CN112949563A - Non-invasive load identification method based on variable point detection and improved KNN algorithm - Google Patents

Non-invasive load identification method based on variable point detection and improved KNN algorithm Download PDF

Info

Publication number
CN112949563A
CN112949563A CN202110317472.XA CN202110317472A CN112949563A CN 112949563 A CN112949563 A CN 112949563A CN 202110317472 A CN202110317472 A CN 202110317472A CN 112949563 A CN112949563 A CN 112949563A
Authority
CN
China
Prior art keywords
period
voltage
load
current
electric
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.)
Granted
Application number
CN202110317472.XA
Other languages
Chinese (zh)
Other versions
CN112949563B (en
Inventor
孙耀杰
胡敏政
延菲
李昕然
马磊
孙洁
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fudan University
Zhuhai Fudan Innovation Research Institute
Original Assignee
Fudan University
Zhuhai Fudan Innovation Research Institute
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Fudan University, Zhuhai Fudan Innovation Research Institute filed Critical Fudan University
Priority to CN202110317472.XA priority Critical patent/CN112949563B/en
Publication of CN112949563A publication Critical patent/CN112949563A/en
Application granted granted Critical
Publication of CN112949563B publication Critical patent/CN112949563B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/088—Non-supervised learning, e.g. competitive learning
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S20/00—Management or operation of end-user stationary applications or the last stages of power distribution; Controlling, monitoring or operating thereof
    • Y04S20/20—End-user application control systems
    • Y04S20/242—Home appliances

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Signal Processing (AREA)
  • Supply And Distribution Of Alternating Current (AREA)

Abstract

The invention is a non-invasive load identification method based on variable point detection and improved KNN algorithm, firstly, the switching time point of an electric appliance is judged by using a variable point detection method, and then the current and voltage before and after the time point are differentiated by using the specific time point; and then, carrying out feature extraction by using the obtained current-voltage difference value, finally calculating the comprehensive similarity between the unknown load and various electrical appliances in the established feature set, and taking the maximum similarity as the judgment result of the unknown load. The invention takes into account the computational power requirements of non-intrusive load detection in practical engineering applications. The invention adopts a variable point detection method with lower computational power demand and a pattern recognition algorithm, has low sensor quantity required by non-invasive load detection, can reduce the cost of household energy management, and has important significance for formulating the scheduling strategy of the electric equipment and the energy storage system, reducing the electricity consumption cost, realizing peak clipping and valley filling and promoting the safe and stable operation of the regional power grid.

