CN113569952A - Non-invasive load identification method and system - Google Patents
Non-invasive load identification method and system Download PDFInfo
- Publication number
- CN113569952A CN113569952A CN202110861575.2A CN202110861575A CN113569952A CN 113569952 A CN113569952 A CN 113569952A CN 202110861575 A CN202110861575 A CN 202110861575A CN 113569952 A CN113569952 A CN 113569952A
- Authority
- CN
- China
- Prior art keywords
- data set
- actual
- experimental
- classifier
- sample data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000002474 experimental method Methods 0.000 claims description 20
- 230000005611 electricity Effects 0.000 claims description 14
- 238000011156 evaluation Methods 0.000 claims description 13
- 238000012544 monitoring process Methods 0.000 claims description 12
- 238000002372 labelling Methods 0.000 claims description 6
- 230000000694 effects Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000012360 testing method Methods 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
- G06F18/24—Classification techniques
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Economics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a non-invasive load identification method and a system, a non-invasive terminal acquires an actual sample data set of an experimental household appliance, the actual sample data set and the experimental sample data set are combined, the combined sample data set trains a classifier trained for one time, the non-invasive terminal utilizes the classifier trained for two times to identify the actual equipment category of the experimental household appliance, the experimental sample data set slides forwards in the combined sample data set for presetting the number of samples, a part of data set marked by an original laboratory is removed, newly acquired actual data is added, so that a training sample is closer to an experimental household, and the classifier is continuously optimized and iterated by using the updated experimental sample data set until the performance of the classifier reaches the optimum. The invention continuously updates the samples by utilizing the acquired real family data, reduces the dependence on a large number of marked data sets and effectively identifies the types of various household appliance loads.
Description
Technical Field
The invention relates to the field of intelligent power utilization, in particular to a non-invasive load identification method and system.
Background
There are 2 common implementation methods for load monitoring, and Intrusive Load Monitoring (ILM) is mainly to monitor equipment by installing sensors on each piece of electric equipment; the essential of non-intrusive load monitoring (NILM) is to install a sensor at the entrance of a power consumer to acquire load information, identify the power consumption information of the consumer through an algorithm, install a data acquisition sensing device at the entrance of the consumer, analyze the current and voltage data of the consumer to acquire the working operation state of the internal electrical appliances of the consumer, and thus know the operation condition and the power consumption law of the electrical appliances in the house of the consumer. Non-invasive load monitoring technique can be simpler the realization to the monitoring of resident's power consumption load data, can also go deep into the operational mode and the consumption of understanding each electrical apparatus, energy supplier can formulate energy-conserving policy according to the comprehensive information of domestic appliance operation, the consumer can do more reasonable cost saving selection according to supplier's feedback, can realize electric wire netting and user's nimble two-way interaction, it is significant to smart power grids's development, to the home subscriber, also have clear understanding to self power consumption information, thereby can the more reasonable power consumption.
In recent years, with the proposal of an effective training algorithm of a deep network, deep learning succeeds in multiple fields by means of strong representation learning capacity and is widely applied. NILM is a multi-label classification problem that is run by multiple devices simultaneously, but deep learning often requires the use of large sets of labeled data, which is time consuming and labor intensive to gather and annotate.
Disclosure of Invention
The invention aims to provide a non-invasive load identification method and a non-invasive load identification system, which utilize collected real family data to continuously update samples so as to reduce the dependence on a large number of marked data sets and effectively identify the types of various loads.
In order to achieve the purpose, the invention provides the following scheme:
a non-intrusive load identification method, the non-intrusive load identification method being applied to a non-intrusive load monitoring system, the method comprising:
acquiring an experiment sample data set of the household appliances of the experiment family through a non-invasive cloud; the experimental sample data set comprises an experimental characteristic data set of the power consumption of the household appliance and an experimental equipment category set of the household appliance;
training a classifier by using the experimental sample data set to obtain a once-trained classifier, and downloading the once-trained classifier to a non-invasive terminal;
acquiring an actual characteristic data set of the electricity consumption of the experimental household appliance by using the non-invasive terminal, and inputting the actual characteristic data set into the classifier trained at one time to obtain an actual equipment class set of the experimental household appliance;
forming an actual sample data set of the experimental household appliance by the actual characteristic data set and the actual equipment category set, and transmitting the actual sample data set to the non-invasive cloud;
merging the experimental sample data set and the actual sample data set at the non-invasive cloud end, and training the once-trained classifier by using the merged sample data set to obtain a secondarily-trained classifier;
downloading the secondarily trained classifier as a new primarily trained classifier to a non-invasive terminal;
sliding the experiment sample data set forward in the combined sample data set by a preset sample number to obtain a new experiment sample data set, returning to the step of obtaining an actual characteristic data set of the electricity consumption of the experimental household appliance by using the non-invasive terminal, inputting the actual characteristic data set into the classifier trained at one time to obtain an actual equipment classification set of the experimental household appliance until the identification performance of the classifier on the household appliance reaches the optimum, and obtaining the optimized classifier;
and obtaining the final actual equipment category of each household appliance of the target family by using the optimized classifier through the non-invasive terminal.
