CN111209976B - Non-invasive load monitoring method and system - Google Patents

Non-invasive load monitoring method and system Download PDF

Info

Publication number
CN111209976B
CN111209976B CN202010035793.6A CN202010035793A CN111209976B CN 111209976 B CN111209976 B CN 111209976B CN 202010035793 A CN202010035793 A CN 202010035793A CN 111209976 B CN111209976 B CN 111209976B
Authority
CN
China
Prior art keywords
load
switching
types
type
current
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010035793.6A
Other languages
Chinese (zh)
Other versions
CN111209976A (en
Inventor
武昕
焦点
高宇辰
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.)
North China Electric Power University
Original Assignee
North China Electric Power University
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 North China Electric Power University filed Critical North China Electric Power University
Priority to CN202010035793.6A priority Critical patent/CN111209976B/en
Publication of CN111209976A publication Critical patent/CN111209976A/en
Application granted granted Critical
Publication of CN111209976B publication Critical patent/CN111209976B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • 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
    • 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply
    • 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
    • Y04S20/222—Demand response systems, e.g. load shedding, peak shaving

Abstract

The invention discloses a non-invasive load monitoring method and a non-invasive load monitoring system. The method comprises the following steps: acquiring total voltage and total current at each user power inlet, and judging whether user load switching operation occurs according to the total current; if switching operation occurs, determining a current waveform and a voltage waveform of a switching load according to a difference value of current before and after switching of the load and the total voltage; extracting load characteristics of the switching load according to the current waveform and the voltage waveform of the switching load; judging load types of load characteristics of the switching load by adopting a trained load classification model to obtain the types of the switching load; storing the current waveform and the voltage waveform of the switching load and the types of the switching load into a user edge load feature library; and monitoring the user load according to the user edge load feature library. By adopting the method and the system provided by the invention, the difference of the use loads of different users is considered, and the edge load feature library of each user is established, so that the load monitoring precision can be effectively improved.

