CN112560977A - Non-invasive load decomposition method based on sparse classifier hierarchical algorithm - Google Patents
Non-invasive load decomposition method based on sparse classifier hierarchical algorithm Download PDFInfo
- Publication number
- CN112560977A CN112560977A CN202011538200.4A CN202011538200A CN112560977A CN 112560977 A CN112560977 A CN 112560977A CN 202011538200 A CN202011538200 A CN 202011538200A CN 112560977 A CN112560977 A CN 112560977A
- Authority
- CN
- China
- Prior art keywords
- electric appliance
- active power
- reactive power
- frequency
- power
- 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
Links
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 32
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 26
- 239000011159 matrix material Substances 0.000 claims abstract description 33
- 238000004458 analytical method Methods 0.000 claims abstract description 7
- 238000007621 cluster analysis Methods 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 description 5
- 238000011156 evaluation Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 240000007594 Oryza sativa Species 0.000 description 1
- 235000007164 Oryza sativa Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 235000009566 rice Nutrition 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/28—Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
-
- 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)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a non-invasive load decomposition method based on a sparse classifier hierarchical algorithm, which comprises the following steps of: s1, acquiring voltage and current data of the electric appliance; s2, calculating low-frequency active power, low-frequency reactive power and high-frequency current harmonic waves of the electric appliance; s3, clustering analysis; s4, constructing a feature dictionary; s5, constructing a sparse matrix; s6, power identification; s7, judging the result, and if the coincidence number in the judgment result is equal to 1, outputting the judgment result; otherwise, executing step S8; and S8, identifying current harmonics and outputting an identification result. By adopting the non-invasive load decomposition method based on the sparse classifier hierarchical algorithm, whether current harmonic identification is introduced or not can be determined by comprehensively considering the active power identification result and the reactive power identification result of the electric appliances when partial household electric appliances with similar power exist in a user, so that the electric appliance identification result is ensured without long-time and large-calculation-amount electric appliance identification when the power is similar.
Description
Technical Field
The invention relates to a smart grid technology, in particular to a non-intrusive load decomposition method based on a sparse classifier hierarchical algorithm.
Background
The key technology of intelligent power utilization is mainly embodied in Advanced Metering Infrastructure (AMI), system and terminal technology, intelligent power utilization bidirectional interactive operation mode and support technology, and the mutual influence of a user power utilization environment and a power utilization mode.
With the comprehensive popularization of AMI technology, the acquisition of fine-grained aggregated power of users becomes possible, which provides a chance for deep mining of user load information. In the construction of a smart power grid, the perception and the acquisition of the Load information of the equipment in a user are key links in information flow exchange and information deep mining, and Non-Intrusive Load Monitoring (NILM) provides a method for acquiring the energy consumption data of each equipment in a family only by collecting data such as aggregated power, current and the like.
The non-intrusive load decomposition includes low frequency non-intrusive load decomposition and high frequency load decomposition. In the low-frequency non-invasive load decomposition, a large amount of prior knowledge is usually required for identification, and the low-frequency non-invasive load decomposition is difficult to realize in practical application. And the load decomposition calculation amount of high frequency is large, and the identification time is long. And aiming at the problem that the user side has electrical appliances with similar power, how to make the identification result reach the best under the condition of ensuring the decomposition time becomes the problem which needs to be solved urgently.
Disclosure of Invention
The invention aims to provide a non-invasive load decomposition method based on a sparse classifier hierarchical algorithm, which can comprehensively consider the active power identification and reactive power identification results of electric appliances to determine whether to introduce current harmonic identification when partial household electric appliances with similar power exist in a user, thereby ensuring that the electric appliance identification result is ensured without long-time and large-computation-amount electric appliance identification when the power is similar.
In order to achieve the above object, the present invention provides a non-invasive load decomposition method based on a sparse classifier hierarchical algorithm, comprising the following steps:
s1, acquiring voltage and current data of the electric appliance;
s2, calculating low-frequency active power, low-frequency reactive power and high-frequency current harmonic waves of the electric appliance according to the collected voltage and current data;
s3, performing cluster analysis on the low-frequency active power, the low-frequency reactive power and the high-frequency current harmonic of the electric appliance according to a Mean-shift algorithm;
s4, constructing an electric appliance low-frequency active power feature dictionary, a low-frequency reactive power feature dictionary and a high-frequency current harmonic feature dictionary according to the clustering analysis result;
s5, constructing a sparse matrix according to the clustering analysis result and the characteristic that different states of the same electric appliance cannot be used simultaneously;
s6, performing power identification according to the sparse matrix, the low-frequency active power feature dictionary and the low-frequency reactive power feature dictionary;
s7, judging results according to the active power identification results and the reactive power identification results, and if the coincidence number in the judgment results is equal to 1, outputting the judgment results; otherwise, executing step S8;
and S8, performing current harmonic recognition according to the sparse matrix and the high-frequency current harmonic feature dictionary, and outputting a recognition result.