Further, the experimental characteristic data set of the household electrical appliance electricity consumption is f ═ { I ═ Irms,Urms,S,P,Q,φ,PF,FW,HAR2nd,HAR3rd,…,HAR19th};
Wherein f is an experimental characteristic data set of household appliance power utilization, IrmsIs the effective value of the current, UrmsIs the effective value of voltage, S is admittance, P is active power, Q is reactive power, phi is phase angle, PF is power factor, FM is fundamental wave, HAR2ndIs the 2 nd harmonic, HAR3rdIs the 3 rd harmonic, HAR19thIs the 19 th harmonic.
Further, the step of forming an actual sample data set of the experimental home appliance by using the actual feature data set and the actual device category set specifically includes:
labeling the actual characteristic data set according to the actual equipment category set and the Internet of things label of the experimental household appliance to obtain an actual label set of the experimental household appliance;
and the actual characteristic data set and the actual label set form an actual sample data set of the experimental household appliance.
Further, the non-invasive cloud is used for merging the experimental sample data set and the actual sample data set, training the classifier trained for the first time by using the merged sample data set, and obtaining a classifier trained for the second time, and then the method further comprises the following steps:
evaluating the secondarily trained classifier by taking the accuracy and the F-score as evaluation indexes;
and downloading the secondarily trained classifier with the accuracy rate greater than or equal to the accuracy rate threshold and the F-score greater than or equal to the F score threshold as a new primarily trained classifier to the non-invasive terminal.
A non-intrusive load identification system, the non-intrusive load identification system being applied to a non-intrusive load monitoring system, the system comprising:
the experimental sample data set acquisition module is used for acquiring an experimental sample data set of the experimental household appliance through a non-invasive cloud; the experimental sample data set comprises an experimental characteristic data set of the power consumption of the household appliance and an experimental equipment category set of the household appliance;
the classifier obtaining module is used for training the classifier by using the experimental sample data set to obtain a trained classifier, and downloading the trained classifier to a non-invasive terminal;
the experimental household appliance actual equipment category set obtaining module is used for obtaining an actual characteristic data set of the experimental household appliance electricity utilization by utilizing the non-invasive terminal, inputting the actual characteristic data set into the classifier which is trained at one time, and obtaining an actual equipment category set of the experimental household appliance;
the actual sample data set forming module is used for forming an actual sample data set of the experimental household appliance by the actual characteristic data set and the actual equipment category set and transmitting the actual sample data set to the non-invasive cloud terminal;
the classifier obtaining module which is trained for the second time is used for merging the experimental sample data set and the actual sample data set at the non-invasive cloud end, and training the classifier which is trained for the first time by using the merged sample data set to obtain a classifier which is trained for the second time;
the classifier replacing module is used for downloading the secondarily trained classifier serving as a new primarily trained classifier to a non-invasive terminal;
the optimized classifier obtaining module is used for sliding the experiment sample data set forward in the combined sample data set by the preset number of samples to obtain a new experiment sample data set, and returning to the step of obtaining an actual characteristic data set of the electricity consumption of the experimental household appliance by using the non-invasive terminal, inputting the actual characteristic data set into the classifier trained at one time to obtain an actual equipment classification set of the experimental household appliance until the identification performance of the classifier on the household appliance is optimal, and obtaining the optimized classifier;
and the final actual equipment category obtaining module is used for obtaining the final actual equipment category of each household appliance of the target family by using the optimized classifier through the non-invasive terminal.
Further, the experimental characteristic data set of the household electrical appliance electricity consumption is f ═ { I ═ Irms,Urms,S,P,Q,φ,PF,FW,HAR2nd,HAR3rd,…,HAR19th};
Wherein f is an experimental characteristic data set of household appliance power utilization, IrmsIs the effective value of the current, UrmsIs the effective value of voltage, S is admittance, P is active power, Q is reactive power, phi is phase angle, PF is power factor, FM is fundamental wave, HAR2ndIs the 2 nd harmonic, HAR3rdIs the 3 rd harmonic, HAR19thIs the 19 th harmonic.
Further, the actual sample data set constructing module specifically includes:
the actual label set obtaining submodule is used for labeling the actual characteristic data set according to the actual equipment category set and the Internet of things label of the experimental household appliance to obtain an actual label set of the experimental household appliance;
and the actual sample data set forming submodule is used for forming an actual sample data set of the experimental household appliance by the actual characteristic data set and the actual label set.