Description

Non-invasive load monitoring method and system
Technical Field
The invention relates to the technical field of load monitoring, in particular to a non-invasive load monitoring method and system.
Background
As the progress of urban and industrial processes has been gradually accelerated, the global power usage has continued to increase. The total power consumption of the common resident users has a huge ratio in the total power consumption, and the problems of low power consumption efficiency, serious power resource waste and the like exist. The effective demand side management can improve the electricity utilization efficiency of the demand side; and the power resources can be reasonably configured. With the advance of management work on the demand side, residential user domain load monitoring is an important factor for realizing intelligent management on the demand side. The actual energy consumption level of various loads in the users can be known through load monitoring, scientific collection and management of energy efficiency data are realized, and the electricity consumption conditions of various users are mastered.
In recent years, non-invasive load monitoring has the advantages of simplicity, easiness in implementation, low economic cost, high reliability and good data integrity. And (3) carrying out data analysis on the power signals collected at the power inlet of the user, and tracking the power consumption condition and detail information of the internal load of the user by utilizing the power parameter characteristics during the load operation and combining the information in the characteristic library which is manually constructed in advance through a decomposition identification algorithm. However, in practice, the workload of a mode of manually constructing a feature library for each user in advance is huge, the load types, brands and models are various, the circuit parameters are different, the features presented in operation are greatly different, the difference of the loads used by different users is not considered in the existing non-invasive load monitoring method, and the load monitoring precision is low.
Disclosure of Invention
The invention aims to provide a non-invasive load monitoring method and system, which take the difference of loads used by different users into consideration, and establish an edge load feature library of each user, so that the load monitoring precision can be effectively improved.
In order to achieve the above object, the present invention provides the following solutions:
a non-invasive load monitoring method, comprising:
acquiring total voltage and total current at each user power inlet;
judging whether user load switching operation occurs according to the total current; if switching operation occurs, determining a current waveform and a voltage waveform of a switching load according to a difference value of current before and after switching of the load and the total voltage; if the switching operation does not occur, returning to the step of acquiring the total voltage and the total current at the power inlet of each user;
extracting load characteristics of the switching load according to the current waveform and the voltage waveform of the switching load;
judging the load types of the load characteristics of the switching load by adopting a trained load classification model to obtain the types of the switching load;
storing the current waveform and the voltage waveform of the switching load and the type of the switching load into a user edge load feature library;
and monitoring the user load according to the user edge load feature library.
Optionally, the load characteristic specifically includes:
load entity characteristics and load attribute characteristics;
the load entity characteristics comprise linear and nonlinear characteristics, continuous operation and intermittent operation characteristics, fixed operation time and non-fixed operation time characteristics of the load;
the characteristic attribute features comprise active power, reactive power, power factor, current distortion rate and pearson correlation coefficient of the load.
Optionally, the load feature of the switching load adopts a trained load classification model to judge the load type, so as to obtain the type of the switching load, which specifically includes:
judging load types of the load characteristics of the switching load by adopting a trained first load classification model to obtain pre-generated load types;
judging whether the pre-generated load types have similar load types or not according to the determined similar load sets; if similar load types exist, judging the load types of the load characteristics of the switching load by adopting a trained second load classification model to obtain the types of the switching load; if the similar load types do not exist, the pre-generated load types are determined as the types of switching loads.
Optionally, the training method of the first load classification model specifically includes:
acquiring a plurality of loads to be trained;
extracting load characteristics of the load to be trained;
clustering the load to be trained according to the load characteristic similarity of the load to be trained to obtain clustered multiple load types;
and training the load characteristics of the load to be trained by adopting a gradient descent method according to the clustered multiple load types to obtain a trained first load classification model and a load characteristic coefficient under each load type.
Optionally, the method for determining the similar load specifically includes:
and determining a measurement value according to the load characteristic coefficient by adopting the following formula:
in the method, in the process of the invention,representing a measurement value, ++>For load type omega n1 N1 represents the n1 st load type, < >>For load type omega n2 N2 represents the n2 nd load type, load characteristic coefficients representing the mth load characteristic;
and determining the load with the measurement value smaller than a preset value as a similar load.
Optionally, the training method of the second load classification model specifically includes:
extracting load characteristics of the similar load;
and training the load characteristics of the similar loads by adopting a gradient descent method according to the load types of the similar loads to obtain a trained second load classification model and load characteristic coefficients among the similar loads.
The present invention also provides a non-invasive load monitoring system comprising:
the total voltage and total current acquisition module is used for acquiring the total voltage and total current at the power inlet of each user;
the user load switching operation judging module is used for judging whether the user load switching operation occurs or not according to the total current; if the switching operation occurs, executing a current-voltage waveform determining module; if the switching operation does not occur, executing a total voltage and total current acquisition module;
the current-voltage waveform determining module is used for determining the current waveform and the voltage waveform of the switching load according to the difference value of the current before and after the switching of the load and the total voltage;
the load characteristic extraction module of the switching load is used for extracting the load characteristic of the switching load according to the current waveform and the voltage waveform of the switching load;
the switching load type determining module is used for judging the load type of the load characteristics of the switching load by adopting a trained load classification model to obtain the type of the switching load;
the user edge load feature library building module is used for storing the current waveform and the voltage waveform of the switching load and the types of the switching load into the user edge load feature library;
and the load monitoring module is used for monitoring the user load according to the user edge load feature library.
Optionally, the type determining module of the switching load specifically includes:
the pre-generated load type determining unit is used for judging the load type of the load characteristics of the switching load by adopting a trained first load classification model to obtain the pre-generated load type;
a similar load type judging unit for judging whether the pre-generated load type has a similar load type according to the determined similar load set; if the similar load types exist, executing a secondary load type determining unit; if the similar load types do not exist, a type determining unit for executing switching loads;
the secondary load type determining unit is used for judging the load type of the load characteristics of the switching load by adopting a trained second load classification model to obtain the type of the switching load;
and a kind determination unit for determining the pre-generated load kind as the kind of the switching load.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides a non-invasive load monitoring method and a system, which only need to acquire the total voltage and the total current at the power inlet of each user, obtain the current waveform and the voltage waveform of the switched load according to the difference value of the current before and after the switching of the load and the total voltage, do not need to enter the house of each user, do not need to power off other equipment when one equipment voltage and current waveform is measured, and realize the non-invasive load monitoring. The load characteristics of the switching load are extracted through the current waveform and the voltage waveform of the switching load, the load types of the switching load are judged by adopting a trained load classification model, the types of the switching load are obtained, the current waveform and the voltage waveform of the switching load and the types of the switching load are stored in a user edge load characteristic library, the monitoring of the user load is realized, the statistics can be carried out for the electricity consumption of each type of load, and the guidance on electricity saving is facilitated.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a non-invasive load monitoring method in accordance with an embodiment of the present invention;
FIG. 2 is a block diagram of a non-invasive load monitoring system in accordance with an embodiment of the present invention;
FIG. 3 is a schematic diagram of hollow load characteristics in an embodiment of the present invention;
fig. 4 is a schematic diagram of load characteristics of 4 kinds of household appliances in an embodiment of the present invention;
FIG. 5 is a graph showing a total current change curve of a user acquired within a sampling time of 2.25h according to an embodiment of the present invention;
fig. 6 is a waveform diagram of load voltage and current in an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention aims to provide a non-invasive load monitoring method and system, which take the difference of loads used by different users into consideration, and establish an edge load feature library of each user, so that the load monitoring precision can be effectively improved.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