Preferably, step S2 specifically includes the following steps:
s20, calculating the active power P of the electric appliance according to the voltage and current data;
s21, calculating reactive power Q of the electric appliance according to the voltage and current data;
s22, calculating the specific formula of the current harmonic X of the electric appliance according to the voltage and current data, wherein the specific formula is as follows:
in the formula, C0Is a direct current component, Cmsin (m ω t + θ) denotes the m harmonic, CmIs the amplitude of the m-th harmonic, and θ is the initial phase of the m-th harmonic, with an angular frequency of ω.
Preferably, step S4 specifically includes the following steps:
s40, establishing a low-frequency active power feature dictionary, a low-frequency reactive power feature dictionary and a high-frequency current harmonic feature dictionary of the electric appliance i, wherein the specific formulas are as follows:
Pi＝[p1,p2,…,pL]
Qi＝[q1,q2,…,qL]
Xi＝[x1,x2,…,xL]
in the formula, p1，p2,…pLThe active power dictionary values of all the states of the electrical appliance i are obtained; q. q.s1,q2,…,qLDictionary values of reactive power for all states of the electrical appliance i; x is the number of1,x2,…,xLCurrent harmonic dictionary values for all states of the electrical appliance i; wherein p isL,qLPower value, x, representing the state of the applianceLRepresenting the first 15 current harmonics of the state of the electric appliance;
s41, establishing a low-frequency active power feature dictionary, a low-frequency reactive power feature dictionary and a high-frequency current harmonic feature dictionary of all electrical equipment, wherein the specific formulas are as follows:
PD＝[P1,P2,P3,…,PN]
QD＝[Q1,Q2,Q3,…,QN]
XD＝[X1,X2,X3,…,XN]
in the formula, PDDictionary of active power, P, representing all electrical appliances1，P2,…PNThe active power dictionary values of all the states of the N electric appliances are obtained; qDDictionary of reactive powers, Q, representing all electrical appliances1，Q2,…QNDictionary values of reactive power for all states of the N electrical appliances; xDDictionary of high-frequency current harmonics, X, representing all electrical appliances1，X2,…XNCurrent harmonic dictionary values for all states of the N appliances.
Preferably, the sparse matrix formula constructed in step S5 is:
in the formula, each column of the sparse matrix A represents a combination of appliances, and at most four 1 s exist in each column, which means that at most four appliances are simultaneously turned on.
Preferably, step S6 specifically includes the following steps:
s60 active power identification
S600, calculating the minimum Euclidean distance between the input active power and the input active power according to the sparse matrix and the low-frequency active power feature dictionary;
s601, obtaining an active power threshold value according to the set penalty coefficient, and calculating all electric appliance combinations smaller than or equal to the active power threshold value;
s61 reactive power identification
S610, calculating the minimum Euclidean distance from the input reactive power according to the sparse matrix and the low-frequency reactive power feature dictionary;
s611, obtaining a reactive power threshold value according to the set penalty coefficient, and calculating all electric appliance combinations smaller than or equal to the reactive power threshold value.
Preferably, step S6 specifically includes the following steps:
s60 active power identification
S600, the specific formula of the minimum Euclidean distance is as follows:
dp＝min|P-PD×A|
wherein P represents input active power, PDX A represents the active power value of all the electrical appliances in combination, dpThe Euclidean distance representing the minimum active power of the input active power and the active power of all electric appliance combinations;
s601, calculating all electric appliance combinations with active power thresholds less than or equal to the active power thresholds, wherein the specific formula is as follows:
Ap←|P-PD×A|≤dp+ε
wherein P represents input active power, PDX A represents the active power value of all the electrical appliances in combination, dp+ ε represents the active power threshold, ε represents the penalty factor, APRepresenting active power and input active power of combinations of electrical appliances in a sparse matrix AAll electric appliance combinations with Euclidean distance less than or equal to the active power threshold;
s61 reactive power identification
S610, the specific formula of the minimum Euclidean distance is as follows:
dq＝min|Q-QD×A|
wherein Q represents the reactive power of the input, QDX A represents the reactive power value of all the electrical appliances in combination, dqThe Euclidean distance representing the minimum reactive power of the input reactive power and the reactive power of all electric appliance combinations;
s611, calculating all electric appliance combinations smaller than or equal to the reactive power threshold value, wherein the specific formula is as follows:
Aq←|Q-QD×A|≤dq+ε
wherein Q represents the reactive power of the input, QDX A represents the reactive power value of all the electrical appliances in combination, dq+ ε represents the reactive power threshold, ε represents the penalty factor, AqAnd all the electric appliance combinations with Euclidean distances between the reactive power of the electric appliance combinations in the sparse matrix A and the input reactive power smaller than or equal to the reactive power threshold value.
Preferably, step S7 specifically includes the following steps: judging the number of the electric appliance combinations in the intersection of all the electric appliance combinations with the active power threshold value and all the electric appliance combinations with the reactive power threshold value, and if the electric appliance combinations are equal to 1, directly outputting a power identification result; otherwise, step S8 is executed.