Further, the system further comprises:
the evaluation module is used for evaluating the secondarily trained classifier by taking the accuracy and the F-score as evaluation indexes;
and the downloading module is used for downloading the secondarily trained classifier with the accuracy rate larger than or equal to the accuracy rate threshold and the F-score larger than or equal to the F score threshold to the non-invasive terminal as a new primarily trained classifier.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides a non-invasive load identification method and a system, firstly storing an experimental sample data set of an experimental household appliance at a non-invasive cloud end, training a classifier by using the experimental sample data set, downloading the classifier trained for one time to a non-invasive terminal, acquiring an actual characteristic data set of the power consumption of the experimental household appliance by the non-invasive terminal, acquiring an actual equipment class set of the experimental household appliance through the trained classifier for one time, merging the actual sample data set and the experimental sample data set, training the classifier trained for one time by using the merged sample data set to obtain a classifier trained for two times, identifying the actual equipment class of the experimental household appliance by using the classifier trained for two times by the non-invasive terminal, and sliding the experimental sample data set in the merged sample data set forward by preset sample number, a part of the originally marked data set of the laboratory is removed, and newly acquired actual data are added, so that the training sample is closer to the experimental family, the experimental sample data set is continuously updated, and the classifier is continuously optimized and iterated by using the updated experimental sample data set until the performance of the classifier reaches the optimum. The invention continuously updates the samples by utilizing the acquired real family data, reduces the dependence on a large number of marked data sets and effectively identifies the types of various household appliance loads.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used 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 it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
Fig. 1 is a system architecture diagram of a non-intrusive load identification method provided by the present invention;
FIG. 2 is a schematic diagram of a non-invasive load identification method according to the present invention;
fig. 3 is a flowchart of a non-intrusive load identification method provided by the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a non-invasive load identification method and a non-invasive load identification system, which utilize collected real family data to continuously update samples so as to reduce the dependence on a large number of marked data sets and effectively identify the types of various loads.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
A non-invasive load identification method, as shown in fig. 1-3, the non-invasive load identification method is applied to a non-invasive load monitoring system, and the method includes:
s101, acquiring an experiment sample data set of the household appliances of the experiment family through a non-invasive cloud; the experimental sample data set comprises an experimental characteristic data set of the household appliance power consumption and an experimental equipment category set of the household appliance; wherein, the experimental characteristic data set of the household electrical appliance power consumption is f ═ { I ═ Irms,Urms,S,P,Q,φ,PF,FW,HAR2nd,HAR3rd,…,HAR19th};
Wherein f is an experimental characteristic data set of household appliance power utilization, IrmsIs the effective value of the current, UrmsIs the effective value of voltage, S is admittance, P is active power, Q is reactive power, phi is phase angle, PF is power factor, FM is fundamental wave, HAR2ndIs the 2 nd harmonic, HAR3rdIs the 3 rd harmonic, HAR19thIs the 19 th harmonic.
And S102, training the classifier by using the experimental sample data set to obtain the classifier which is trained once, and downloading the classifier which is trained once to the non-invasive terminal.
S103, acquiring an actual characteristic data set of the power consumption of the experimental household appliance by using the non-invasive terminal, and inputting the actual characteristic data set into a classifier which is trained once to obtain an actual equipment category set of the experimental household appliance.
S104, forming an actual sample data set of the experimental household appliance by the actual characteristic data set and the actual equipment category set, and transmitting the actual sample data set to a non-invasive cloud terminal, wherein the method specifically comprises the following steps:
labeling the actual characteristic data set according to the actual equipment category set and the Internet of things label of the experimental household appliance to obtain an actual label set of the experimental household appliance;
the actual characteristic data set and the actual label set form an actual sample data set of the experimental household appliance.
And S105, merging the experimental sample data set and the actual sample data set at the non-invasive cloud end, and training the classifier trained for the first time by using the merged sample data set to obtain the classifier trained for the second time.
After S105, further comprising:
evaluating the secondarily trained classifier by taking the accuracy and the F-score as evaluation indexes;
and downloading the secondarily trained classifier with the accuracy rate greater than or equal to the accuracy rate threshold and the F-score greater than or equal to the F score threshold as a new primarily trained classifier to the non-invasive terminal.
And S106, downloading the classifier which is trained twice as a new classifier which is trained once to the non-invasive terminal.
