CN111598145A - Non-invasive load monitoring method based on mixed probability label time-varying constraint distribution - Google Patents

Non-invasive load monitoring method based on mixed probability label time-varying constraint distribution Download PDF

Info

Publication number
CN111598145A
CN111598145A CN202010347984.6A CN202010347984A CN111598145A CN 111598145 A CN111598145 A CN 111598145A CN 202010347984 A CN202010347984 A CN 202010347984A CN 111598145 A CN111598145 A CN 111598145A
Authority
CN
China
Prior art keywords
time
load
state
distribution
equipment
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
CN202010347984.6A
Other languages
Chinese (zh)
Other versions
CN111598145B (en
Inventor
江思伟
谢振平
司修利
刘莉
蔡佳燕
袁宏亮
王珺
林栋�
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.)
Neovoltaic Energy Nantong Co ltd
Original Assignee
Neovoltaic Energy Nantong Co ltd
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 Neovoltaic Energy Nantong Co ltd filed Critical Neovoltaic Energy Nantong Co ltd
Priority to CN202010347984.6A priority Critical patent/CN111598145B/en
Publication of CN111598145A publication Critical patent/CN111598145A/en
Application granted granted Critical
Publication of CN111598145B publication Critical patent/CN111598145B/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
    • 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
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2132—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on discrimination criteria, e.g. discriminant analysis
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R19/00—Arrangements for measuring currents or voltages or for indicating presence or sign thereof
    • G01R19/0046—Arrangements for measuring currents or voltages or for indicating presence or sign thereof characterised by a specific application or detail not covered by any other subgroup of G01R19/00
    • G01R19/0053—Noise discrimination; Analog sampling; Measuring transients
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R21/00—Arrangements for measuring electric power or power factor
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R21/00—Arrangements for measuring electric power or power factor
    • G01R21/001—Measuring real or reactive component; Measuring apparent energy
    • G01R21/002—Measuring real component
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R21/00—Arrangements for measuring electric power or power factor
    • G01R21/001—Measuring real or reactive component; Measuring apparent energy
    • G01R21/003—Measuring reactive component
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Power Engineering (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Supply And Distribution Of Alternating Current (AREA)

Abstract

The invention discloses a non-invasive load monitoring method based on mixed probability label time-varying constraint distribution, which considers that mixed equipment signal characteristics can be modeled by using a mixed Gaussian model at a single time point and introduces an equipment working state label joint distribution mechanism on a time domain, provides a novel monitoring method which is more efficient than an HMM method and can carry out joint correlation identification analysis on power load time-space domain signal characteristics, reduces the influence of the characteristics of low-power-consumption load equipment covered by high-power-consumption load equipment by using a differential characteristic extraction method, carries out modeling by using the mixed Gaussian model at the single time point and introduces the equipment working state label joint distribution mechanism on the time domain to carry out joint correlation identification analysis on the power load time-space domain signal characteristics, thereby improving the identification accuracy of load equipment.