Preferably, the specific formula of the number of electrical appliances in step S7 is as follows:
n，Apq＝f(Ap，Aq)
wherein f () represents taking Ap、AQNumber of combinations of coincidences, n representing number of coincidences, ApqRepresenting the corresponding electric appliance combination, if n is equal to 1, indicating that the power identification result is good, and directly outputting a judgment result if only one electric appliance combination in the database accords with the power characteristic of an identification object; if n is greater than 1, execution proceeds to step S8.
Preferably, step S8 specifically includes the following steps:
performing current harmonic recognition according to the sparse matrix and the high-frequency current harmonic feature dictionary, wherein the specific formula is as follows:
dX＝min|X-XDApq|
wherein X represents the harmonic of the input current, XD×ApqRepresenting the combined harmonic wave of the electric appliance meeting the identification of active power and reactive power, total 15 harmonic waves, dxRepresenting the minimum Euclidean distance between the input current harmonic and the current harmonic of all the electric appliance combinations;
finding the corresponding electric appliance combination according to the minimum Euclidean distance of the current harmonic waves:
Ax←|X-XD×Apq|＝dx
in the formula AxIs represented by ApqThe combination of electrical appliances in the set corresponding to the minimum Euclidean distance of the input current harmonics, AxAnd outputting the identification result for the final result of the electrical appliance identification.
Therefore, the beneficial effects of the invention are as follows:
1) the method comprises the steps of firstly obtaining voltage and current data of the electric appliance, secondly calculating low-frequency active power, low-frequency reactive power and high-frequency current harmonic data of the electric appliance, then carrying out cluster analysis on the low-frequency active power, the low-frequency reactive power and the high-frequency current harmonic data of the electric appliance according to a Mean-shift algorithm so as to construct a low-frequency active power characteristic dictionary, a low-frequency reactive power characteristic dictionary, a high-frequency current harmonic characteristic dictionary and a sparse matrix of the electric appliance, and finally determining the electric appliance combination corresponding to each electric appliance data by adopting a hierarchical load identification method according to the sparse matrix, the low-frequency active power characteristic dictionary, the low-frequency reactive power characteristic dictionary and the high-frequency current harmonic characteristic dictionary, judging whether current harmonic identification needs to be introduced or not according to the reactive power identification result of the electric appliance under the condition that the power of the electric, the method has the advantages that long-time and large-calculation-amount electric appliance identification is not needed, the identification result of the electric appliance is guaranteed, and the problems that when low-frequency data is used for load decomposition, a large amount of priori knowledge is needed to improve the decomposition precision, and the decomposition precision is difficult to achieve in practical application, and when high-frequency data is used for load decomposition, the identification time is too long, and the calculation amount is too large are solved.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
FIG. 1 is a flowchart of a non-intrusive load decomposition method based on a sparse classifier pyramid algorithm according to an embodiment of the present invention;
fig. 2 is a detailed flowchart of step S6 of a non-intrusive load decomposition method based on a sparse classifier pyramid algorithm according to an embodiment of the present invention.
Detailed Description
The present invention will be further described with reference to the accompanying drawings, and it should be noted that the present embodiment is based on the technical solution, and the detailed implementation and the specific operation process are provided, but the protection scope of the present invention is not limited to the present embodiment.
Fig. 1 is a flowchart of a non-invasive load decomposition method based on a sparse classifier hierarchical algorithm according to an embodiment of the present invention, and fig. 2 is a detailed flowchart of step S6 of the non-invasive load decomposition method based on the sparse classifier hierarchical algorithm according to the embodiment of the present invention, as shown in fig. 1 and fig. 2, the present invention includes the following steps:
s1, acquiring voltage and current data of the electric appliance;
s2, calculating low-frequency active power, low-frequency reactive power and high-frequency current harmonic waves of the electric appliance according to the collected voltage and current data;
s3, performing cluster analysis on the low-frequency active power, the low-frequency reactive power and the high-frequency current harmonic of the electric appliance according to a Mean-shift algorithm;
s4, constructing an electric appliance low-frequency active power feature dictionary, a low-frequency reactive power feature dictionary and a high-frequency current harmonic feature dictionary according to the clustering analysis result;
s5, constructing a sparse matrix according to the clustering analysis result and the characteristic that different states of the same electric appliance cannot be used simultaneously;
s6, performing power identification according to the sparse matrix, the low-frequency active power feature dictionary and the low-frequency reactive power feature dictionary;
s7, judging results according to the active power identification results and the reactive power identification results, and if the coincidence number in the judgment results is equal to 1, outputting the judgment results; otherwise, executing step S8;
and S8, performing current harmonic recognition according to the sparse matrix and the high-frequency current harmonic feature dictionary, and outputting a recognition result.