S107, sliding the experiment sample data set forward in the combined sample data set by a preset sample number to obtain a new experiment sample data set, returning to the step of obtaining an actual characteristic data set of the electricity consumption of the experiment household appliance by using a non-invasive terminal, inputting the actual characteristic data set into a classifier trained at one time to obtain an actual equipment classification set of the experiment household appliance until the identification performance of the classifier on the household appliance reaches the optimum, and obtaining the optimized classifier;
and S108, obtaining the final actual equipment category of each household appliance of the target family by using the optimized classifier through the non-invasive terminal.
The following describes the operation of the present invention in detail with a specific embodiment.
Step 1: labeled data set training classifier
Given all household appliance data sets D (including characteristic data corresponding to various types of household appliances) in a laboratory, selecting part of the household appliances in the laboratory through analysis of a target household, acquiring data such as voltage and current at high frequency in order to obtain more comprehensive household appliance characteristic information, and calculating through basic data to obtain data such as active power and reactive power to form a marked data set { XL, YL } { X1: L, Y1: L }, wherein X is a subset of all household appliance data spaces D, Y is a label set (the content of the label set is the type of the household appliance), L represents a sample, training a classifier P on the cloud by using the X and Y data sets, verifying the classifier P, and downloading the classifier P to a NILM terminal.
And extracting appropriate characteristics from the acquired mixed signal to perform load identification on the household appliance, wherein the characteristics are represented by f, and the expression is as follows:
f={Irms,Urms,S,P,Q,φ,PF,FW,HAR2nd,HAR3rd,…,HAR19th} (1)
in the formula: i isrmsRepresenting the effective value of the current, UrmsRepresenting the effective value of the voltage, S representing the admittance, P representing the active power, Q representing the reactive power, phi representing the phase angle, PF representing the power factor, FM representing the fundamental, HAR3rdRepresents the 3 rd harmonic, HAR19thRepresenting the 19 th harmonic.
Step 2: uploading family data and identification results to cloud database
The method comprises the steps that an NILM terminal collects actual electricity consumption data of a target family, characteristic data XU is extracted, U represents the number of collected real family samples, household appliances are identified by a classifier which is trained in the step 1 and downloaded to the NILM terminal, an identification result is uploaded to a cloud database, meanwhile, the collected characteristic data XU of the target family is also uploaded to the cloud, and the NILM terminal uploads target family data and identification of the household appliances in real time.
And step 3: new data set training classifier
The method comprises the steps of marking the collected unmarked actual family sample data, using Internet of things labels (household appliances which can be networked exist in a family, and household appliance start-stop information on the Internet of things household cloud of a resident user can provide some label data, as shown in fig. 1) and identification results uploaded by an NILM terminal to collect target family data sets XU (X L +1: and L + U is marked to obtain a label set YU ═ YL +1: l + U, the labeled data set { XL, YL } ═ X1: l, Y1: l and the newly acquired tagged family data set { XU, YU } ═ XL +1: l + U, YL +1: l + U } are merged into a new data set, which is updated with { XL + U, YL + U } ═ X1: l + U, Y1: l +: u, the classifier P is retrained by the data set, thereby increasing its accuracy.
And 4, step 4: classifier evaluation
The cloud platform needs to continuously evaluate the performance of the classifier P, and if the classifier P is qualified, a new trained classifier is downloaded to the NILM terminal to identify the household appliance; if P is not qualified, the classifier is not updated.
The effect of the classification was evaluated, here using accuracy and F-score as indicators.
The accuracy can be used to evaluate the overall effect of classification, and represents the ratio of the correct prediction results, i.e. the ratio Acc of the number of correct classifications to the total number of classifications. The accuracy calculation formula is as follows:
in the formula, Tr represents the total number of correct predictions of the positive samples and the negative samples after the convolutional neural network model is determined, Tw represents the total number of incorrect predictions of the positive samples and the negative samples after the convolutional neural network model is determined, and Tw + Tr represents the total number of test samples.
F-score represents the harmonic mean evaluation index of precision (precision) and recall (recall) in the model. Wherein, the accuracy rate represents the proportion of the actual positive samples in the samples with positive prediction results, and is represented by P, and the expression is as follows:
where Trp represents the number of positive samples that are correctly predicted, and Twp represents the number of positive samples that are identified as positive, but actually negative samples.
And R represents the recall ratio, i.e., the number of correctly predicted positive samples out of all positive samples, which can be obtained by the following formula.
Where Trp represents the number of positive samples that are predicted to be correct, and Twn represents the number of positive samples that are recognized as negative, but actually positive.
F-score is the calculation of the weighted average of precision and recall and is expressed as:
since the precision rate and the recall rate are mutually exclusive, the F-score index is selected to balance the precision rate and the recall rate.