Description

Non-invasive load monitoring method based on mixed probability label time-varying constraint distribution
Technical Field
The invention belongs to the field of information processing, and particularly relates to a non-invasive load monitoring method based on mixed probability label time-varying constraint distribution.
Background
The non-intrusive load monitoring (NILM) technology considers that energy consumption data aggregated by all devices are obtained from single-point signal measurement, and reversely calculates the working conditions of each device in the system load according to the consumption characteristics of a single electrical device under the condition that other sensing devices are not installed, and becomes an important technical means for improving the quality of power supply service and ensuring the safety of the power supply service at present.
Existing NILM techniques are largely divided into two categories, event-based and non-event-based. The event-based monitoring algorithm comprises two steps of event detection and classification, such as accumulation and detection, robust Bayesian detection, generalized likelihood ratio detection and the like, wherein the event detection performance can greatly influence the subsequent classification effect. The NILM method, which is not based on event detection, does not require the step of event detection, but directly analyzes the composite power signal characteristics of all samples. Among them, Hidden Markov Models (HMM) and their improved models are a widely used effective method. The HMM-based method considers using signal data at longer sampling points, the computational complexity of the device mixing state will grow exponentially as the number of devices increases, and the real-time computation efficiency is not high.
The NILM technology has certain applications in different fields, such as power systems, water flow monitoring systems, ventilation monitoring systems, solar power generation, and the like. Recently, the NILM technology has been studied and improved at home and abroad.
The method also has some research reports on the NILM in China, considers that the total current signal is mixed and superposed when the load operates independently, converts the load decomposition problem into the blind source separation problem, performs whitening processing on the mixed current signal, constructs a de-mixing matrix, and decomposes the electricity consumption of residents. Differential feature extraction is also used for converting the differential feature extraction into single load feature identification, and the load equipment identification is realized through fuzzy clustering, but the identification performance of the method on the low-power-consumption load equipment is still not ideal.
Disclosure of Invention
The purpose of the invention is as follows: the invention provides a non-invasive load monitoring method based on mixed probability label time-varying constraint distribution, which considers that the signal characteristics of mixed equipment at a single time point can be modeled by using a mixed Gaussian model, and introduces an equipment working state label joint distribution mechanism on a time domain, provides a novel monitoring method which is simpler than an HMM method and can carry out joint correlation identification analysis on the signal characteristics of a power load time-space domain, takes the data of different load equipment under a non-invasive environment as a training sample, extracts a load characteristic sample through principal component analysis, divides the load equipment into two types according to a comprehensive evaluation value and a sample ranking graph, and combines a Fisher judgment criterion to realize the characteristic separation of the load equipment of different types; specifically, the influence of the covering characteristics of low-power-consumption load equipment by high-power-consumption load equipment is reduced by utilizing a differential characteristic extraction method, modeling is carried out on a single time point by utilizing a Gaussian mixture model, and a joint distribution mechanism of equipment working state labels on a time domain is introduced to carry out joint correlation identification analysis on the power load time-space domain signal characteristics, so that the identification accuracy of the load equipment is improved.
The technical scheme is as follows: the invention discloses a non-invasive load monitoring method based on mixed probability label time-varying constraint distribution, which specifically comprises the following steps:
(1) based on a Gaussian mixture probability model and a time-varying label optimal coupling distribution mechanism, the state of the power load equipment is identified through an iterative optimization distribution method, a mixed probability label time-varying constraint distribution strategy is introduced, the transient characteristic and the steady-state characteristic of the load equipment are extracted, and the model realizes the self-adaptive division modeling of continuous numerical data through a maximum self-interpretation reduction target and further analyzes the continuous numerical data.
(2) Performing difference characteristic extraction on total active power P, total reactive power Q, C-phase active power PC, C-phase reactive power QC and harmonic current IC;
(3) training data of all load equipment in independent operation in different states to obtain corresponding load equipment characteristics, namely a Gaussian model parameter mean value mu and a covariance sigma;
(4) judging the running state of the load equipment by a calculation formula by using a single-point Gaussian mixture model;
(5) at time T (T)>2) By calculating the cumulative error of the formula to accumulate the error ETMinimization of the error, if the accumulated error ETWhen the time is reduced, the step (4) is carried out again;
(6) when the error E is accumulatedTGradually converge to less than log2And T, finishing the calculation at the moment, finishing backtracking distribution, and identifying the working state of the corresponding load equipment.