Description

Non-invasive load identification method based on variable point detection and improved KNN algorithm
Technical Field
The invention belongs to the field of electrical engineering, and particularly relates to a non-intrusive load identification method based on variable point detection and an improved KNN algorithm.
Background
The household energy management system is an important component of the smart grid as an extension of the smart grid on a resident side, and the realization of many functions of the smart grid needs the support of the household energy management system. Non-intrusive Load Monitoring (NILM) only needs to install a measuring device at a main switch of a service line, acquires terminal voltage and total current of a user, and then distributes total energy consumption information to each piece of electric equipment by using a data analysis technology. Therefore, for the household energy management system, a non-invasive load identification method with high accuracy, low cost and suitability for complex practical scenes is still very important, and the contradiction between the computing power requirement of the balance algorithm and the installation and maintenance cost is still a difficult problem of the household energy management system.
Disclosure of Invention
The invention aims to provide a non-invasive load identification method based on variable point detection and improved KNN algorithm, which is rapid, low in cost, high in accuracy and strong in robustness, so that the rapidity and economy of non-invasive load decomposition are remarkably improved, and the load decomposition accuracy rate is more than 95%.
The invention provides a non-intrusive load identification method based on variable point detection and an improved KNN algorithm, which comprises the following specific steps:
(1) the non-invasive load identification method is realized by a non-invasive load identification system, the system consists of a power supply system, an electric signal acquisition system, a bus deconcentrator, a load decomposition module and various electric appliances to be detected, the power supply system is connected with the electric signal acquisition system through a first circuit breaker, the electric signal acquisition system is connected with the load decomposition module, the electric signal acquisition system is connected with the bus deconcentrator, the bus deconcentrator is respectively connected with the various electric appliances to be detected through different circuit breakers, and the power supply system is used for supplying power to the electric signal acquisition system; each circuit breaker is used for respectively controlling the switching-in and switching-out actions of the electric appliance; the electric signal acquisition system is used for acquiring electric signal data such as voltage, current and the like at the bus deconcentrator; the electrical appliance to be tested can select various different types of electrical appliances for feature set establishment and testing; the bus deconcentrator is used for merging different types of electric appliances to be tested into a bus, realizing the aliasing of loads and simulating an actual electric-using scene;
(2) establishing a characteristic set, putting a single electric appliance into a non-invasive load identification system, measuring voltage data u (t) and current data i (t) of the single electric appliance during stable operation by using the electric signal measuring system, carrying out Fourier transform on electric appliance data of a period, taking a fundamental wave voltage phase as 0 as a reference point, extracting a voltage instantaneous value u (t) in the period and current instantaneous value data i (t) in the period, and obtaining the voltage data u (t) in the periodT(t) voltage data i in the periodT(t) according to uT(t)，iT(t), making a V-I track, extracting higher harmonics, active power and reactive power, and taking the binarized V-I track, higher harmonics, active power and reactive power as the feature set of the electrical appliance; repeating the steps for different kinds of electric appliances to respectively obtain feature sets of different electric appliances;
(3) performing aliasing input test on a plurality of electric appliances, inputting or cutting the plurality of electric appliances within a period of time t, obtaining a voltage instantaneous sampling discrete value U (i) of the period within t seconds and a current instantaneous sampling discrete value I (i) of the period through an electric signal acquisition system, and calculating active power, wherein a formula (1) is a definition expression of the active power, and a formula (2) is an active power calculation method under discrete data;
Figure BDA0002991757300000021
wherein: p is the active power in the period, T is the period, u (T), i (T) are the voltage and current instantaneous values of the period respectively;
the active power calculation method adopted in the step is shown in formula (2):
Figure BDA0002991757300000022
p is the active power in the period, T is the period, U (i), I (i) are instantaneous sampling discrete values of the voltage and the current of the period, and Fs is the sampling frequency;