Examples
The invention provides a non-invasive load monitoring and determining method, fig. 1 is a flow chart of a non-invasive load monitoring method in an embodiment of the invention, as shown in fig. 1, the method comprises the following steps:
step 101: the total voltage and total current at each user power inlet is obtained.
When a plurality of electric appliances are in a working state in a household, the electric appliances are in parallel connection with each other, and each branch is in a stable state. When a certain electric appliance is not in use, the branch circuit can be regarded as an open circuit, and the passing current is zero. The electric appliances on different branches do not affect each other, so the voltages of all branches are equal, and the total voltage U is acquired mix (t) equals the branch voltage U l (t) collecting the obtained total current I mix (t) for the respective branch currents I l The sum of (t), namely:
U mix (t)＝U l (t)l＝1,2,…,L
wherein L is the total number of the electric appliances in operation.
Step 102: judging whether the user load switching operation occurs according to the total current; if the switching operation occurs, executing step 103; if no switching operation occurs, the process returns to step 101.
The current intensity is the most obvious variation when switching occurs with a load in the circuit. Cycle amperage I RMS The calculation formula is as follows:
ΔI＝I RMS (q+1)-I RMS (q)
wherein T is the sampling point number of one period, I mix (n) is the sampled value of the current for that period, q is the q-th period of the current. η is a threshold value for determining that the steady-state current is suddenly changed, and whether the current is suddenly changed (i.e., load switching) is determined according to the threshold value η=0.1. And judging that the load switching is carried out when the delta I is larger than the threshold value eta, and judging that the load switching is not carried out when the delta I is smaller than the threshold value eta.
Step 103: and determining the current waveform and the voltage waveform of the switching load according to the difference value of the current before and after the switching of the load and the total voltage.
The fundamental phase angle of the load steady-state current is determined by the initial phase of the voltage during measurement, and the steady-state current needs to be measured under the voltage with the same initial phase angle when the difference between the steady-state current before and after switching of the load is calculated to obtain an independent load current waveform due to the current superposition.
U(e)>0,U(e-1)<0
Where e represents the current sampling point satisfying the total current formula.
Searching current sampling points of the two periods before and after the period q when the load switching happens, namely the (q-1) th period and the (q+1) th period, according to a total current formula, and respectively taking data of one periodAnd->
Wherein I is k (n) current waveform representing switching load, U k (n) represents the voltage waveform of the switching load.
Step 104: and extracting the load characteristics of the switching load according to the current waveform and the voltage waveform of the switching load.
Load characteristics, specifically including: load entity characteristics and load attribute characteristics. The functional characteristics of the electric appliance are taken as load entity characteristics, and the operation characteristics of the electric appliance are taken as load attribute characteristics.
Load entity characteristics include linear and nonlinear characteristics of the load, continuous and intermittent operation characteristics, fixed and non-fixed operation time characteristics.
The characteristic attribute features comprise active power, reactive power, power factor, current distortion rate and pearson correlation coefficient of the load.
Taking the operation characteristics of the electric appliance as the load attribute characteristics, independently collecting the voltage and current signals of each common electric appliance, and extracting the steady-state operation characteristics of the electric appliance:
F k ＝Γ(U k ,I k )
wherein: k is the kth electrical load; u (U) k And I k The voltage and current signals of the load k are respectively; f is a feature vector, F= [ F ] 1 ,f 2 ,…,f m ,…,f M ]F is the load characteristic, M is the number of the load characteristic; r is a feature extraction function including, but not limited to, the features set forth in the following equation:
wherein P is k And Q k Active and reactive power; PF (physical filter) k Is a power factor; THD (total heat transfer) k Is the current distortion rate; r is (r) k A pearson correlation coefficient between the current waveform and the constructed standard sine wave; i k,r Representing structured standard sine wave vectors, I k,r Length, peak value and frequency and I k The same applies.And->Respectively is I k And I k,r Standard deviation of (2); phi is the phase angle of the voltage and the current, i is the harmonic frequency, and k is the kth electrical load; u (U) k And I k Voltage, current signal of load k, < ->Is the power factor angle of the load k.
Step 105: and judging the load types of the switching loads by adopting a trained load classification model to the load characteristics of the switching loads, so as to obtain the types of the switching loads.
Step 105 specifically includes:
and judging the load types of the load characteristics of the switching load by adopting a trained first load classification model to obtain the pre-generated load types.