Wherein, step S2 specifically includes the following steps:
s20, calculating the active power P of the electric appliance according to the voltage and current data;
s21, calculating reactive power Q of the electric appliance according to the voltage and current data;
s22, calculating the specific formula of the current harmonic X of the electric appliance according to the voltage and current data, wherein the specific formula is as follows:
in the formula, C0Is a direct current component, Cmsin (m ω t + θ) denotes the m harmonic, CmIs the amplitude of the m-th harmonic, and θ is the initial phase of the m-th harmonic, with an angular frequency of ω.
Step S4 specifically includes the following steps:
s40, establishing a low-frequency active power feature dictionary, a low-frequency reactive power feature dictionary and a high-frequency current harmonic feature dictionary of the electric appliance i, wherein the specific formulas are as follows:
Pi＝[p1,p2,…,pL]
Qi＝[q1,q2,…,qL]
Xi＝[x1,x2,…,xL]
in the formula, p1，p2,…pLThe active power dictionary values of all the states of the electrical appliance i are obtained; q. q.s1,q2,…,qLDictionary values of reactive power for all states of the electrical appliance i; x is the number of1,x2,…,xLFor current harmonic words of all states of the appliance iA typical value; wherein p isL,qLPower value, x, representing the state of the applianceLRepresenting the first 15 current harmonics of the state of the electric appliance;
s41, establishing a low-frequency active power feature dictionary, a low-frequency reactive power feature dictionary and a high-frequency current harmonic feature dictionary of all electrical equipment, wherein the specific formulas are as follows:
PD＝[P1,P2,P3,…,PN]
QD＝[Q1,Q2,Q3,…,QN]
XD＝[X1,X2,X3,…,XN]
in the formula, PDDictionary of active power, P, representing all electrical appliances1，P2,…PNThe active power dictionary values of all the states of the N electric appliances are obtained; qDDictionary of reactive powers, Q, representing all electrical appliances1，Q2,…QNDictionary values of reactive power for all states of the N electrical appliances; xDDictionary of high-frequency current harmonics, X, representing all electrical appliances1，X2,…XNCurrent harmonic dictionary values for all states of the N appliances.
The sparse matrix formula constructed in step S5 is:
in the formula, each column of the sparse matrix A represents a combination of appliances, and at most four 1 s exist in each column, which means that at most four appliances are simultaneously turned on.
Step S6 specifically includes the following steps:
s60 active power identification
S600, calculating the minimum Euclidean distance between the input active power and the input active power according to the sparse matrix and the low-frequency active power feature dictionary;
s601, obtaining an active power threshold value according to the set penalty coefficient, and calculating all electric appliance combinations smaller than or equal to the active power threshold value;
s61 reactive power identification
S610, calculating the minimum Euclidean distance from the input reactive power according to the sparse matrix and the low-frequency reactive power feature dictionary;
s611, obtaining a reactive power threshold value according to the set penalty coefficient, and calculating all electric appliance combinations smaller than or equal to the reactive power threshold value.
More specifically, step S6 specifically includes the following steps:
s60 active power identification
S600, the specific formula of the minimum Euclidean distance is as follows:
dp＝min|P-PD×A|
wherein P represents input active power, PDX A represents the active power value of all the electrical appliances in combination, dpThe Euclidean distance representing the minimum active power of the input active power and the active power of all electric appliance combinations;
s601, calculating all electric appliance combinations with active power thresholds less than or equal to the active power thresholds, wherein the specific formula is as follows:
Ap←|P-PD×A|≤dp+ε
wherein P represents input active power, PDX A represents the active power value of all the electrical appliances in combination, dp+ ε represents the active power threshold, ε represents the penalty factor, APRepresenting all the electric appliance combinations with Euclidean distances between the active power of the electric appliance combinations in the sparse matrix A and the input active power smaller than or equal to an active power threshold value;
s61 reactive power identification
S610, the specific formula of the minimum Euclidean distance is as follows:
dq＝min|Q-QD×A|
wherein Q represents the reactive power of the input, QDX A represents the reactive power value of all the electrical appliances in combination, dqThe Euclidean distance representing the minimum reactive power of the input reactive power and the reactive power of all electric appliance combinations;
s611, calculating all electric appliance combinations smaller than or equal to the reactive power threshold value, wherein the specific formula is as follows:
Aq←|Q-QD×A|≤dq+ε
wherein Q represents the reactive power of the input, QDX A represents the reactive power value of all the electrical appliances in combination, dq+ ε represents the reactive power threshold, ε represents the penalty factor, AqAnd all the electric appliance combinations with Euclidean distances between the reactive power of the electric appliance combinations in the sparse matrix A and the input reactive power smaller than or equal to the reactive power threshold value.
Step S7 specifically includes the following steps: judging the number of the electric appliance combinations in the intersection of all the electric appliance combinations with the active power threshold value and all the electric appliance combinations with the reactive power threshold value, and if the electric appliance combinations are equal to 1, directly outputting a power identification result; otherwise, step S8 is executed.