(4) the method comprises the steps of variable point detection and characteristic extraction, wherein the time point of active power P change is determined by adopting a bilateral CUSUM algorithm or a standard deviation multiple method based on active power in a time period t, namely the time point t of electric appliance inputm(ii) a The extraction time intervals are respectively [ tm，tm+1]And [ tm-1，tm]Current-voltage signal of one period in the period, and voltage of one period before the change point is UoffThe current in a period before the change point is IoffThe voltage in a period after the change point is UonThe current in a period after the change point is IonDefined as the abscissa of the V-I locus
Figure BDA0002991757300000023
Ordinate Im＝Ion-IoffTo U withm-ImMaking the V-I trajectory of the unknown load as a trajectory feature, and applying to the Um、ImPerforming Fourier transform, and extracting harmonic amplitude, reactive power and the like as amplitude characteristics;
(5) and (4) load identification, namely calculating comprehensive similarity of the track characteristics and the amplitude characteristics in the step (4) and the electrical characteristics in the known electrical appliance characteristic set respectively, adopting an improved KNN algorithm to calculate the comprehensive similarity which is to be detected and accords with various known load characteristic sets respectively, and taking the result of the maximum similarity as the prediction result of the unknown load.
In the invention, the load identification method in the step (5) adopts a KNN algorithm for averagely distributing weights, belongs to an improved KNN algorithm, and comprises the following steps of:
(5.1) the sum of the weights of the training samples of each category is set to be 1, when the samples are averagely distributed, the weight distributed to each sample is only related to the number of the samples contained in the category to which the sample belongs, and the weight (T) is weightj) The calculation method is as follows:
weight(Tj)＝1/size(CTj) (3)
wherein, size (C)Tj) Represents TjThe number of training samples contained in the category;
(5.2) calculating the V-I track similarity and the amplitude similarity of the sample to be detected and all training samples, and respectively recording as Sim1 and Sim 2:
sim1＝1/(1+dist1) (4)
sim2＝1/(1+dist2) (5)
wherein dist1 and dist2 are distances of a distance and an amplitude of a V-I track between two samples respectively, and are Euclidean distances. (ii) a
(5.3) arranging the training samples according to the size of Sim1 in a descending order, and taking the first K training samples with the largest track similarity as the K nearest neighbor of the current test sample;
(5.4) calculating the current test sample and the K nearest neighbor TjThe combined similarity Sim of the trajectory feature and the amplitude feature of (j ═ 1, 2.., K):
sim(a，Tj)＝sim1(a，Tj)×weight(Tj)+sim2(a，Tj) (4)
and (5.5) counting the total comprehensive similarity of the sample to be detected and each class in the K nearest neighbors, and taking the class with the maximum total comprehensive similarity as a prediction result.
In the invention, the characteristic set establishment in the step (2) takes various factors into consideration, and a method for combining the intuitive track characteristic with the amplitude characteristic with strong physical significance is selected. The track characteristics have higher distinguishing efficiency on the properties of the electric appliance, such as distinguishing pure resistive load from non-pure resistive load; meanwhile, the amplitude characteristics can realize further load subdivision through differences of reactive power, harmonic amplitude and the like under the condition that V-I tracks are similar. In the invention, the active power calculation method in the step (3) is suitable for discrete sampling, can be combined with various data acquisition systems for use, and has low requirement on calculated amount.
In the invention, the variable point detection method in the step (4) can adopt a bilateral CUSUM algorithm or a standard deviation multiple method, has certain flexibility, has definite physical significance for the characteristic extraction method, has strong robustness, has low requirements on software and hardware, and is suitable for being widely used under the conditions of low computational power and low cost.
The invention has the beneficial effects that: firstly, the method improves the existing KNN algorithm, overcomes the problem that the category with less samples is easy to misjudge when the data set has unbalance, namely the number difference of the samples of different categories is larger, and ensures the accuracy of load identification under the condition of using a low-cost unsupervised algorithm based on the algorithm; secondly, the method combines various electrical characteristics, establishes a characteristic set and has a wider application range; finally, the invention uses the high-efficiency variable point detection algorithm, and can accurately identify the load switching time point.