And 5: forward slip sample
The new classifier P has better performance, after the evaluation is qualified, the training set is made to slide forward by I samples (I < L) to obtain a new data set { XL, YL }' (XI: L + I, YI: L + I), the new sample data set after the sliding removes a part of the data set marked by the original laboratory, and newly acquired actual data is added, so that the training sample is closer to a target family, the performance of the classifier trained by the new data set becomes better along with the increase of the sliding times of the samples, and finally the trained and optimized new classifier P is downloaded to a non-invasive terminal.
Step 6: go to step 2
And uploading the family data and the identification result to the cloud again, and then continuously optimizing and iterating the classifier until the performance of the classifier reaches the optimum.
The method comprises the steps of firstly collecting a marked data set in a laboratory, predicting the type and the number of target household appliances according to expert judgment or internet query, then analyzing the marked data set, selecting a feature set and training a classifier, installing an NILM terminal in a target household, collecting aggregated data of the household appliances to form a data set, transmitting the feature data to a cloud platform, then connecting the laboratory data set and the household data set, updating the classifier, sliding the training set forwards after the evaluation is qualified, finally downloading a new classifier to the NILM terminal, and repeatedly and iteratively updating the classifier to achieve the optimum. The invention can reduce the dependence on a large number of marked data sets under the condition of non-invasive power utilization data acquisition, thereby effectively identifying various loads of users, providing powerful support for load management of a power grid company and being an effective implementation method for non-invasive load monitoring.
The invention also provides a non-invasive load identification system, which is applied to a non-invasive load monitoring system and comprises:
the experimental sample data set acquisition module is used for acquiring an experimental sample data set of the experimental household appliance through a non-invasive cloud; the experimental sample data set comprises an experimental characteristic data set of the household appliance power consumption and an experimental equipment category set of the household appliance;
the classifier obtaining module is used for training the classifier by utilizing the experimental sample data set to obtain the classifier which is trained for the first time and downloading the classifier which is trained for the first time to the non-invasive terminal;
the experimental household appliance power utilization system comprises an experimental household appliance actual equipment category set obtaining module, a power utilization control module and a power utilization control module, wherein the experimental household appliance actual equipment category set obtaining module is used for obtaining an actual characteristic data set of experimental household appliance power utilization by utilizing a non-invasive terminal, inputting the actual characteristic data set into a classifier which is trained once, and obtaining an actual equipment category set of the experimental household appliance;
the actual sample data set forming module is used for forming an actual sample data set of the experimental household appliance by the actual characteristic data set and the actual equipment category set and transmitting the actual sample data set to the non-invasive cloud;
the classifier obtaining module which is trained for the second time is used for merging the experimental sample data set and the actual sample data set at the non-invasive cloud end, and training the classifier which is trained for the first time by using the merged sample data set to obtain the classifier which is trained for the second time;
the classifier replacing module is used for downloading the secondarily trained classifier serving as a new primarily trained classifier to the non-invasive terminal;
the optimized classifier obtaining module is used for sliding the experiment sample data set forward in the combined sample data set by the preset number of samples to obtain a new experiment sample data set, and returning to the step of obtaining an actual characteristic data set of the electricity consumption of the experimental household appliance by using the non-invasive terminal, inputting the actual characteristic data set into a classifier trained once to obtain an actual equipment class set of the experimental household appliance until the identification performance of the classifier on the household appliance is optimal, and obtaining the optimized classifier;
and the final actual equipment category obtaining module is used for obtaining the final actual equipment category of each household appliance of the target family by using the optimized classifier through the non-invasive terminal.
The experimental characteristic data set of the household electrical appliance electricity consumption is f ═ Irms,Urms,S,P,Q,φ,PF,FW,HAR2nd,HAR3rd,…,HAR19th};
Wherein f is an experimental characteristic data set of household appliance power utilization, IrmsIs the effective value of the current, UrmsIs the effective value of voltage, S is admittance, P is active power, Q is reactive power, phi is phase angle, PF is power factor, FM is fundamental wave, HAR2ndIs the 2 nd harmonic, HAR3rdIs the 3 rd harmonic, HAR19thIs the 19 th harmonic.
The actual sample data set constructing module specifically comprises:
the actual label set obtaining submodule is used for labeling the actual characteristic data set according to the actual equipment category set and the Internet of things label of the experimental household appliance to obtain an actual label set of the experimental household appliance;
and the actual sample data set forming submodule is used for forming an actual sample data set of the experimental household appliance by the actual characteristic data set and the actual label set.