More specifically, the specific formula of the number of electrical appliance combinations in step S7 is as follows:
n,Apq＝f(Ap,Aq)
wherein f () represents taking Ap、AQNumber of combinations of coincidences, n representing number of coincidences, ApqRepresenting the corresponding electric appliance combination, if n is equal to 1, indicating that the power identification result is good, and directly outputting a judgment result if only one electric appliance combination in the database accords with the power characteristic of an identification object; if n is greater than 1, execution proceeds to step S8.
Step S8 specifically includes the following steps:
performing current harmonic recognition according to the sparse matrix and the high-frequency current harmonic feature dictionary, wherein the specific formula is as follows:
dX＝min|X-XDApq|
wherein X represents the harmonic of the input current, XD×ApqRepresenting the combined harmonic wave of the electric appliance meeting the identification of active power and reactive power, total 15 harmonic waves, dxRepresenting the minimum Euclidean distance between the input current harmonic and the current harmonic of all the electric appliance combinations;
finding the corresponding electric appliance combination according to the minimum Euclidean distance of the current harmonic waves:
Ax←|X-XD×Apq|＝dx
in the formula AxIs represented by ApqThe combination of electrical appliances in the set corresponding to the minimum Euclidean distance of the input current harmonics, AxAnd outputting the identification result for the final result of the electrical appliance identification.
Example one
In this embodiment, a total of 6 kinds of household appliances are provided, and the effectiveness of the method is tested by sampling voltage and current data of a user electric meter by using a USB type a/D data acquisition card, wherein the sampling frequency is 6.4 kHZ. In the data acquisition process, 6 electric equipment are ordered: the electric hair drier, the electric rice cooker, the microwave oven, the television, the electromagnetic oven and the dust collector are started and stopped randomly for 200 times, and the monitoring time is 24 hours. The random start-stop event represents the number of times the running state of the electrical appliance changes. The dictionary values obtained after the load feature template construction process are shown in table 1:
TABLE 1 electric appliance power dictionary value table
The evaluation indexes used in table 1 include: the Accuracy (Accuracy), Precision (Precision), Recall (Recall) and F1 scores are respectively calculated as follows:
in the formula, TP represents the total number of times the load is open in real data and also open in prediction; FP represents the total number of times the load is turned off in real data and on in predictions; TN represents the total number of times the load is turned off in real data and also turned off in anticipation; FN represents the total number of times the load is on in real data and off in anticipation. These metrics show how accurately an algorithm can predict the on/off state of a device. Finally, the identification result of each electric appliance is obtained, and the evaluation indexes are shown in the following table 2:
TABLE 2 evaluation index comparison Table
Therefore, by adopting the non-invasive load decomposition method based on the hierarchical algorithm of the sparse classifier, whether current harmonic identification is introduced or not can be determined by comprehensively considering the active power identification result and the reactive power identification result of the electric appliances when partial household electric appliances with similar power exist in a user, so that the electric appliance identification result is ensured without long-time and large-calculation-amount electric appliance identification when the power is similar.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting the same, and although the present invention is described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that: modifications and equivalents may be made to the invention without departing from the spirit and scope of the invention.
Claims (9)
1. A non-invasive load decomposition method based on a sparse classifier hierarchical algorithm is characterized by comprising the following steps: the method comprises the following steps:
s1, acquiring voltage and current data of the electric appliance;
s2, calculating low-frequency active power, low-frequency reactive power and high-frequency current harmonic waves of the electric appliance according to the collected voltage and current data;
s3, performing cluster analysis on the low-frequency active power, the low-frequency reactive power and the high-frequency current harmonic of the electric appliance according to a Mean-shift algorithm;
s4, constructing an electric appliance low-frequency active power feature dictionary, a low-frequency reactive power feature dictionary and a high-frequency current harmonic feature dictionary according to the clustering analysis result;
s5, constructing a sparse matrix according to the clustering analysis result and the characteristic that different states of the same electric appliance cannot be used simultaneously;
s6, performing power identification according to the sparse matrix, the low-frequency active power feature dictionary and the low-frequency reactive power feature dictionary;
s7, judging results according to the active power identification results and the reactive power identification results, and if the coincidence number in the judgment results is equal to 1, outputting the judgment results; otherwise, executing step S8;
and S8, performing current harmonic recognition according to the sparse matrix and the high-frequency current harmonic feature dictionary, and outputting a recognition result.
2. The sparse classifier pyramid algorithm-based non-invasive load decomposition method according to claim 1, wherein: step S2 specifically includes the following steps:
s20, calculating the active power P of the electric appliance according to the voltage and current data;
s21, calculating reactive power Q of the electric appliance according to the voltage and current data;
s22, calculating the specific formula of the current harmonic X of the electric appliance according to the voltage and current data, wherein the specific formula is as follows:
in the formula, C0Is a direct current component, Cmsin (m ω t + θ) denotes the m harmonic, CmIs the amplitude of the m-th harmonic, and θ is the initial phase of the m-th harmonic, with an angular frequency of ω.