Drawings
FIG. 1 is a schematic system design diagram of a non-intrusive load identification method based on switching event change point detection and electrical appliance characteristic extraction adopted in the invention;
FIG. 2 is a flow chart of the improved KNN algorithm for realizing load decomposition;
FIG. 3 is an overall experimental flow diagram of the inventive load split;
FIG. 4 is a flow diagram of the feature set establishment of the invention;
figure 5 shows the Macro-F1 values identified for a single load for the four algorithms.
Reference numbers in the figures: the system comprises a power supply system 1, a general commercial power, an electric signal acquisition device 2, a wire divider 3, various to-be-tested electric appliances 4-1, 4-2 and the like, an incandescent lamp 4-1, an LED lamp 4-2, a dryer 4-3 and a load decomposition module 5. The four algorithms are respectively:
algorithm 1: conventional kNN considering only trajectory features;
and 2, algorithm: considering only the weighted kNN of the trajectory features;
algorithm 3: conventional kNN considering trajectory and amplitude features;
and algorithm 4: the weighting kNN of the trajectory features and the magnitude features is considered.
Detailed Description
The invention is further illustrated by the following examples in conjunction with the accompanying drawings.
Example 1: as shown in fig. 1, the schematic diagram of the non-intrusive load identification method based on switching event change point detection and the improved KNN algorithm is composed of a mains supply system, an electrical signal acquisition device, a load decomposition module, a circuit breaker, and a plurality of electrical appliances to be detected (in this embodiment, three electrical appliances are selected, namely, an LED lamp, an incandescent lamp, and a dryer), wherein 4-1 represents the incandescent lamp, 4-2 represents the LED lamp, and 4-3 represents the dryer.
Step (1), firstly closing a first breaker switch S1 and a second breaker S2, independently connecting an incandescent lamp, extracting feature sets of the incandescent lamp through data processing of a measuring system and a load decomposition module, wherein each set of features comprises a V-I track, 3 harmonic waves, 5 harmonic waves and 7 harmonic wave amplitudes in stable operation, and repeating an experiment n4-1Secondly, the incandescent lamp feature set contains n in total4-1Data for data, where the experiment was repeated 94 times, n4-1＝94；
And (2) opening a second breaker switch S2, closing a third breaker switch S3, independently accessing the LED lamps, extracting feature sets of the LED lamps through data processing of a measurement system and a load decomposition module, wherein each set of features comprises a V-I track, 3-order harmonic, 5-order harmonic and 7-order harmonic amplitude during stable operation, and repeating an experiment n4-2Secondly, the LED lamp feature set contains n in total4-2Data for data, where experiment was repeated 138 times, n4-2＝138；
Step (3), opening the third breaker switch S3, closing the fourth breaker switch S4, continuing to use the method of step (1) and step (2) to obtain the feature set of the dryer,containing n4-3Data for the group, where experiment was repeated 125 times, n4-3＝125；
And (4) closing any combination of the first breaker switch S1, the second breaker S2, the third breaker S3 or the fourth breaker S4 to obtain one switching combination of the three electrical appliances within a period of time. Through the load decomposition module, firstly, an active power catastrophe point and an electrical appliance switching action point are obtained, wherein the active power catastrophe point and the electrical appliance switching action point are as follows according to the input sequence: the power spectrum in 20 seconds is calculated according to a formula (2) by using an incandescent lamp, an LED lamp and a dryer, wherein the sampling frequency Fs is 6250Hz, and the total sampling points are 125000;
step (5), according to the flow of fig. 2, extracting features near each switching point, and setting the weights as:
Figure BDA0002991757300000051
Figure BDA0002991757300000052
for three electrical appliances, the weight sum of each sample is 1. By utilizing the improved KNN algorithm, the comprehensive similarity is respectively calculated with each load, for example, the comprehensive similarity of the load thrown at a certain time and an incandescent lamp is as follows:
sim4-1＝sim14-1×weight4-1+sim24-1
finally, judging the class of the three types of loads to which the switching electric appliance belongs each time;
in the experimental process, after the system shown in fig. 1 is built, parameter recording is performed. In example 1, data recorded in step (1), step (2) and step (3) is used for feature set establishment; a random switching process is generated in the step (4), namely, the switching combination of the three kinds of electric appliances is realized; and (5) judging the class of each load switching by using the improved KNN algorithm.