The system further comprises:
the evaluation module is used for evaluating the secondarily trained classifier by taking the accuracy and the F-score as evaluation indexes;
and the downloading module is used for downloading the secondarily trained classifier with the accuracy rate larger than or equal to the accuracy rate threshold and the F-score larger than or equal to the F score threshold to the non-invasive terminal as a new primarily trained classifier.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (8)
1. A non-invasive load identification method is applied to a non-invasive load monitoring system, and comprises the following steps:
acquiring an experiment sample data set of the household appliances of the experiment family through a non-invasive cloud; the experimental sample data set comprises an experimental characteristic data set of the power consumption of the household appliance and an experimental equipment category set of the household appliance;
training a classifier by using the experimental sample data set to obtain a once-trained classifier, and downloading the once-trained classifier to a non-invasive terminal;
acquiring an actual characteristic data set of the electricity consumption of the experimental household appliance by using the non-invasive terminal, and inputting the actual characteristic data set into the classifier trained at one time to obtain an actual equipment class set of the experimental household appliance;
forming an actual sample data set of the experimental household appliance by the actual characteristic data set and the actual equipment category set, and transmitting the actual sample data set to the non-invasive cloud;
merging the experimental sample data set and the actual sample data set at the non-invasive cloud end, and training the once-trained classifier by using the merged sample data set to obtain a secondarily-trained classifier;
downloading the secondarily trained classifier as a new primarily trained classifier to a non-invasive terminal;
sliding the experiment sample data set forward in the combined sample data set by a preset sample number to obtain a new experiment sample data set, returning to the step of obtaining an actual characteristic data set of the electricity consumption of the experimental household appliance by using the non-invasive terminal, inputting the actual characteristic data set into the classifier trained at one time to obtain an actual equipment classification set of the experimental household appliance until the identification performance of the classifier on the household appliance reaches the optimum, and obtaining the optimized classifier;
and obtaining the final actual equipment category of each household appliance of the target family by using the optimized classifier through the non-invasive terminal.
2. The non-invasive load identification method according to claim 1, wherein the experimental characteristic dataset of the power consumption of the household appliance is f ═ { I ═ Irms,Urms,S,P,Q,φ,PF,FW,HAR2nd,HAR3rd,…,HAR19th};
Wherein f is an experimental characteristic data set of household appliance power utilization, IrmsIs the effective value of the current, UrmsIs the effective value of voltage, S is admittance, P is active power, Q is reactive power, phi is phase angle, PF is power factor, FM is fundamental wave, HAR2ndIs the 2 nd harmonic, HAR3rdIs the 3 rd harmonic, HAR19thIs the 19 th harmonic.
3. The non-invasive load recognition method according to claim 1, wherein the step of forming the actual feature data set and the actual device class set into an actual sample data set of the experimental home appliance specifically comprises:
labeling the actual characteristic data set according to the actual equipment category set and the Internet of things label of the experimental household appliance to obtain an actual label set of the experimental household appliance;
and the actual characteristic data set and the actual label set form an actual sample data set of the experimental household appliance.
4. The method according to claim 1, wherein the non-invasive cloud is configured to merge the experimental sample data set and the actual sample data set, and train the once-trained classifier using the merged sample data set to obtain a twice-trained classifier, and then further comprising:
evaluating the secondarily trained classifier by taking the accuracy and the F-score as evaluation indexes;
and downloading the secondarily trained classifier with the accuracy rate greater than or equal to the accuracy rate threshold and the F-score greater than or equal to the F score threshold as a new primarily trained classifier to the non-invasive terminal.
5. A non-intrusive load identification system, the non-intrusive load identification system being applied to a non-intrusive load monitoring system, the system comprising:
the experimental sample data set acquisition module is used for acquiring an experimental sample data set of the experimental household appliance through a non-invasive cloud; the experimental sample data set comprises an experimental characteristic data set of the power consumption of the household appliance and an experimental equipment category set of the household appliance;
the classifier obtaining module is used for training the classifier by using the experimental sample data set to obtain a trained classifier, and downloading the trained classifier to a non-invasive terminal;
the experimental household appliance actual equipment category set obtaining module is used for obtaining an actual characteristic data set of the experimental household appliance electricity utilization by utilizing the non-invasive terminal, inputting the actual characteristic data set into the classifier which is trained at one time, and obtaining an actual equipment category set of the experimental household appliance;
the actual sample data set forming module is used for forming an actual sample data set of the experimental household appliance by the actual characteristic data set and the actual equipment category set and transmitting the actual sample data set to the non-invasive cloud terminal;
the classifier obtaining module which is trained for the second time is used for merging the experimental sample data set and the actual sample data set at the non-invasive cloud end, and training the classifier which is trained for the first time by using the merged sample data set to obtain a classifier which is trained for the second time;
the classifier replacing module is used for downloading the secondarily trained classifier serving as a new primarily trained classifier to a non-invasive terminal;
the optimized classifier obtaining module is used for sliding the experiment sample data set forward in the combined sample data set by the preset number of samples to obtain a new experiment sample data set, and returning to the step of obtaining an actual characteristic data set of the electricity consumption of the experimental household appliance by using the non-invasive terminal, inputting the actual characteristic data set into the classifier trained at one time to obtain an actual equipment classification set of the experimental household appliance until the identification performance of the classifier on the household appliance is optimal, and obtaining the optimized classifier;
and the final actual equipment category obtaining module is used for obtaining the final actual equipment category of each household appliance of the target family by using the optimized classifier through the non-invasive terminal.
6. The non-invasive load recognition system of claim 5, wherein the experimental characteristic dataset of appliance power usage is f ═ Irms,Urms,S,P,Q,φ,PF,FW,HAR2nd,HAR3rd,…,HAR19th};
Wherein f is an experimental characteristic data set of household appliance power utilization, IrmsIs the effective value of the current, UrmsIs the effective value of voltage, S is admittance, P is active power, Q is reactive power, phi is phase angle, PF is power factor, FM is fundamental wave, HAR2ndIs the 2 nd harmonic, HAR3rdIs the 3 rd harmonic, HAR19thIs the 19 th harmonic.
7. The non-intrusive load recognition system of claim 5, wherein the actual sample data set comprises a module, specifically comprising:
the actual label set obtaining submodule is used for labeling the actual characteristic data set according to the actual equipment category set and the Internet of things label of the experimental household appliance to obtain an actual label set of the experimental household appliance;
and the actual sample data set forming submodule is used for forming an actual sample data set of the experimental household appliance by the actual characteristic data set and the actual label set.
8. The non-intrusive load recognition system of claim 5, further comprising:
the evaluation module is used for evaluating the secondarily trained classifier by taking the accuracy and the F-score as evaluation indexes;
and the downloading module is used for downloading the secondarily trained classifier with the accuracy rate larger than or equal to the accuracy rate threshold and the F-score larger than or equal to the F score threshold to the non-invasive terminal as a new primarily trained classifier.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110861575.2A CN113569952A (en) | 2021-07-29 | 2021-07-29 | Non-invasive load identification method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110861575.2A CN113569952A (en) | 2021-07-29 | 2021-07-29 | Non-invasive load identification method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113569952A true CN113569952A (en) | 2021-10-29 |
Family
ID=78168824
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110861575.2A Pending CN113569952A (en) | 2021-07-29 | 2021-07-29 | Non-invasive load identification method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113569952A (en) |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106790248A (en) * | 2017-01-23 | 2017-05-31 | 中南大学 | A kind of network inbreak detection method based on the online extreme learning machine of double adaptive regularization |
CN109739844A (en) * | 2018-12-26 | 2019-05-10 | 西安电子科技大学 | Data classification method based on decaying weight |
CN110119545A (en) * | 2019-04-24 | 2019-08-13 | 西安理工大学 | A kind of non-intrusive electrical load recognition methods based on stack self-encoding encoder |
CN110376457A (en) * | 2019-06-28 | 2019-10-25 | 同济大学 | Non-intrusion type load monitoring method and device based on semi-supervised learning algorithm |
CN110445126A (en) * | 2019-06-25 | 2019-11-12 | 中国电力科学研究院有限公司 | A kind of non-intrusion type load decomposition method and system |
GB202101718D0 (en) * | 2018-06-22 | 2021-03-24 | Moixa Energy Holdings Ltd | Systems for machine learning, optimising and managing local multi-asset flexibility of distributed energy storage resources-divisional |
US10983496B1 (en) * | 2020-10-13 | 2021-04-20 | King Abdulaziz University | Apparatus and method for non-intrusive load monitoring aided hybrid high and low frequency approaches |
WO2021101630A1 (en) * | 2019-11-24 | 2021-05-27 | Purdue Research Foundation | Intelligent non-intrusive load monitoring system |
WO2021107422A1 (en) * | 2019-11-28 | 2021-06-03 | 한국전자기술연구원 | Nonintrusive load monitoring method using energy consumption data |
WO2022025743A1 (en) * | 2020-07-29 | 2022-02-03 | Université Internationale de RABAT | Intelligent system for detecting and identifying appliances in operation by non-intrusive bimodal monitoring of the electrical signal |
-
2021
- 2021-07-29 CN CN202110861575.2A patent/CN113569952A/en active Pending
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106790248A (en) * | 2017-01-23 | 2017-05-31 | 中南大学 | A kind of network inbreak detection method based on the online extreme learning machine of double adaptive regularization |
GB202101718D0 (en) * | 2018-06-22 | 2021-03-24 | Moixa Energy Holdings Ltd | Systems for machine learning, optimising and managing local multi-asset flexibility of distributed energy storage resources-divisional |
CN109739844A (en) * | 2018-12-26 | 2019-05-10 | 西安电子科技大学 | Data classification method based on decaying weight |