3. The sparse classifier pyramid algorithm-based non-invasive load decomposition method according to claim 1, wherein: step S4 specifically includes the following steps:
s40, establishing a low-frequency active power feature dictionary, a low-frequency reactive power feature dictionary and a high-frequency current harmonic feature dictionary of the electric appliance i, wherein the specific formulas are as follows:
Pi＝[p1,p2,…,pL]
Qi＝[q1,q2,…,qL]
Xi＝[x1,x2,…,xL]
in the formula, p1，p2,…pLThe active power dictionary values of all the states of the electrical appliance i are obtained; q. q.s1,q2,…,qLDictionary values of reactive power for all states of the electrical appliance i; x is the number of1,x2,…,xLCurrent harmonic dictionary values for all states of the electrical appliance i; wherein p isL,qLPower value, x, representing the state of the applianceLRepresenting the first 15 current harmonics of the state of the electric appliance;
s41, establishing a low-frequency active power feature dictionary, a low-frequency reactive power feature dictionary and a high-frequency current harmonic feature dictionary of all electrical equipment, wherein the specific formulas are as follows:
PD＝[P1,P2,P3,…,PN]
QD＝[Q1,Q2,Q3,…,QN]
XD＝[X1,X2,X3,…,XN]
in the formula, PDDictionary of active power, P, representing all electrical appliances1，P2,…PNThe active power dictionary values of all the states of the N electric appliances are obtained; qDDictionary of reactive powers, Q, representing all electrical appliances1，Q2,…QNDictionary values of reactive power for all states of the N electrical appliances; xDDictionary of high-frequency current harmonics, X, representing all electrical appliances1，X2,…XNCurrent harmonic dictionary values for all states of the N appliances.
4. The sparse classifier pyramid algorithm-based non-invasive load decomposition method according to claim 1, wherein: the sparse matrix formula constructed in step S5 is:
in the formula, each column of the sparse matrix A represents a combination of appliances, and at most four 1 s exist in each column, which means that at most four appliances are simultaneously turned on.
5. The sparse classifier pyramid algorithm-based non-invasive load decomposition method according to claim 1, wherein: step S6 specifically includes the following steps:
s60 active power identification
S600, calculating the minimum Euclidean distance between the input active power and the input active power according to the sparse matrix and the low-frequency active power feature dictionary;
s601, obtaining an active power threshold value according to the set penalty coefficient, and calculating all electric appliance combinations smaller than or equal to the active power threshold value;
s61 reactive power identification
S610, calculating the minimum Euclidean distance from the input reactive power according to the sparse matrix and the low-frequency reactive power feature dictionary;
s611, obtaining a reactive power threshold value according to the set penalty coefficient, and calculating all electric appliance combinations smaller than or equal to the reactive power threshold value.
6. The sparse classifier pyramid algorithm-based non-invasive load decomposition method according to claim 5, wherein: step S6 specifically includes the following steps:
s60 active power identification
S600, the specific formula of the minimum Euclidean distance is as follows:
dp＝min|P-PD×A|
wherein P represents input active power, PDX A represents the active power value of all the electrical appliances in combination, dpThe Euclidean distance representing the minimum active power of the input active power and the active power of all electric appliance combinations;
s601, calculating all electric appliance combinations with active power thresholds less than or equal to the active power thresholds, wherein the specific formula is as follows:
Ap←|P-PD×A|≤dp+ε
wherein P represents input active power, PDX A represents the active power value of all the electrical appliances in combination, dp+ ε represents the active power threshold, ε represents the penalty factor, APRepresenting all the electric appliance combinations with Euclidean distances between the active power of the electric appliance combinations in the sparse matrix A and the input active power smaller than or equal to an active power threshold value;
s61 reactive power identification
S610, the specific formula of the minimum Euclidean distance is as follows:
dq＝min|Q-QD×A|
wherein Q represents the reactive power of the input, QDX A represents the reactive power value of all the electrical appliances in combination, dqThe Euclidean distance representing the minimum reactive power of the input reactive power and the reactive power of all electric appliance combinations;
s611, calculating all electric appliance combinations smaller than or equal to the reactive power threshold value, wherein the specific formula is as follows:
Aq←|Q-QD×A|≤dq+ε
wherein Q represents the reactive power of the input, QDX A represents the reactive power value of all the electrical appliances in combination, dq+ ε represents the reactive power threshold, ε represents the penalty factor, AqAnd all the electric appliance combinations with Euclidean distances between the reactive power of the electric appliance combinations in the sparse matrix A and the input reactive power smaller than or equal to the reactive power threshold value.
7. The sparse classifier pyramid algorithm-based non-invasive load decomposition method according to claim 1, wherein: step S7 specifically includes the following steps: judging the number of the electric appliance combinations in the intersection of all the electric appliance combinations with the active power threshold value and all the electric appliance combinations with the reactive power threshold value, and if the electric appliance combinations are equal to 1, directly outputting a power identification result; otherwise, step S8 is executed.
8. The sparse classifier pyramid algorithm-based non-invasive load decomposition method of claim 7, wherein: the specific formula of the number of electrical appliance combinations in step S7 is as follows:
n，Apq＝f(Ap，Aq)
wherein f () represents taking Ap、AQNumber of combinations of coincidences, n representing number of coincidences, ApqRepresenting the corresponding electric appliance combination, if n is equal to 1, indicating that the power identification result is good, and directly outputting a judgment result if only one electric appliance combination in the database accords with the power characteristic of an identification object; if n is greater than 1, execution proceeds to step S8.
9. The sparse classifier pyramid algorithm-based non-invasive load decomposition method according to claim 1, wherein: step S8 specifically includes the following steps:
performing current harmonic recognition according to the sparse matrix and the high-frequency current harmonic feature dictionary, wherein the specific formula is as follows:
dX＝min|X-XDApq|
wherein X represents the harmonic of the input current, XD×ApqRepresenting the combined harmonic wave of the electric appliance meeting the identification of active power and reactive power, total 15 harmonic waves, dxRepresenting the minimum Euclidean distance between the input current harmonic and the current harmonic of all the electric appliance combinations;
finding the corresponding electric appliance combination according to the minimum Euclidean distance of the current harmonic waves:
Ax←|X-XD×Apq|＝dx
in the formula AxIs represented by ApqIn a collectionElectric appliance combination corresponding to minimum Euclidean distance of input current harmonic wave, AxAnd outputting the identification result for the final result of the electrical appliance identification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011538200.4A CN112560977B (en) | 2020-12-23 | 2020-12-23 | Non-invasive load decomposition method based on sparse classifier hierarchical algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011538200.4A CN112560977B (en) | 2020-12-23 | 2020-12-23 | Non-invasive load decomposition method based on sparse classifier hierarchical algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112560977A true CN112560977A (en) | 2021-03-26 |
| CN112560977B CN112560977B (en) | 2022-09-27 |
Family
ID=75030978
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011538200.4A Active CN112560977B (en) | 2020-12-23 | 2020-12-23 | Non-invasive load decomposition method based on sparse classifier hierarchical algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112560977B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113094931A (en) * | 2021-04-29 | 2021-07-09 | 杭州电力设备制造有限公司 | Non-invasive load decomposition method, device and equipment |
| CN113408341A (en) * | 2021-05-10 | 2021-09-17 | 深圳供电局有限公司 | Load identification method and device, computer equipment and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015059272A1 (en) * | 2013-10-24 | 2015-04-30 | Universite Libre De Bruxelles | Improved non-intrusive appliance load monitoring method and device |
| WO2016079229A1 (en) * | 2014-11-21 | 2016-05-26 | Universite Libre De Bruxelles | Improved non-intrusive appliance load monitoring method and device |
| WO2019155214A1 (en) * | 2018-02-07 | 2019-08-15 | Green Running Limited | Method and apparatus for power signal disaggragation using a convolutional neural network |
| CN110514889A (en) * | 2019-07-19 | 2019-11-29 | 浙江万胜智能科技股份有限公司 | A kind of method and system of non-intrusion type household electricity remained capacity |
| CN110954744A (en) * | 2019-11-18 | 2020-04-03 | 浙江工业大学 | Non-invasive load monitoring method based on event detection |
| CN111007347A (en) * | 2019-12-27 | 2020-04-14 | 广东电网有限责任公司电力科学研究院 | Non-invasive load identification device and related equipment |
| CN111461189A (en) * | 2020-03-23 | 2020-07-28 | 东南大学 | Resident rhythm detection method based on non-invasive measurement technology |
-
2020
- 2020-12-23 CN CN202011538200.4A patent/CN112560977B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015059272A1 (en) * | 2013-10-24 | 2015-04-30 | Universite Libre De Bruxelles | Improved non-intrusive appliance load monitoring method and device |
| WO2016079229A1 (en) * | 2014-11-21 | 2016-05-26 | Universite Libre De Bruxelles | Improved non-intrusive appliance load monitoring method and device |
| WO2019155214A1 (en) * | 2018-02-07 | 2019-08-15 | Green Running Limited | Method and apparatus for power signal disaggragation using a convolutional neural network |
| CN110514889A (en) * | 2019-07-19 | 2019-11-29 | 浙江万胜智能科技股份有限公司 | A kind of method and system of non-intrusion type household electricity remained capacity |
| CN110954744A (en) * | 2019-11-18 | 2020-04-03 | 浙江工业大学 | Non-invasive load monitoring method based on event detection |
| CN111007347A (en) * | 2019-12-27 | 2020-04-14 | 广东电网有限责任公司电力科学研究院 | Non-invasive load identification device and related equipment |
| CN111461189A (en) * | 2020-03-23 | 2020-07-28 | 东南大学 | Resident rhythm detection method based on non-invasive measurement technology |
Non-Patent Citations (3)
| Title |
|---|
| LINGLI 等: ""Multi-objective evolutionary algorithms applied to non-intrusive load monitoring"", 《ELECTRIC POWER SYSTEMS RESEARCH》 * |
| SHIKHA SINGH 等: ""Non-Intrusive Load Monitoring via Multi-Label Sparse Representation-Based Classification"", 《IEEE TRANSACTIONS ON SMART GRID》 * |
| 刘天宇: ""计及设备用电特性和用户用电模式的负荷分解方法"", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113094931A (en) * | 2021-04-29 | 2021-07-09 | 杭州电力设备制造有限公司 | Non-invasive load decomposition method, device and equipment |
| CN113094931B (en) * | 2021-04-29 | 2022-04-22 | 杭州电力设备制造有限公司 | Non-invasive load decomposition method, device and equipment |
| CN113408341A (en) * | 2021-05-10 | 2021-09-17 | 深圳供电局有限公司 | Load identification method and device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112560977B (en) | 2022-09-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103001230B (en) | Non-invasive power load monitoring and decomposing current mode matching method | |
| CN112560977B (en) | Non-invasive load decomposition method based on sparse classifier hierarchical algorithm | |
| Chen et al. | Intelligent home-appliance recognition over IoT cloud network | |
| CN109470957B (en) | Non-invasive electrical appliance identification method based on Frey's distance | |
| CN110726875B (en) | New energy flexible direct-current grid-connected transient harmonic detection method and system | |
| Dash et al. | An appliance load disaggregation scheme using automatic state detection enabled enhanced integer programming | |
| Qu et al. | A residual convolutional neural network with multi-block for appliance recognition in non-intrusive load identification | |
| CN113762355B (en) | User abnormal electricity behavior detection method based on non-invasive load decomposition | |
| Han et al. | Non‐intrusive load monitoring by voltage–current trajectory enabled asymmetric deep supervised hashing | |
| CN113627289A (en) | Deep learning-based power equipment identification method and device | |
| CN111934318B (en) | Non-invasive power load decomposition method, apparatus, device and storage medium | |
| CN111199014B (en) | Time sequence based seq2point NILM method and device | |
| CN116861316A (en) | Electrical appliance monitoring method and device | |
| CN110009231A (en) | A kind of power load recognition methods and device based on repeatability and entropy weight distinction | |
| Zhihua et al. | Research on non-invasive load identification method based on VMD | |
| Xu et al. | Edge-based NILM system with MDMR filter-based feature selection | |
| Liu et al. | Enhancing residential electricity safety and management: a novel non-intrusive load monitoring-based methodology for accurate appliance operational state identification | |
| CN116188199A (en) | Non-invasive load decomposition method and system | |
| CN115825602A (en) | Load identification method and system for comprehensive multivariate information similarity analysis | |
| CN110896218B (en) | Harmonic modeling method and system for establishing collective residential load | |
| CN110542855B (en) | Load switch event detection method and system based on discrete cosine transform | |
| CN113762619B (en) | Distribution Internet of things user load identification method, system, equipment and storage medium | |
| Li et al. | Non-intrusive load monitoring for consistent shape loads based on convolutional neural network | |
| Pengfei et al. | Non-intrusive load monitoring method based on LMD and KNN algorithm | |
| CN109194367A (en) | Power signal reconstructing method in Energy Decomposition |
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 | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20210326 Assignee: Yunnan Yinhuan Power Engineering Co.,Ltd. Assignor: NORTH CHINA ELECTRIC POWER University (BAODING) Contract record no.: X2023420000264 Denomination of invention: A non-invasive load decomposition method based on sparse classifier hierarchical algorithm Granted publication date: 20220927 License type: Common License Record date: 20230726 |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20210326 Assignee: Yunnan Yadian Technology Co.,Ltd. Assignor: NORTH CHINA ELECTRIC POWER University (BAODING) Contract record no.: X2023420000285 Denomination of invention: A non-invasive load decomposition method based on sparse classifier hierarchical algorithm Granted publication date: 20220927 License type: Common License Record date: 20230816 |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20210326 Assignee: HEFEI LIFA ELECTRICAL TECHNOLOGY Co.,Ltd. Assignor: NORTH CHINA ELECTRIC POWER University (BAODING) Contract record no.: X2023980047629 Denomination of invention: A non-invasive load decomposition method based on sparse classifier hierarchical algorithm Granted publication date: 20220927 License type: Common License Record date: 20231120 |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20210326 Assignee: BAODING ELECTRICITY VALLEY TECHNOLOGY Co.,Ltd. Assignor: NORTH CHINA ELECTRIC POWER University (BAODING) Contract record no.: X2024990000173 Denomination of invention: A non-invasive load decomposition method based on sparse classifier hierarchical algorithm Granted publication date: 20220927 License type: Common License Record date: 20240422 |