Claims (2)

1. A non-intrusive load identification method based on variable point detection and improved KNN algorithm is characterized by comprising the following specific steps:
(1) the non-invasive load identification method is realized by a non-invasive load identification system, the system consists of a power supply system, an electric signal acquisition system, a bus deconcentrator, a load decomposition module and various electric appliances to be detected, the power supply system is connected with the electric signal acquisition system through a first circuit breaker, the electric signal acquisition system is connected with the load decomposition module, the electric signal acquisition system is connected with the bus deconcentrator, the bus deconcentrator is respectively connected with the various electric appliances to be detected through different circuit breakers, and the power supply system is used for supplying power to the electric signal acquisition system; each circuit breaker is used for respectively controlling the switching-in and switching-out actions of the electric appliance; the electric signal acquisition system is used for acquiring electric signal data such as voltage, current and the like at the bus deconcentrator; the electrical appliance to be tested can select various different types of electrical appliances for feature set establishment and testing; the bus deconcentrator is used for merging different types of electric appliances to be tested into a bus, realizing the aliasing of loads and simulating an actual electric-using scene;
(2) establishing a characteristic set, putting a single electric appliance into a non-invasive load identification system, measuring voltage data u (t) and current data i (t) of the single electric appliance during stable operation by using the electric signal measuring system, carrying out Fourier transform on electric appliance data of a period, taking a fundamental wave voltage phase as 0 as a reference point, extracting a voltage instantaneous value u (t) in the period and current instantaneous value data i (t) in the period, and obtaining the voltage data u (t) in the periodT(t) voltage data i in the periodT(t) according to uT(t)，iT(t), making a V-I track, extracting higher harmonics, active power and reactive power, and taking the binarized V-I track, higher harmonics, active power and reactive power as the feature set of the electrical appliance; repeating the steps for different kinds of electric appliances to respectively obtain feature sets of different electric appliances;
(3) performing aliasing input test on a plurality of electric appliances, inputting or cutting the plurality of electric appliances within a period of time t, obtaining a voltage instantaneous sampling discrete value U (i) of the period within t seconds and a current instantaneous sampling discrete value I (i) of the period through an electric signal acquisition system, and calculating active power, wherein a formula (1) is a definition expression of the active power, and a formula (2) is an active power calculation method under discrete data;
Figure FDA0002991757290000011
wherein: p is the active power in the period, T is the period, u (T), i (T) are the voltage and current instantaneous values of the period respectively;
the active power calculation method adopted in the step is shown in formula (2):
Figure FDA0002991757290000012
p is the active power in the period, T is the period, U (i), I (i) are instantaneous sampling discrete values of the voltage and the current of the period, and Fs is the sampling frequency;
(4) the method comprises the steps of variable point detection and characteristic extraction, wherein the time point of active power P change is determined by adopting a bilateral CUSUM algorithm or a standard deviation multiple method based on active power in a time period t, namely the time point t of electric appliance inputm(ii) a The extraction time intervals are respectively [ tm，tm+1]And [ tm-1，tm]Current-voltage signal of one period in the period, and voltage of one period before the change point is UoffThe current in a period before the change point is IoffThe voltage in a period after the change point is UonThe current in a period after the change point is IonDefined as the abscissa of the V-I locus
Figure FDA0002991757290000021
Ordinate Im＝Ion-IoffTo U withm-ImMaking the V-I trajectory of the unknown load as a trajectory feature, and applying to the Um、ImPerforming Fourier transform, and extracting harmonic amplitude, reactive power and the like as amplitude characteristics;
(5) and (4) load identification, namely calculating comprehensive similarity of the track characteristics and the amplitude characteristics in the step (4) and the electrical characteristics in the known electrical appliance characteristic set respectively, adopting an improved KNN algorithm to calculate the comprehensive similarity which is to be detected and accords with various known load characteristic sets respectively, and taking the result of the maximum similarity as the prediction result of the unknown load.
2. The method of claim 1, wherein: the load identification method in the step (5) adopts a KNN algorithm for averagely distributing weights, belongs to an improved KNN algorithm, and comprises the following steps of:
(5.1) the sum of the weights of the training samples of each category is set to be 1, when the samples are averagely distributed, the weight distributed to each sample is only related to the number of the samples contained in the category to which the sample belongs, and the weight (T) is weightj) The calculation method is as follows:
weight(Tj)＝1/size(CTj) (3)
wherein, size (C)Tj) Represents TjThe number of training samples contained in the category;
(5.2) calculating the V-I track similarity and the amplitude similarity of the sample to be detected and all training samples, and respectively recording as Sim1 and Sim 2:
sim1＝1/(1+dist1) (4)
sim2＝1/(1+dist2) (5)
wherein dist1 and dist2 are distances of a distance and an amplitude of a V-I track between two samples respectively, and are Euclidean distances. (ii) a
(5.3) arranging the training samples according to the size of Sim1 in a descending order, and taking the first K training samples with the largest track similarity as the K nearest neighbor of the current test sample;
(5.4) calculating the current test sample and the K nearest neighbor TjThe combined similarity Sim of the trajectory feature and the amplitude feature of (j ═ 1, 2.., K):
sim(a，Tj)＝sim1(a，Tj)×weight(Tj)+sim2(a，Tj) (4)
and (5.5) counting the total comprehensive similarity of the sample to be detected and each class in the K nearest neighbors, and taking the class with the maximum total comprehensive similarity as a prediction result.
CN202110317472.XA 2021-03-25 2021-03-25 Non-invasive load identification method based on variable point detection and improved KNN algorithm Active CN112949563B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110317472.XA CN112949563B (en) 2021-03-25 2021-03-25 Non-invasive load identification method based on variable point detection and improved KNN algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110317472.XA CN112949563B (en) 2021-03-25 2021-03-25 Non-invasive load identification method based on variable point detection and improved KNN algorithm

Publications (2)

Publication Number Publication Date
CN112949563A true CN112949563A (en) 2021-06-11
CN112949563B CN112949563B (en) 2022-09-16

Family

ID=76226684

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110317472.XA Active CN112949563B (en) 2021-03-25 2021-03-25 Non-invasive load identification method based on variable point detection and improved KNN algorithm

Country Status (1)

Country Link
CN (1) CN112949563B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114252682A (en) * 2021-12-21 2022-03-29 上海中和软件有限公司 Large building abnormal load control method based on non-invasive load monitoring