CN110119545A (en) * | 2019-04-24 | 2019-08-13 | 西安理工大学 | A kind of non-intrusive electrical load recognition methods based on stack self-encoding encoder |
CN110445126A (en) * | 2019-06-25 | 2019-11-12 | 中国电力科学研究院有限公司 | A kind of non-intrusion type load decomposition method and system |
CN110376457A (en) * | 2019-06-28 | 2019-10-25 | 同济大学 | Non-intrusion type load monitoring method and device based on semi-supervised learning algorithm |
WO2021101630A1 (en) * | 2019-11-24 | 2021-05-27 | Purdue Research Foundation | Intelligent non-intrusive load monitoring system |
WO2021107422A1 (en) * | 2019-11-28 | 2021-06-03 | 한국전자기술연구원 | Nonintrusive load monitoring method using energy consumption data |
WO2022025743A1 (en) * | 2020-07-29 | 2022-02-03 | Université Internationale de RABAT | Intelligent system for detecting and identifying appliances in operation by non-intrusive bimodal monitoring of the electrical signal |
US10983496B1 (en) * | 2020-10-13 | 2021-04-20 | King Abdulaziz University | Apparatus and method for non-intrusive load monitoring aided hybrid high and low frequency approaches |
Non-Patent Citations (3)
Title |
---|
HIMEUR, Y.; ALSALEMI, A.; (...); DIMITRAKOPOULOS, G.: "On the applicability of 2D local binary patterns for identifying electrical appliances in non-intrusive load monitoring", INTELLIGENT SYSTEMS AND APPLICATIONS. 2020 INTELLIGENT SYSTEMS CONFERENCE (INTELLISYS) 2021, 25 August 2020 (2020-08-25), pages 188 - 205 * |
WU, X等: "Multi-Label Classification Based on Random Forest Algorithm for Non-Intrusive Load Monitoring System", 《 PROCESSES》, vol. 7, no. 6, 30 June 2019 (2019-06-30), pages 1 - 14 * |
谭啸风等: "基于数据流的非侵入式负荷识别与分类", 《电源技术》, vol. 40, no. 05, 20 May 2016 (2016-05-20), pages 1110 - 1112 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109145949A (en) | Non-intrusive electrical load monitoring and decomposition method and system based on integrated study | |
CN112434799B (en) | Non-invasive load identification method based on full convolution neural network | |
CN113723844B (en) | Low-voltage station theoretical line loss calculation method based on ensemble learning | |
CN113554361B (en) | Comprehensive energy system data processing and calculating method and processing system | |
CN111563827A (en) | Load decomposition method based on electrical appliance physical characteristics and residential electricity consumption behaviors | |
CN115170000A (en) | Remote monitoring method and system based on electric energy meter communication module | |
CN113780684A (en) | Intelligent building user energy consumption behavior prediction method based on LSTM neural network | |
CN115130830B (en) | Non-invasive load decomposition method based on cascade width learning and sparrow algorithm | |
Han et al. | Non‐intrusive load monitoring by voltage–current trajectory enabled asymmetric deep supervised hashing | |
Li et al. | Load profile inpainting for missing load data restoration and baseline estimation | |
CN108830499A (en) | A kind of non-intrusion type load decomposition method and system | |
Liu et al. | Enhanced NILM load pattern extraction via variable-length motif discovery | |
CN116681186A (en) | Power quality analysis method and device based on intelligent terminal | |
CN113569952A (en) | Non-invasive load identification method and system | |
CN111239484A (en) | Non-invasive load electricity consumption information acquisition method for non-resident users | |
CN116561569A (en) | Industrial power load identification method based on EO feature selection and AdaBoost algorithm | |
CN115841338A (en) | Method and device for determining abnormal electricity utilization behavior and non-volatile storage medium | |
CN115687999A (en) | Non-invasive load sensing method fusing multi-time scale electric appliance characteristics | |
CN115564046A (en) | Non-invasive load monitoring method based on multitask learning | |
CN115622034A (en) | Internet of things meter power load identification method based on improved genetic neural network | |
Li et al. | An snn ontology based environment monitoring method for intelligent irrigation system | |
Jürgel et al. | Non-Intrusive Load Monitoring in the OpenHAB smart home framework | |
Hao et al. | Active Sensing for Transformer Model in Sparse Mobile CrowdSensing | |
CN113762619B (en) | Distribution Internet of things user load identification method, system, equipment and storage medium | |
CN117833243B (en) | Method, system, equipment and medium for predicting short-term demand of electric power |
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 |