Further, the transient characteristics of step (1) include instantaneous voltage, instantaneous current, instantaneous power and voltage noise; the steady state characteristics include active power, reactive power, current harmonics, voltage waveforms.
Further, the step (3) is specifically as follows: the Gaussian Mixture Model (GMM) has a probability distribution Model of the form,
Figure BDA0002470029180000031
wherein the content of the first and second substances,
Figure BDA0002470029180000032
Ν(x|μk,Σk) The kth component, π, of the Gaussian mixture modelkA weight for each component; a single probability density function of
Figure BDA0002470029180000033
In the formula, mukAs the mean value of the characteristics of the load equipment, sigmakIs the load device feature covariance, Ν (x | μk,Σk) The value of (d) is a probability density function of a single point gaussian distribution. To arbitrary load equipment ekUsing the power consumption data in different running states as training samples to obtain corresponding mean values mukSum covariance Σk。
Further, the step (4) is specifically as follows:set pikIs 1/k, then the probability of each load device is:
Figure BDA0002470029180000034
Figure BDA0002470029180000035
then, the operation state of all the load devices at any time t can be calculated as
πk＝y(i,k)/Nk(8)。
Further, when pik<At 1/k, the load equipment is considered not to be in operation, when pik>At 1/k, the load equipment may or may not be in operation;
thus, all the load devices { e } at any time t1,e2,e3LeN-2,eN-1,eNThe running state of St＝{st,1,st,2,st,3,Lst,N-2,st,N-1,st,N},st,i∈ {0,1}, i ═ 1,2, LN, when siWhen 1, the load device is operated in this state, when siWhen 0, the load device is not operated in this state; the multiple states of the same device are considered to be different devices, i.e. for any load device eiM states as a set of m load devices { e }i,ei+1Lei+m-2,ei+m-1When eiWhen the device is operated in a certain state, the corresponding device is in an operating state, and other devices are in a closed state; the operating state of only one electrical appliance between any two adjacent moments is changed, and the operating state at the moment t +1 is St+1＝{st+1,1,st+1,2,st+1,3,Lst+1,N-2,st+1,N-1,st+1,N},st+1,i∈ {0,1}, i ═ 1,2, LN, operating state S due to time ttOnly the possibility of N operating states, state st+1A state space diagram at time t + 1; form ofThe state space is the state in which all possible operating states of the load device between time t and time t +1 can occur.
Further, in the step (5), a single-point mixed gaussian model is used, and the probability vectors W of the load equipment at the time T are obtained according to the formulas (6), (7) and (8) calculated in the step (4)T＝{π1,π2…πkTherein of
Figure BDA0002470029180000041
The accumulated error of the state spaces of all the electric appliances before the backtracking time T is as follows:
Figure BDA0002470029180000042
at time T, according to St≤TAnd WTObtaining a temporarily selected running state, and distributing the running state at the moment of T +1 on the basis of the temporarily selected running state; if the running state at the time T cannot be distributed to the running state at the time T +1, the time T-1 is traced back forward.
Further, the step (6) is specifically as follows: will accumulate the error ETMinimization as model training target when accumulating error ETGradually converge and are less than log2And T, finishing backtracking distribution and identifying the working state of the corresponding load equipment.
Has the advantages that: the invention discloses a non-invasive load monitoring method based on mixed probability label time-varying constraint distribution, which utilizes a differential feature extraction method to reduce the influence of the covering features of low-power-consumption load equipment by high-power-consumption load equipment, utilizes a mixed Gaussian model to perform modeling at a single time point, introduces an equipment working state label joint distribution mechanism on a time domain, and performs joint correlation identification analysis on the power load time-space domain signal features, thereby improving the identification accuracy of the load equipment.
Drawings
FIG. 1 is a graph of a hybrid probabilistic tag time-varying constraint distribution strategy;
fig. 2 is a state space diagram of successive moments of the load device.
Detailed Description
When multiple load devices run simultaneously, the high-power load device can cover the characteristics of the low-power load device, and the identification degree of the load device is not high. Aiming at the problem, an optimal coupling distribution mechanism based on a Gaussian mixture probability model and a time-varying label is provided, and the state of the power load equipment is identified through an iterative optimization distribution method. A mixed probabilistic label time-varying constraint assignment strategy is introduced as shown in figure 1.
W in FIG. 1TProbability vectors of each load device obtained based on a Gaussian mixture probability model at time T, St≤TPossible operating states of all load units at time T, L (S)t≤T,WT)PP kinds of operation state spaces { L) of load equipment at time TT,0,LT,1,LT,2,…LT,P-1,LT,P}。WT+1Probability vectors of each load device obtained based on the Gaussian mixture probability model at the moment of T +1, St≤T+1Possible operating states of all load units before T +1, L (S)t≤T+1,WT+1)QQ kinds of operation state spaces { L) of load equipment at the moment of T +1T+1,0,LT+1,1,LT+1,2,…LT+1,Q-1,LT+1,Q}. At time T, according to St≤TAnd WTObtaining a temporarily selected running state LT,0Based on this, according to WT+1Allocating a preferred operating state LT+1,0When L is presentT,0If the reasonable distribution can not be carried out in the Q running states at the time T +1, the candidate running state space { L } is selected by backtracking to the time TT,1,LT,2,…LT,P-1,LT,PAnd obtaining the optimal coupling distribution. If the P running states can not be reasonably coupled and distributed at the time T +1 at the time T, the time T-1 is traced back forwards, and so on.