Judging whether the pre-generated load types have similar load types according to the determined similar load sets; if similar load types exist, judging the load types of the load characteristics of the switching load by adopting a trained second load classification model to obtain the types of the switching load; if the similar load types do not exist, the pre-generated load type is determined as the type of the switching load.
The training method of the first load classification model specifically comprises the following steps:
1) And acquiring a plurality of loads to be trained.
2) And extracting the load characteristics of the load to be trained.
3) Clustering the load to be trained according to the load characteristic similarity of the load to be trained to obtain clustered multiple load types.
And (3) carrying out normalization processing on the calculated characteristic f due to different numerical ranges of different characteristics, namely:
f * ＝(f-f min )/(f max -f min )
wherein f max And f min The maximum and minimum values of the load characteristics are respectively; f (f) * Is the characteristic value after characteristic normalization, F * ＝[f 1 * ,f 2 * ,…,f m * ,…,f M * ]Is the normalized feature vector.
Measuring the load similarity d between the characteristic vector and different load characteristic information by using a minimum residual error method k , ω The method comprises the following steps:
d k,ω ＝||F k * -F ω * ||
wherein F is k * To normalize the eigenvector of the afterload k, F ω * Is a characteristic information vector with the load category omega stored in the map.
Class ω with knowledge fusion to the load k k Obtained by the following formula:
ω k ＝arg mind k,ω
the load characteristic information in the map is updated to finish the knowledge fusion of the load characteristic information, namely:
4) And training the load characteristics of the load to be trained by adopting a gradient descent method according to the clustered multiple load types to obtain a trained first load classification model and a load characteristic coefficient under each load type.
The load characteristics are weighted linearly as an objective function by:
wherein f ω,m Is the m-th characteristic in the omega-th load, w ω,m Coefficient of mth characteristic in omega type load, b ω Training bias for class omega load, p ω And outputting results for each load node.
The result is output in the form of a probability vector through the softmax layer, the loss is calculated through the least squares method,
namely:
loss＝||Label-S||
wherein N represents N loads in the map, loss is calculated loss, label is training real Label, S is training predictive Label, S= [ S ] 1 ,S 2 ,…S ω …]。
The characteristic coefficients and training bias are updated by a gradient descent method, namely:
wherein w is new And b new For updated characteristic coefficients and offsets, w old And b old For the pre-update feature coefficients and offsets, α is the training step size.
And (5) repeatedly training and iterating until the parameters are converged, and obtaining the characteristic coefficients of the load knowledge graph.
The method for determining the similar load specifically comprises the following steps:
1) And determining a measurement value according to the load characteristic coefficient by adopting the following formula:
in the method, in the process of the invention,representing a measurement value, ++>For load type omega n1 N1 represents the n1 st load type, < >>For load type omega n2 N2 represents the n2 nd load type, load characteristic coefficients representing the mth load characteristic;
2) And determining the load with the measurement value smaller than the preset value as a similar load.
The training method of the second load classification model specifically comprises the following steps:
1) And extracting load characteristics of similar loads.
2) And training the load characteristics of the similar loads by adopting a gradient descent method according to the load types of the similar loads to obtain a trained second load classification model and load characteristic coefficients among the similar loads.
Training method adopting first load classification model for two kinds of load omega with similar loads n1 And omega n2 Retraining, namely changing the upper limit of the training load type into 2, and determining characteristic coefficients among similar loads after the training parameters are converged
Based on the extracted load characteristic f according to the load characteristic coefficient k，m And judging the load type. The method comprises the following steps:
L ω ＝arg maxp ω
wherein p is ω Is the probability that the load k belongs to category ω, L ω Is the judgment type of the load k.
Step 106: and storing the current waveform and the voltage waveform of the switching load into a user edge load characteristic library.
Step 107: and monitoring the user load according to the user edge load feature library.
The invention only needs to obtain the total voltage and the total current at the power inlet of each user, obtains the current waveform and the voltage waveform of the switching load according to the difference value of the current before and after the switching of the load and the total voltage, does not need to enter the house of each user, does not need to power off other equipment when measuring the voltage and the current waveform of one equipment, and realizes non-invasive load monitoring. The load characteristics of the switching load are extracted through the current waveform and the voltage waveform of the switching load, the load types of the switching load are judged by adopting a trained load classification model, the types of the switching load are obtained, the current waveform and the voltage waveform of the switching load and the types of the switching load are stored in a user edge load characteristic library, the monitoring of the user load is realized, the statistics can be carried out for the electricity consumption of each type of load, and the guidance on electricity saving is facilitated.