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108257044A (en) * 2017-09-19 2018-07-06 济南大学 A kind of non-intrusion type load decomposition method based on steady-state current model
CN109387712A (en) * 2018-10-09 2019-02-26 厦门理工学院 Non-intrusion type cutting load testing and decomposition method based on state matrix decision tree
CN110504679A (en) * 2019-07-25 2019-11-26 深圳供电局有限公司 A kind of non-intrusion type load discrimination method based on KM matching algorithm
CN112505511A (en) * 2020-12-14 2021-03-16 天津求实智源科技有限公司 Non-invasive low-voltage fault arc detection and positioning method and system

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108257044A (en) * 2017-09-19 2018-07-06 济南大学 A kind of non-intrusion type load decomposition method based on steady-state current model
CN109387712A (en) * 2018-10-09 2019-02-26 厦门理工学院 Non-intrusion type cutting load testing and decomposition method based on state matrix decision tree
CN110504679A (en) * 2019-07-25 2019-11-26 深圳供电局有限公司 A kind of non-intrusion type load discrimination method based on KM matching algorithm
CN112505511A (en) * 2020-12-14 2021-03-16 天津求实智源科技有限公司 Non-invasive low-voltage fault arc detection and positioning method and system

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
MENGQI LU等: "A Hybrid Event Detection Approach for Non-Intrusive Load Monitoring", 《IEEE TRANSACTIONS ON SMART GRID》 *
PENG TAO等: "Multi-level non-intrusive load indentification based on k-NN", 《2019 IEEE 3RD CONFERENCE ON ENERGY INTERNET AND ENERGY SYSTEM INTEGRATION》 *
郑义林等: "改进最小二乘变点识别法在负荷分解的应用", 《计算机测量与控制》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114252682A (en) * 2021-12-21 2022-03-29 上海中和软件有限公司 Large building abnormal load control method based on non-invasive load monitoring
CN114252682B (en) * 2021-12-21 2024-04-05 上海中和软件有限公司 Large building abnormal load control method based on non-invasive load monitoring

Also Published As

Publication number Publication date
CN112949563B (en) 2022-09-16

Similar Documents

Publication Publication Date Title
Shaik et al. Power quality assessment and event detection in hybrid power system
CN105067966B (en) The low-voltage alternating-current fault arc detection method of feature based modal components energy spectrometer
CN108616120B (en) Non-invasive power load decomposition method based on RBF neural network
KR101642373B1 (en) Method and device for the non-intrusive determination of the electrical power comsumed by an installation, by analysing load transients
Najmeddine et al. State of art on load monitoring methods
Zhang et al. A real-time classification method of power quality disturbances
CN110514889A (en) A kind of method and system of non-intrusion type household electricity remained capacity
Sirojan et al. Embedded edge computing for real-time smart meter data analytics
CN106093565B (en) A kind of electricity subentry measurement method and device based on steady state characteristic Waveform Matching
Lee Electric load information system based on non-intrusive power monitoring
CN109470957B (en) Non-invasive electrical appliance identification method based on Frey's distance
CN112924790B (en) Fourier transform-based non-invasive load identification method adopting non-invasive load identification device
CN111722028A (en) Load identification method based on high-frequency data
CN109299134B (en) Method for identifying electric appliance under non-invasive monitoring system
CN110571806B (en) Feature extraction and identification method for load category of power distribution network
CN112949563B (en) Non-invasive load identification method based on variable point detection and improved KNN algorithm
CN114252682B (en) Large building abnormal load control method based on non-invasive load monitoring
CN114942344A (en) Non-invasive electrical appliance identification method, system, medium, equipment and terminal
CN113193654A (en) Event-driven non-intrusive power load monitoring method based on transient and steady state combination characteristics
CN109541346A (en) A kind of non-intrusion type electrical load under steady state condition identifies method for improving
CN110244150B (en) Non-invasive electrical appliance identification method based on root mean square and standard deviation
Chang et al. Energy spectrum-based wavelet transform for non-intrusive demand monitoring and load identification
Gao et al. High-impedance arc fault modeling for distribution networks based on dynamic geometry dimension
CN112730984B (en) Low-voltage distribution network phase identification method based on intelligent electric meter
Li et al. A training-free non-intrusive load monitoring approach for high-frequency measurements based on graph signal processing

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