The non-invasive load power monitoring mainly comprises three parts, namely load data acquisition, load data processing and load monitoring and identification. There may be some specificity in the operational characteristics of the different load devices in the NILM system, so that these valid characteristics can be extracted from the raw power signal data. The operating characteristics of the electrical load devices can be largely classified into transient characteristics and steady-state characteristics. Transient characteristics include instantaneous voltage, instantaneous current, instantaneous power, voltage noise, etc., and steady-state characteristics include active power, reactive power, current harmonics, voltage waveforms, etc. Table 1 shows a common transient characteristic extraction method, and a common steady-state load characteristic method is shown in table 2.
Table 1 transient feature extraction method and advantages and disadvantages thereof
Figure BDA0002470029180000061
TABLE 2 Steady-State feature extraction method
Figure BDA0002470029180000062
Figure BDA0002470029180000071
The operating characteristics of the electrical load devices can be largely classified into transient characteristics and steady-state characteristics. The method specifically considers extracting total active power P, total reactive power Q, C-phase active power PC, C-phase reactive power QC and harmonic current IC.
The Gaussian Mixture Model (GMM) has a probability distribution Model of the form,
Figure BDA0002470029180000072
wherein the content of the first and second substances,
Figure BDA0002470029180000073
Ν(x|μk,Σk) The kth component, π, of the Gaussian mixture modelkIs the weight of each component.
A single probability density function of
Figure BDA0002470029180000074
In the formula, mukAs the mean value of the characteristics of the load equipment, sigmakIs the load device feature covariance, Ν (x | μk,Σk) The value of (d) is a probability density function of a single point gaussian distribution. To arbitrary load equipment ekUsing the power consumption data in different running states as training samples to obtain corresponding mean values mukSum covariance Σk. Set pikIs 1/k, then the probability of each load device is:
Figure BDA0002470029180000075
Figure BDA0002470029180000076
then, the operation state of all the load devices at any time t can be calculated as
πk＝y(i,k)/Nk(8)
When pik<At 1/k, the load equipment is considered not to be in operation, when pik>At 1/k, the load device may or may not be operating.
Thus, all the load devices { e } at any time t1,e2,e3LeN-2,eN-1,eNThe running state of St＝{st,1,st,2,st,3,Lst,N-2,st,N-1,st,N},st,i∈ {0,1}, i ═ 1,2, LN, when siWhen 1, the load device is operated in this state, when siWhen 0, the load device is not operated in this state. The multiple states of the same device are considered to be different devices, i.e. for any load device eiM states as a set of m load devices { e }i,ei+1Lei+m-2,ei+m-1When eiWhen the device is operated in a certain state, the corresponding device is in an operating state, and other devices are in a closed state. With only one current between any two adjacent momentsThe working state of the device is changed, and the running state at the moment of t +1 is St+1＝{st+1,1,st+1,2,st+1,3,Lst+1,N-2,st+1,N-1,st+1,N},st+1,i∈ {0,1}, i ═ 1,2, LN, operating state S due to time ttOnly the possibility of N operating states, state st+1The state space at time t +1 is shown in fig. 2, where the blue circles represent that the load devices are operating, the white circles represent that the load devices are not operating, and the state space is the state in which the operating states of the load devices between time t and time t +1 are all possible.
First, according to equations (6), (7) and (8), a probability vector W of the load device at time T is obtainedT＝{π1,π2…πk}，
Wherein
Figure BDA0002470029180000081
The accumulated error of the state spaces of all the electric appliances before the backtracking time T is as follows:
Figure BDA0002470029180000082
at time T, according to St≤TAnd WTA temporarily selected operating state is obtained, and the operating state at the time T +1 is allocated on the basis of the temporarily selected operating state. If the running state at the time T cannot be distributed to the running state at the time T +1, the time T-1 is traced back forward. Will accumulate the error ETMinimization as model training target when accumulating error ETGradually converge and are less than log2And T, finishing backtracking distribution and identifying the working state of the corresponding load equipment.
In summary, the non-intrusive load monitoring method based on the hybrid probability label time-varying constraint distribution is as follows:
(1) based on a Gaussian mixture probability model and a time-varying label optimal coupling distribution mechanism, the state of the power load equipment is identified through an iterative optimization distribution method, a mixed probability label time-varying constraint distribution strategy is introduced, the transient characteristic and the steady-state characteristic of the load equipment are extracted, and the model realizes the self-adaptive division modeling of continuous numerical data through a maximum self-interpretation reduction target and further analyzes the continuous numerical data.
(2) Performing difference characteristic extraction on total active power P, total reactive power Q, C-phase active power PC, C-phase reactive power QC and harmonic current IC;
(3) training data of all load equipment in independent operation in different states to obtain corresponding load equipment characteristics, namely a Gaussian model parameter mean value mu and a covariance sigma;
(4) judging the running state of the load equipment by a calculation formula by using a single-point Gaussian mixture model;
(5) at time T (T)>2) By calculating the cumulative error of the formula to accumulate the error ETMinimization of the error, if the accumulated error ETWhen the time is reduced, the step (4) is carried out again;
(6) when the error E is accumulatedTGradually converge to less than log2And T, finishing the calculation at the moment, finishing backtracking distribution, and identifying the working state of the corresponding load equipment.
Although the present invention has been described with reference to a preferred embodiment, it should be understood that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims (7)