The present invention also provides a non-invasive load monitoring system, fig. 2 is a structural diagram of the non-invasive load monitoring system in the embodiment of the present invention, as shown in fig. 2, the system includes:
a total voltage and total current acquisition module 201 for acquiring a total voltage and total current at each user power inlet.
A user load switching operation judging module 202, configured to judge whether a user load switching operation occurs according to the total current; if the switching operation occurs, the current-voltage waveform determining module 203 is executed; if no switching operation occurs, the total voltage and total current acquisition module 201 is executed.
The current-voltage waveform determining module 203 is configured to determine a current waveform and a voltage waveform of the switching load according to the difference value between the current before and after the switching of the load and the total voltage.
The load characteristic extraction module 204 of the switching load is used for extracting the load characteristic of the switching load according to the current waveform and the voltage waveform of the switching load.
The type determining module 205 of the switching load is configured to determine a load type of the load feature of the switching load by using a trained load classification model, so as to obtain a type of the switching load.
The type determining module 205 of the switching load specifically includes:
the pre-generated load type determining unit is used for judging load types of load characteristics of switching loads by adopting a trained first load classification model to obtain pre-generated load types;
a similar load type judging unit for judging whether the pre-generated load type has a similar load type according to the determined similar load set; if the similar load types exist, executing a secondary load type determining unit; if the similar load types do not exist, a type determining unit for executing switching loads;
the secondary load type determining unit is used for judging the load type of the load characteristics of the switching load by adopting a trained second load classification model to obtain the type of the switching load;
and a kind determination unit for determining the kind of the pre-generated load as the kind of the switching load.
The user edge load feature library building module 206 is configured to store the current waveform and the voltage waveform of the switching load and the type of the switching load into the user edge load feature library.
The load monitoring module 207 is configured to monitor a user load according to the user edge load feature library.
For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
Fig. 3 is a schematic view of load characteristics of an air conditioner, and fig. 4 is a schematic view of load characteristics of 4 kinds of household appliances. In fig. 4, only load characteristics of 4 common household appliances (air conditioner, television, computer and electric cooker) are shown, and in fig. 4, a load characteristic knowledge graph is constructed according to four selected loads and eight characteristics thereof, wherein a dotted line represents a physical characteristic, a solid line represents an attribute characteristic, and according to coefficient similarity, the television and the computer belong to similar loads. The numbers in fig. 3 and 4 represent the probability that the load belongs to the load characteristic.
Fig. 5 is a graph showing the total current change of the user acquired during a sampling time of 2.25h, and shows the load switching events detected by using abrupt changes in the current effective value, and during this time, there are 5 load switching events. According to the feature extraction and the feature clustering, the event 2 and the event 3 are switched for the same category of loads.
Fig. 6 is a waveform diagram of load voltage and current resulting from extraction of each switching event in fig. 5, wherein switching loads represented by event 3 and event 4 are clustered into similar loads according to waveform characteristics thereof. From the waveform, event 2 represents the load off and event 3 represents the load on.
Table 1 is a classification confusion matrix for judging the type of the load waveform shown in fig. 5 by the load characteristic knowledge graph. As can be seen from the event probability of the table 1, the load type of the event 1 is an electric cooker, the load types of the event 2 and the event 3 are televisions, the load type of the event 4 is a computer, and the load type of the event 5 is an air conditioner. Event 2 and event 3 are the same type of load and event 3 and event 4 are similar loads.
TABLE 1 event probability
Probability of Electric cooker Television set Computer with a computer program Air conditioner
Event 1 0.9186 0.0551 0.0027 0.0234
Event 2 0.4390 0.6855 0.2684 0.0021
Event 3 0.4770 0.7313 0.2149 0.0610
Event 4 0.0061 0.2930 0.9468 0.1790
Event 5 0.0071 0.0318 0.0042 0.9569
The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In summary, the present description should not be construed as limiting the invention.

