of each power terminal to be detected, wherein the starting time is the power-on time t of the equipmentstartTo the moment t when the steady-state power is reachedpDifference t ofp-tstartIs denoted by ton;
S23, counting and calculating the operation time of each power terminal to be detected, wherein the operation time is the power-on time t of the equipmentstartTo the power-off time tstopDifference t ofstop-tstartIs denoted by trun;
S24, counting and calculating the power consumption of each power terminal to be detected, wherein the power consumption is the power-on time t of the equipmentstartTo the power-off time tstopIntegral of power P to operating time ^ integraltP, is marked as w;
and S25, constructing a power consumption behavior data set of the power terminal, wherein the data set comprises the steady-state power, the starting time, the running duration and the power consumption.
4. The method of claim 1, wherein the method comprises the following steps: the step S3 includes the following sub-steps:
s31, taking a sliding window for the specific power terminal power time sequence, detecting an event, and carrying out event detection at a power point PiTaking the window power sequence S ═ Pi-N…Pi…Pi+N]N is PiThe length of the window at two sides, 2N +1 is the total length of the sliding window; calculating the variance of SvarCalculating the average power S as Smean. If S isvar>>αSmeanIf alpha is a threshold control coefficient, judging that the power is suddenly changed;
s32, extracting a power terminal window current time sequence array sample from the power abrupt change point, performing Fourier series expansion, taking each current harmonic amplitude as a load characteristic, and recording as x ═ x (x ═ x-1,x2,…,xn) Wherein n is the number of odd harmonics with the largest amplitude;
s33, load characteristic sample x is usedj=(xj1,xj2,…,xjn) As input, the load classifier M, the sample label y, is trainedjWhere j is the number of detected power terminals.
5. The method of claim 1, wherein the method comprises the following steps: the step S4 includes the following sub-steps:
s41, taking a sliding window for the total outlet power time sequence, and detecting an event at a power pointTaking window power sequenceN isThe length of the window at two sides, 2N +1 is the total length of the sliding window; to pairCalculate its variance asAverage power ofIf it is notWhereinIf the power is the threshold control coefficient, judging that the power is suddenly changed;
s42, extracting a current time sequence array sample of a power main outlet window for the power abrupt change point, performing Fourier series expansion, taking the amplitude of each current harmonic as a load characteristic, and recording as x ═ x1,x2,…,xn) Wherein n is the number of odd harmonics with the largest amplitude;
s43, load characteristic sample x ═ x1,x2,…,xn) As input, y is output by the trained load classifier MjAnd obtaining the type of the specific electric power terminal equipment.
6. The method of claim 1, wherein the method comprises the following steps: the step S5 includes the following sub-steps:
s51, calculating the steady-state power of the specific power terminal according to the event detected by the total outlet power and the obtained specific power terminal, and recording the steady-state power as p';
s52, calculating the starting time of the specific power terminal according to the event detected at the total outlet of the step S4 and the obtained specific power terminal, and recording the starting time as ton′;
S53, according to the events detected at the total outlet of the step S4 and the obtained specific power terminal, calculating the running time of the specific power terminal, and recording the running time as the running timetrun′;
And S54, calculating the electric energy consumption of the specific electric power terminal according to the event detected at the total outlet of the step S4 and the obtained specific electric power terminal, and recording the electric energy consumption as w'.
7. The method of claim 1, wherein the method comprises the following steps: the step S6 includes the following sub-steps:
s61, according to the historical electricity utilization information of each power terminal in the step S2 and the electricity utilization information of the specific power terminal detected by the total exit in the step S5, carrying out steady-state power detection, if so, carrying out steady-state power detectionThe electrical equipment operates with power abnormality, whereinpIs a threshold control coefficient;
s62, according to the historical electricity utilization information of each power terminal in the step S2 and the electricity utilization information of the specific power terminal detected by the total exit in the step S5, the startup time is detected, if so, the startup time is detectedThe powered device startup process takes an exception, whereinIs a threshold control coefficient;
s63, according to the historical electricity utilization information of each power terminal in the step S2 and the electricity utilization information of the specific power terminal detected by the total outlet in the step S5, detecting the running time, if so, detecting the running timeThe running time of the electric equipment is abnormal, wherein trun,minThe shortest time length t of single operation of the power terminal in the historical databaserun,maxThe maximum operation time of the power terminal in the historical database is the maximum time of single operation;
s64, according to each electric power terminal in the step S2The terminal history power consumption information and the power consumption information of the specific power terminal detected in the step S5, if it is detected that the power consumption is not enoughThe power consumption of the electric device is abnormal, whereinwIs a threshold control coefficient.
8. The method of claim 4, wherein the malicious behavior detection method is based on the non-intrusive power terminal timing monitoring, and comprises: the step S33 includes:
s331, forming a training set X ═ X according to the load characteristic samples1,x2,…,xmAnd the corresponding power terminal type label sequence Y is { Y ═ Y1,y2,…,ymIs the desired output; initializing sample weight D1={d11,d12,…,d1mTherein ofm is the number of load characteristic samples in the training set;
s332, selecting the weak classifier H with the lowest current error as the kth basic classifier HkAccording to weak classifier HkClassification result of (G)t={y′1,y′2,…,y′mAnd the desired output Y ═ Y1,y2,…,ymComparing to obtain misclassified samples, and calculating HtClassification error ofWherein K is 1,2, …, and K is the number of weak classifiers;
s333, calculating the kth weak classifier HkHas a weight coefficient ofSample weight d (k) ═ d for the kth weak classifierk1,dk2,…,dkmUpdate the corresponding k +1 th weakClassifier Hk+1The sample set weight coefficients of (a) are:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010503563.8A CN111932051A (en) | 2020-06-05 | 2020-06-05 | Malicious behavior detection method based on non-invasive power terminal time sequence monitoring |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010503563.8A CN111932051A (en) | 2020-06-05 | 2020-06-05 | Malicious behavior detection method based on non-invasive power terminal time sequence monitoring |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111932051A true CN111932051A (en) | 2020-11-13 |
Family
ID=73317063
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010503563.8A Pending CN111932051A (en) | 2020-06-05 | 2020-06-05 | Malicious behavior detection method based on non-invasive power terminal time sequence monitoring |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111932051A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114090406A (en) * | 2021-11-29 | 2022-02-25 | 中国电力科学研究院有限公司 | Electric power Internet of things equipment behavior safety detection method, system, equipment and storage medium |
CN114158029A (en) * | 2021-12-03 | 2022-03-08 | 国网上海市电力公司 | Power acquisition terminal network flow monitoring method based on big data analysis |
CN114202011A (en) * | 2021-10-30 | 2022-03-18 | 湖南江军科技有限责任公司 | Power utilization safety sensing method |
CN114881167A (en) * | 2022-05-24 | 2022-08-09 | 北京百度网讯科技有限公司 | Abnormality detection method, abnormality detection device, electronic apparatus, and medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110086776A (en) * | 2019-03-22 | 2019-08-02 | 国网河南省电力公司经济技术研究院 | Intelligent substation Network Intrusion Detection System and detection method based on deep learning |
US20190361065A1 (en) * | 2018-05-22 | 2019-11-28 | Guangdong University Of Technology | Method and device for reliability assessment of wide area protection system |
CN110555369A (en) * | 2019-07-16 | 2019-12-10 | 浙江工业大学 | MLCDTL-based non-intrusive load identification method |
CN110954744A (en) * | 2019-11-18 | 2020-04-03 | 浙江工业大学 | Non-invasive load monitoring method based on event detection |
-
2020
- 2020-06-05 CN CN202010503563.8A patent/CN111932051A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190361065A1 (en) * | 2018-05-22 | 2019-11-28 | Guangdong University Of Technology | Method and device for reliability assessment of wide area protection system |
CN110086776A (en) * | 2019-03-22 | 2019-08-02 | 国网河南省电力公司经济技术研究院 | Intelligent substation Network Intrusion Detection System and detection method based on deep learning |
CN110555369A (en) * | 2019-07-16 | 2019-12-10 | 浙江工业大学 | MLCDTL-based non-intrusive load identification method |
CN110954744A (en) * | 2019-11-18 | 2020-04-03 | 浙江工业大学 | Non-invasive load monitoring method based on event detection |
Non-Patent Citations (3)
Title |
---|
刘明: "基于行为特征的非侵入式负荷识别方法研究", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 * |
周明等: "基于非侵入式负荷监测的居民用电行为分析", 《电网技术》 * |
王爱囡等: "一种非侵入式电力负荷监测***的设计与实现", 《西安理工大学学报》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114202011A (en) * | 2021-10-30 | 2022-03-18 | 湖南江军科技有限责任公司 | Power utilization safety sensing method |
CN114090406A (en) * | 2021-11-29 | 2022-02-25 | 中国电力科学研究院有限公司 | Electric power Internet of things equipment behavior safety detection method, system, equipment and storage medium |
CN114158029A (en) * | 2021-12-03 | 2022-03-08 | 国网上海市电力公司 | Power acquisition terminal network flow monitoring method based on big data analysis |
CN114881167A (en) * | 2022-05-24 | 2022-08-09 | 北京百度网讯科技有限公司 | Abnormality detection method, abnormality detection device, electronic apparatus, and medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111932051A (en) | Malicious behavior detection method based on non-invasive power terminal time sequence monitoring | |
Zufferey et al. | Machine learning approaches for electric appliance classification | |
CN104966161B (en) | A kind of power quality recorder data calculation and analysis methods based on gauss hybrid models | |
CN111382789B (en) | Power load identification method and system based on machine learning | |
CN110416995B (en) | Non-invasive load decomposition method and device | |
CN111027408A (en) | Load identification method based on support vector machine and V-I curve characteristics | |
CN112598303A (en) | Non-invasive load decomposition method based on combination of 1D convolutional neural network and LSTM | |
CN110569876A (en) | Non-invasive load identification method and device and computing equipment | |
Rehman et al. | Comparative evaluation of machine learning models and input feature space for non-intrusive load monitoring | |
Dash et al. | An appliance load disaggregation scheme using automatic state detection enabled enhanced integer programming | |
Chen et al. | Non-intrusive load monitoring based on feature extraction of change-point and xgboost classifier | |
Schirmer et al. | Double Fourier integral analysis based convolutional neural network regression for high-frequency energy disaggregation | |
CN111579978A (en) | System and method for realizing relay fault identification based on artificial intelligence technology | |
CN114859169A (en) | Intelligent identification method and system for distribution transformer outgoing line load and storage medium | |
Yu et al. | Multi-time-scale shapelet-based feature extraction for non-intrusive load monitoring | |
CN113193654A (en) | Event-driven non-intrusive power load monitoring method based on transient and steady state combination characteristics | |
CN116681186A (en) | Power quality analysis method and device based on intelligent terminal | |
CN111239484A (en) | Non-invasive load electricity consumption information acquisition method for non-resident users | |
Kommey et al. | An artificial intelligence‐based non‐intrusive load monitoring of energy consumption in an electrical energy system using a modified K‐Nearest Neighbour algorithm | |
CN116365707A (en) | Intelligent power consumption monitoring system | |
Weißhaar et al. | Device classification for NILM using FIT-PS compared with standard signal forms | |
CN115687999A (en) | Non-invasive load sensing method fusing multi-time scale electric appliance characteristics | |
CN115169405A (en) | Hotel guest room equipment fault diagnosis method and system based on support vector machine | |
Yang et al. | Transient fault diagnosis of track circuit based on MFCC-DTW | |
CN114676783A (en) | Load identification method based on single classification and fuzzy width 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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201113 |