1. The non-intrusive load monitoring method based on the mixed probability label time-varying constraint distribution is characterized in that: the method specifically comprises the following steps:
(1) identifying the state of the power load equipment by an iterative optimization distribution method based on a Gaussian mixed probability model and a time-varying label optimal coupling distribution mechanism, introducing a mixed probability label time-varying constraint distribution strategy, and extracting transient characteristics and steady-state characteristics of the load equipment;
(2) performing difference characteristic extraction on total active power P, total reactive power Q, C-phase active power PC, C-phase reactive power QC and harmonic current IC;
(3) training data of all load equipment in independent operation in different states to obtain corresponding load equipment characteristics, namely a Gaussian model parameter mean value mu and a covariance sigma;
(4) judging the running state of the load equipment by a calculation formula by using a single-point Gaussian mixture model;
(5) at time T (T)>2) By calculating the cumulative error of the formula to accumulate the error ETMinimization of the error, if the accumulated error ETWhen the time is reduced, the step (4) is carried out again;
(6) when the error E is accumulatedTGradually converge to less than log2And T, finishing the calculation at the moment, finishing backtracking distribution, and identifying the working state of the corresponding load equipment.
2. The non-intrusive load monitoring method based on hybrid probabilistic label time-varying constraint distribution as recited in claim 1, wherein: the transient characteristics of step (1) include instantaneous voltage, instantaneous current, instantaneous power and voltage noise; the steady state characteristics include active power, reactive power, current harmonics, voltage waveforms.
3. The non-intrusive load monitoring method based on hybrid probabilistic label time-varying constraint distribution as recited in claim 1, wherein: the step (3) is specifically as follows: the Gaussian Mixture Model (GMM) has a probability distribution Model of the form,
Figure FDA0002470029170000011
wherein the content of the first and second substances,
Figure FDA0002470029170000021
Ν(x|μk,Σk) The kth component, π, of the Gaussian mixture modelkA weight for each component; a single probability density function of
Figure FDA0002470029170000022
In the formula, mukAs the mean value of the characteristics of the load equipment, sigmakIs the load device feature covariance, Ν (x | μk,Σk) The value of (a) is a probability density function of single-point Gaussian distribution; to arbitrary load equipment ekUsing the power consumption data in different running states as training samples to obtain corresponding mean values mukSum covariance Σk。
4. The non-intrusive load monitoring method based on hybrid probabilistic label time-varying constraint distribution as recited in claim 1, wherein: the step (4) is specifically as follows: set pikIs 1/k, then the probability of each load device is:
Figure FDA0002470029170000023
Figure FDA0002470029170000024
then, the operation state of all the load devices at any time t can be calculated as
πk＝y(i,k)/Nk(8)。
5. The non-intrusive load monitoring method based on hybrid probabilistic label time-varying constraint distribution as recited in claim 4, wherein: when pik<At 1/k, the load equipment is considered not to be in operation, when pik>At 1/k, the load equipment may or may not be in operation;
thus, all the load devices { e } at any time t1,e2,e3LeN-2,eN-1,eNThe running state of St＝{st,1,st,2,st,3,Lst,N-2,st,N-1,st,N},st,i∈ {0,1}, i ═ 1,2, LN, when siWhen 1, the load device is operated in this state, when siWhen 0, the load device is not operated in this state; the multiple states of the same device are considered to be different devices, i.e. for any load device eiM states as a set of m load devices { e }i,ei+1Lei+m-2,ei+m-1When eiWhen the device is operated in a certain state, the corresponding device is in an operating state, and other devices are in a closed state; the operating state of only one electrical appliance between any two adjacent moments is changed, and the operating state at the moment t +1 is St+1＝{st+1,1,st+1,2,st+1,3,Lst+1,N-2,st+1,N-1,st+1,N},st+1,i∈ {0,1}, i ═ 1,2, LN, operating state S due to time ttOnly the possibility of N operating states, state st+1A state space diagram at time t + 1; the state space is the state in which all possible operating states of the load device between time t and time t +1 can occur.
6. The non-intrusive load monitoring method based on hybrid probabilistic label time-varying constraint distribution as recited in claim 1, wherein: and (5) calculating formulas (6), (7) and (8) in the step (4) by using a single-point mixed Gaussian model to obtain a probability vector W of the load equipment at the moment TT＝{π1,π2…πkTherein of
Figure FDA0002470029170000031
The accumulated error of the state spaces of all the electric appliances before the backtracking time T is as follows:
Figure FDA0002470029170000032
at time T, according to St≤TAnd WTObtaining a temporarily selected running state, and distributing the running state at the moment of T +1 on the basis of the temporarily selected running state; if the running state at the time T cannot be distributed to the running state at the time T +1, the time T-1 is traced back forward.
7. The non-intrusive load monitoring method based on hybrid probabilistic label time-varying constraint distribution as recited in claim 1, wherein: the step (6) is specifically as follows: will accumulate the error ETMinimization as model training target when accumulating error ETGradually converge and are less than log2And T, finishing backtracking distribution and identifying the working state of the corresponding load equipment.
CN202010347984.6A 2020-04-27 2020-04-27 Non-invasive load monitoring method based on mixed probability label time-varying constraint distribution Active CN111598145B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010347984.6A CN111598145B (en) 2020-04-27 2020-04-27 Non-invasive load monitoring method based on mixed probability label time-varying constraint distribution