Claims (6)

1. A method of non-invasive load monitoring, comprising:
acquiring total voltage and total current at each user power inlet;
judging whether user load switching operation occurs according to the total current; if switching operation occurs, determining a current waveform and a voltage waveform of a switching load according to a difference value of current before and after switching of the load and the total voltage; if the switching operation does not occur, returning to the step of acquiring the total voltage and the total current at the power inlet of each user;
extracting load characteristics of the switching load according to the current waveform and the voltage waveform of the switching load;
judging the load types of the load characteristics of the switching load by adopting a trained load classification model to obtain the types of the switching load; the method specifically comprises the following steps:
judging load types of the load characteristics of the switching load by adopting a trained first load classification model to obtain pre-generated load types;
judging whether the pre-generated load types have similar load types or not according to the determined similar load sets; if similar load types exist, judging the load types of the load characteristics of the switching load by adopting a trained second load classification model to obtain the types of the switching load; if the similar load types do not exist, determining the pre-generated load types as the types of switching loads;
storing the current waveform and the voltage waveform of the switching load and the type of the switching load into a user edge load feature library;
and monitoring the user load according to the user edge load feature library.
2. The non-invasive load monitoring method according to claim 1, wherein the load signature, in particular, comprises:
load entity characteristics and load attribute characteristics;
the load entity characteristics comprise linear and nonlinear characteristics, continuous operation and intermittent operation characteristics, fixed operation time and non-fixed operation time characteristics of the load;
the load attribute characteristics include active power, reactive power, power factor, current distortion rate, pearson correlation coefficient of the load.
3. The non-invasive load monitoring method according to claim 1, wherein the training method of the first load classification model specifically comprises:
acquiring a plurality of loads to be trained;
extracting load characteristics of the load to be trained;
clustering the load to be trained according to the load characteristic similarity of the load to be trained to obtain clustered multiple load types;
and training the load characteristics of the load to be trained by adopting a gradient descent method according to the clustered multiple load types to obtain a trained first load classification model and a load characteristic coefficient under each load type.
4. A method of non-invasive load monitoring according to claim 3, characterized in that the method of determining similar loads comprises:
and determining a measurement value according to the load characteristic coefficient by adopting the following formula:
in the method, in the process of the invention,representing a measurement value, ++>For load type omega n1 N1 represents the n1 st load type, < >>For load type omega n2 N2 represents the n2 nd load type, load characteristic coefficients representing the mth load characteristic;
and determining the load with the measurement value smaller than a preset value as a similar load.
5. The method of non-invasive load monitoring according to claim 4, wherein the training method of the second load classification model specifically comprises:
extracting load characteristics of the similar load;
and training the load characteristics of the similar loads by adopting a gradient descent method according to the load types of the similar loads to obtain a trained second load classification model and load characteristic coefficients among the similar loads.
6. A non-invasive load monitoring system, comprising:
the total voltage and total current acquisition module is used for acquiring the total voltage and total current at the power inlet of each user;
the user load switching operation judging module is used for judging whether the user load switching operation occurs or not according to the total current; if the switching operation occurs, executing a current-voltage waveform determining module; if the switching operation does not occur, executing a total voltage and total current acquisition module;
the current-voltage waveform determining module is used for determining the current waveform and the voltage waveform of the switching load according to the difference value of the current before and after the switching of the load and the total voltage;
the load characteristic extraction module of the switching load is used for extracting the load characteristic of the switching load according to the current waveform and the voltage waveform of the switching load;
the switching load type determining module is used for judging the load type of the load characteristics of the switching load by adopting a trained load classification model to obtain the type of the switching load;
the user edge load feature library building module is used for storing the current waveform and the voltage waveform of the switching load and the types of the switching load into the user edge load feature library;
the load monitoring module is used for monitoring the user load according to the user edge load feature library;
the type determining module of the switching load specifically comprises:
the pre-generated load type determining unit is used for judging the load type of the load characteristics of the switching load by adopting a trained first load classification model to obtain the pre-generated load type;
a similar load type judging unit for judging whether the pre-generated load type has a similar load type according to the determined similar load set; if the similar load types exist, executing a secondary load type determining unit; if the similar load types do not exist, a type determining unit for executing switching loads;
the secondary load type determining unit is used for judging the load type of the load characteristics of the switching load by adopting a trained second load classification model to obtain the type of the switching load;
and a kind determination unit for determining the pre-generated load kind as the kind of the switching load.
CN202010035793.6A 2020-01-14 2020-01-14 Non-invasive load monitoring method and system Active CN111209976B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010035793.6A CN111209976B (en) 2020-01-14 2020-01-14 Non-invasive load monitoring method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010035793.6A CN111209976B (en) 2020-01-14 2020-01-14 Non-invasive load monitoring method and system