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010347984.6A CN111598145B (en) 2020-04-27 2020-04-27 Non-invasive load monitoring method based on mixed probability label time-varying constraint distribution

Publications (2)

Publication Number Publication Date
CN111598145A true CN111598145A (en) 2020-08-28
CN111598145B CN111598145B (en) 2022-06-14

Family

ID=72189462

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010347984.6A Active CN111598145B (en) 2020-04-27 2020-04-27 Non-invasive load monitoring method based on mixed probability label time-varying constraint distribution

Country Status (1)

Country Link
CN (1) CN111598145B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115982576A (en) * 2023-03-17 2023-04-18 石家庄科林电气股份有限公司 Malignant load identification method and device and electric energy meter

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110144819A1 (en) * 2009-12-16 2011-06-16 Robert Bosch Gmbh Method for non-intrusive load monitoring using a hybrid systems state estimation approach
CN109461095A (en) * 2018-10-18 2019-03-12 深圳供电局有限公司 Data processing method and equipment for load event detection for non-invasive load identification
CN109541336A (en) * 2018-10-10 2019-03-29 华南理工大学 A kind of multidimensional signal detection method of non-intrusion type load monitoring
CN110445126A (en) * 2019-06-25 2019-11-12 中国电力科学研究院有限公司 A kind of non-intrusion type load decomposition method and system
CN110504679A (en) * 2019-07-25 2019-11-26 深圳供电局有限公司 KM matching algorithm-based non-invasive load identification method
CN110580502A (en) * 2019-08-20 2019-12-17 上海纺织节能环保中心 Factor hidden Markov load decomposition method based on Gaussian mixture
CN110954744A (en) * 2019-11-18 2020-04-03 浙江工业大学 Non-invasive load monitoring method based on event detection

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110144819A1 (en) * 2009-12-16 2011-06-16 Robert Bosch Gmbh Method for non-intrusive load monitoring using a hybrid systems state estimation approach
CN109541336A (en) * 2018-10-10 2019-03-29 华南理工大学 A kind of multidimensional signal detection method of non-intrusion type load monitoring
CN109461095A (en) * 2018-10-18 2019-03-12 深圳供电局有限公司 Data processing method and equipment for load event detection for non-invasive load identification
CN110445126A (en) * 2019-06-25 2019-11-12 中国电力科学研究院有限公司 A kind of non-intrusion type load decomposition method and system
CN110504679A (en) * 2019-07-25 2019-11-26 深圳供电局有限公司 KM matching algorithm-based non-invasive load identification method
CN110580502A (en) * 2019-08-20 2019-12-17 上海纺织节能环保中心 Factor hidden Markov load decomposition method based on Gaussian mixture
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 (1)

* Cited by examiner, † Cited by third party
Title
张露: "非侵入式负荷监测算法研究", 《中国知网硕士电子期刊》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115982576A (en) * 2023-03-17 2023-04-18 石家庄科林电气股份有限公司 Malignant load identification method and device and electric energy meter
CN115982576B (en) * 2023-03-17 2023-06-20 石家庄科林电气股份有限公司 Malignant load identification method and device and electric energy meter

Also Published As

Publication number Publication date
CN111598145B (en) 2022-06-14

Similar Documents

Publication Publication Date Title
CN111428816B (en) Non-invasive load decomposition method
CN110956220B (en) Non-invasive household appliance load identification method
CN109376801B (en) Wind driven generator blade icing diagnosis method based on integrated deep neural network
CN111428355B (en) Modeling method for intelligent comprehensive statistics of power load numbers
CN109345409A (en) A kind of resident&#39;s comprehensive energy efficiency management method based on bandwidth carrier
CN105823948A (en) Non-invasive resident load identification method
CN111563827B (en) Load decomposition method based on physical characteristics of electrical appliance and electricity utilization behaviors of residents
CN106600074A (en) DFHSMM-based non-intrusion type electric power load monitoring method and system
CN109190950B (en) Energy consumption abnormity identification method for communicating electric meter and water meter files
CN109902915A (en) A kind of energy behavior analysis method of the electricity-water-gas based on fuzzy C-mean algorithm model
CN114676742A (en) Power grid abnormal electricity utilization detection method based on attention mechanism and residual error network
CN113902104A (en) Non-invasive load monitoring method combining unsupervised domain self-adaptive strategy and attention mechanism
CN106469421A (en) A kind of wind-powered electricity generation/photovoltaic generation power collects appraisal procedure and the system of topology
Hu et al. Hierarchical fault diagnosis for power systems based on equivalent-input-disturbance approach
CN110380444A (en) Distributing wind-powered electricity generation orderly accesses the method for planning capacity of power grid under a kind of more scenes based on structure changes Copula
CN113887912A (en) Non-invasive load identification method for deeply learning downward embedded equipment
CN111598145B (en) Non-invasive load monitoring method based on mixed probability label time-varying constraint distribution
Kim et al. A study for clustering method to generate Typical Load Profiles for Smart Grid
CN111860684A (en) Power plant equipment fault early warning method and system based on dual networks
Gurbuz et al. Comprehensive non-intrusive load monitoring process: Device event detection, device feature extraction and device identification using KNN, random forest and decision tree
CN115392640A (en) Maintenance-oriented motor generator full life cycle cost efficiency evaluation system
CN116861316B (en) Electrical appliance monitoring method and device
CN112307675B (en) Neural network-based temperature-sensitive load separation identification method and system
CN110163437B (en) Day-ahead photovoltaic power generation power prediction method based on DPK-means
CN112070625A (en) User electricity consumption behavior pattern recognition method and system

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
CB02 Change of applicant information
CB02 Change of applicant information

Address after: 226100 No.888, Jiuhua Road, high tech Zone, Nantong City, Jiangsu Province

Applicant after: Wotai Energy Co.,Ltd.

Address before: 226000 No.888, Jiuhua Road, Tongzhou District, Nantong City, Jiangsu Province

Applicant before: NEOVOLTAIC ENERGY NANTONG Co.,Ltd.

GR01 Patent grant
GR01 Patent grant