Publications (2)

Publication Number Publication Date
CN111209976A CN111209976A (en) 2020-05-29
CN111209976B true CN111209976B (en) 2023-07-21

Family

ID=70786770

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010035793.6A Active CN111209976B (en) 2020-01-14 2020-01-14 Non-invasive load monitoring method and system

Country Status (1)

Country Link
CN (1) CN111209976B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112396007B (en) * 2020-11-24 2021-08-24 国网江苏省电力有限公司营销服务中心 Non-invasive three-threshold detection method and system for load sudden change event of residents
CN112701683A (en) * 2020-12-23 2021-04-23 国网黑龙江省电力有限公司双鸭山供电公司 Non-invasive load detection and decomposition method, system and storage medium
CN113255236B (en) * 2021-07-07 2021-10-08 浙江大学 Non-invasive load self-adaptive identification method based on twin network
CN114239762A (en) * 2022-02-28 2022-03-25 华北电力大学 Non-invasive load identification method and system based on structured load characteristic spectrum
CN115769238A (en) * 2022-06-10 2023-03-07 深圳市正浩创新科技股份有限公司 Load identification method, computer-readable storage medium and device
CN116859169B (en) * 2023-09-01 2023-11-17 国网浙江省电力有限公司余姚市供电公司 Non-invasive power load decomposition method and system

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE10016999A1 (en) * 2000-04-05 2001-10-18 Siemens Ag Switching method for reactive load connected to AC supply

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108021736B (en) * 2017-11-08 2020-06-16 天津大学 Load switching action monitoring method based on sliding window residual error model

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE10016999A1 (en) * 2000-04-05 2001-10-18 Siemens Ag Switching method for reactive load connected to AC supply

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Research on Non-intrusive Load Identification Technology Based on Deep Learning;ShengBo SUN et al.;《2019 IEEE 3rd Conference on Energy Internet and Energy System Integration》;全文 *

Also Published As

Publication number Publication date
CN111209976A (en) 2020-05-29

Similar Documents

Publication Publication Date Title
CN111209976B (en) Non-invasive load monitoring method and system
CN110956220B (en) Non-invasive household appliance load identification method
Heo et al. Toward load identification based on the Hilbert transform and sequence to sequence long short-term memory
CN108899892A (en) One kind being based on CNN non-intrusion type power load decomposition method
CN111242161B (en) Non-invasive non-resident user load identification method based on intelligent learning
CN109633301B (en) Non-invasive electrical appliance load identification method based on quantum genetic optimization
CN111722028A (en) Load identification method based on high-frequency data
CN108918932A (en) Power signal adaptive filter method in load decomposition
CN113361831B (en) Non-invasive load identification electric quantity decomposition method and system based on priority distribution
CN113191253A (en) Non-invasive load identification method based on feature fusion under edge machine learning
CN109377052B (en) Power load matching method, device, equipment and readable storage medium
CN111949707A (en) Shadow field-based hidden Markov model non-invasive load decomposition method
CN110909786A (en) New user load identification method based on characteristic index and decision tree model
CN113193654A (en) Event-driven non-intrusive power load monitoring method based on transient and steady state combination characteristics
Mei et al. Random forest based adaptive non-intrusive load identification
CN111239484A (en) Non-invasive load electricity consumption information acquisition method for non-resident users
CN110009231A (en) A kind of power load recognition methods and device based on repeatability and entropy weight distinction
CN114818827A (en) Non-invasive load decomposition method based on seq2point network
CN110896218B (en) Harmonic modeling method and system for establishing collective residential load
CN110889465B (en) Power demand side equipment identification method and system based on self-adaptive resonant network
CN114186631A (en) Load identification method based on non-invasive intelligent terminal
CN111948446A (en) Self-adaptive load identification method and intelligent electric energy meter
Liu et al. Nonintrusive Load Disaggregation Combining with External Attention Mechanism and Seq2Piont
Li et al. Non-intrusive load monitoring for consistent shape loads based on convolutional neural network
CN115563511B (en) Edge side power fingerprint identification method and device based on machine learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant