CN111027408A - A Load Identification Method Based on Support Vector Machine and V-I Curve Features - Google Patents

A Load Identification Method Based on Support Vector Machine and V-I Curve Features Download PDF

Info

Publication number
CN111027408A
CN111027408A CN201911133730.8A CN201911133730A CN111027408A CN 111027408 A CN111027408 A CN 111027408A CN 201911133730 A CN201911133730 A CN 201911133730A CN 111027408 A CN111027408 A CN 111027408A
Authority
CN
China
Prior art keywords
load
curve
event
harmonic
voltage
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
Application number
CN201911133730.8A
Other languages
Chinese (zh)
Inventor
谢雄威
蒋雯倩
梁捷
杨舟
卿柏元
李刚
韦杏秋
李金瑾
陈珏羽
林秀清
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.)
Electric Power Research Institute of Guangxi Power Grid Co Ltd
Original Assignee
Electric Power Research Institute of Guangxi Power Grid Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Electric Power Research Institute of Guangxi Power Grid Co Ltd filed Critical Electric Power Research Institute of Guangxi Power Grid Co Ltd
Priority to CN201911133730.8A priority Critical patent/CN111027408A/en
Publication of CN111027408A publication Critical patent/CN111027408A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Landscapes

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

Abstract

本发明公开了一种基于支持向量机和V‑I曲线特征的负荷识别方法，涉及负荷识别系统技术领域，利用V‑I曲线与谐波组合作为负荷印记克服了非侵入式负荷识别中可能产生的误辨识的缺点；运用V‑I曲线轨迹的形状特征形成多个负荷印记，增加了负荷辨识的正确率，并辅以电气量经傅里叶变换得出的谐波特征辨识，克服了V‑I曲线不好识别小型负荷的缺点；通过基于SVM的负荷识别更好地处理非线性分类问题，且不会引起“维数灾难”；它可以处理小样本的机器学习，且不至于陷入局部最小以及过学习、欠学习问题，使得负荷识别的结果更具有准确性，提高辨识度。此外，本发明方法以非侵入式为出发点，具有经济实用、容易实现的特点。

Figure 201911133730

The invention discloses a load identification method based on support vector machine and V-I curve characteristics, and relates to the technical field of load identification systems. The combination of V-I curve and harmonics is used as a load imprint to overcome the possibility of occurrence in non-intrusive load identification. The disadvantage of misidentification; using the shape features of the V-I curve trajectory to form multiple load marks, which increases the correct rate of load identification, and supplemented by the harmonic feature identification obtained by the Fourier transform of the electrical quantity, which overcomes the V-I curve. ‑I curve is not good at identifying the disadvantage of small loads; it can better handle nonlinear classification problems through SVM-based load identification without causing the "curse of dimensionality"; it can handle small-sample machine learning without falling into local The minimum and over-learning and under-learning problems make the result of load identification more accurate and improve the recognition degree. In addition, the method of the present invention takes the non-invasive method as the starting point, and has the characteristics of economy, practicality and easy realization.

Figure 201911133730

Description

Load identification method based on support vector machine and V-I curve characteristics
Technical Field
The invention relates to the technical field of load identification systems, in particular to a load identification method based on a support vector machine and V-I curve characteristics.
Background
With the increasing maturity of power system networks and the rapid development of artificial intelligence, smart power grids are growing slowly and rapidly along with the combination of the two. In future intelligent power grid planning, the intelligent power grid planning system is developed towards a fully-automatic power transmission network, and has the capability of monitoring and controlling each user and power grid node and ensuring the bidirectional flow of information and electric energy between all nodes in the whole power transmission and distribution process from a power plant to an end user. Therefore, it is required that good interaction is formed between the grid terminals and the users, and better power management and service are realized. With the Non-intrusive Load monitoring system (nims), Load identification is performed to obtain the real-time power consumption of the user on the premise of not intruding the internal equipment of the user. Through counting the electricity consumption of different loads, a user can know the detailed electricity consumption condition and effectively manage electricity consumption behaviors.
Among them, the extraction of Load Signature (LS) and a good Load recognition classifier are key links for determining the accuracy of Load recognition. At present, scholars at home and abroad develop a series of researches on load characteristic extraction based on steady state and transient state, wherein the steady state characteristics comprise power, V-I waveform, voltage noise, current harmonic wave and the like; transient characteristics include instantaneous voltage, instantaneous power, voltage noise, current, etc. With the continuous development of machine learning, classifiers including linear classifiers, support vector machines, neural networks, etc. are also widely used in load recognition. However, because the single load characteristics of different types of loads often overlap, if the result obtained by the non-intrusive load identification only depending on the single load characteristics is often unreliable, the result is greatly influenced by the false identification.
Disclosure of Invention
The invention aims to provide a load identification method based on a support vector machine and V-I curve characteristics, thereby overcoming the defect that the result obtained by the existing non-invasive load identification only depending on single load characteristics is often unreliable.
In order to achieve the above object, the present invention provides a load identification method based on support vector machine and V-I curve characteristics, comprising the following steps:
s1, acquiring power utilization data, and preprocessing the power utilization data;
s2, judging whether an event occurs in the preprocessed data of S1; if the event occurs, the step enters S3, otherwise, the step enters S1;
s3, performing feature extraction and combination on the voltage and current signals after the event occurs by adopting a V-I curve extraction method and a harmonic feature extraction method to obtain load features;
and S4, taking the load mark obtained in S3 as a load characteristic, carrying out load identification based on an SVM on the event obtained in S2, and identifying to obtain the electric appliance in the working state in the family of the user.
Further, the electricity consumption data includes: current, voltage, and power.
Further, in S2, it is determined that an event occurs according to the change of the effective value of the power, and when the change of the effective value of the power is greater than a threshold, an event occurs, otherwise, no event occurs.
Further, the judging method includes:
setting the apparent power resulting from the preprocessing of S1: s1，…，St，St+1…; event start time tonT seconds, event end time toffT + TL seconds; the step length of each movement of the event detection window is L;
the total apparent power change Δ St＝St+1-St，StTotal apparent power at t seconds;
when Δ St＞Son1At that time, the event detection window starts to move and Δ S is calculatedt+1，ΔSt+2…, up to Δ St+TL＜Son1，Son1Detecting a power change threshold for an event, Son2The minimum event power change value which can be detected;
if St+TL-St＜Son2If the load is changed in the t-t + TL seconds, no event occurs.
Further, the method for extracting the characteristics of the voltage and current signals after the event occurs by using the V-I curve extraction method comprises the following steps:
firstly, smoothing and interpolating voltage and current waveforms within T seconds before and after an event;
then, within T seconds, taking a period of voltage and current waveform every second, carrying out Fourier transform on the voltage waveform, and then taking a point with a fundamental wave voltage phase angle of 0 as an initial sampling point of the voltage and current waveform;
averaging sampling points at the same positions of each period of the voltage and current waveform, and drawing a V-I curve by taking the voltage as an abscissa and the current as an ordinate;
finally, the characteristics of the V-I curve are used as load marks.
Further, the method for extracting the characteristics of the voltage and current signals after the occurrence of the event by using the harmonic characteristic extraction method comprises the following steps: converting the current signal in the time domain into a frequency spectrum signal in the frequency domain through fast Fourier transform, wherein the frequency spectrum signal is shown as a formula (1);
Figure BDA0002279024160000031
in the formula (1), i0Is a direct current component, ikIs the kth harmonic current amplitude, k ω is the kth harmonic component angular frequency,
Figure BDA0002279024160000032
is the initial phase angle of the kth harmonic component; and extracting harmonic components from the spectrum signal to obtain the characteristic information of the load equipment on the frequency domain, and taking the frequency domain characteristic as a load mark.
Further, the harmonic component includes: harmonic order and amplitude.
Further, the harmonic amplitude is the amplitude of the sixth harmonic in the third and fifth harmonic orders.
Further, the S4 includes the following steps:
s41, given the input data and the learning objective: x ═ X1,X2,…,X8}，y＝{y1,y2,…,yNIn which yiIndicating that the identification result i is 1,2, …,8, and N is the number of the electrical appliances; if a hyperplane H serving as a decision boundary exists in a feature space where input data are located, the hyperplane H separates the input data according to a positive class and a negative class, and the distance from a point of any sample to the hyperplane H is greater than or equal to 1, then the classification problem is shown as a formula (2):
Figure BDA0002279024160000041
in the formula (2), omega and b are respectively a normal vector and an intercept of the hyperplane;
s42, optimally classifying the hyperplanes to obtain the hyperplane with the largest classification interval, and simplifying the classification problem into the following optimization problem:
Figure BDA0002279024160000042
solving the formula (3) by using a Lagrange multiplier method, and introducing a Lagrange multiplier α by using omega and b as variablesiNot less than 0, i-1, ·, l yields:
Figure BDA0002279024160000043
converting the problem described by formula (3) to a dual form:
Figure BDA0002279024160000044
Figure BDA0002279024160000045
yi(ω·xi+b)-1≥0,i＝1,···,l (7)
αi≥0,i＝1,···,l (8)
αi(yi(ω·xi+b)-1)＝0,i＝1,···,l (9)
in formulae (5) and (6), LpAn objective function that is a dual form of the above optimization problem;
the following formulas (5), (6), (7), (8) and (9) are simplified:
Figure BDA0002279024160000046
s43, the dual form has the same optimal point as the original optimization problem, so the original optimization problem is converted into:
Figure BDA0002279024160000051
s44, calculating according to the formula (11) to obtain the normal vector of the optimal classification hyperplane:
Figure BDA0002279024160000052
the final discriminant function is:
Figure BDA0002279024160000053
in formula (13), z is the distance from the sample point to the hyperplane; namely, obtaining the classification result of the SVM classifier through f (z).
Compared with the prior art, the invention has the following beneficial effects:
the load identification method based on the support vector machine and the V-I curve characteristics provided by the invention overcomes the defect of false identification possibly generated in non-invasive load identification by using the combination of the V-I curve and the harmonic wave as a load mark; the shape characteristics of the V-I curve track are used for forming a plurality of load marks, the accuracy of load identification is increased, and the harmonic characteristic identification obtained by Fourier transform of the electric quantity is used for identification, so that the defect that the V-I curve cannot identify the small load well is overcome; nonlinear classification problems are better handled through SVM-based load recognition, and dimension disasters are not caused; the method can process machine learning of small samples, and can not fall into the problems of local minimum, over-learning and under-learning, so that the result of load identification (the electrical appliance in a working state in a user family) has higher accuracy, and the identification degree is improved. In addition, the method of the invention takes non-invasive as a starting point and has the characteristics of economy, practicability and easy realization.
Drawings
In order to more clearly illustrate the technical solution of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only one embodiment of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.
FIG. 1 is a flow chart of a load identification method based on a support vector machine and V-I curve characteristics according to the present invention;
FIG. 2 is a V-I plot of a piece of data in a user's home, in accordance with an embodiment of the present invention.
Detailed Description
The technical solutions in the present invention are 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.
As shown in fig. 1, the load identification method based on support vector machine and V-I curve feature provided by the present invention includes the following steps:
s1, acquiring power utilization data, and preprocessing the power utilization data; the electricity consumption data includes: and the current, the voltage, the power and the like are obtained from the intelligent electric meter installed at the user terminal. Noise exists in raw data (electricity consumption data) acquired from a smart meter at a home terminal, which affects extraction of a load mark, so that denoising processing needs to be performed on the raw data.
S2, judging whether an event occurs in the preprocessed data of S1; if there is an event occurring, S3 is entered, otherwise S1 is entered. The invention judges the occurrence of the event by the change of the effective value of the power, compares the change of the effective value of the power with a set threshold value, and if the change of the effective value of the power is greater than the threshold value, the event occurs. The specific judgment method is as follows:
setting the pre-processed apparent power of S1 (S) with the change of the apparent power during the load state transition process1，…，St，St+1…); event start time tonT seconds, event end time toffT + TL seconds; the step length of each movement of the event detection window is L, and the method is set to be L-1 s;
the total apparent power change Δ St＝St+1-St，StTotal apparent power at t seconds; when Δ St＞Son1At that time, the event detection window starts to move and Δ S is calculatedt+1，ΔSt+2…, up to Δ St+TL＜Son1，Son1Detecting a power change threshold for an event, Son2The minimum event power change value which can be detected; if S ist+TL-St＜Son2If the load is changed in the time range from t to t + TL, then TL represents the duration of the event, i.e. no event occurs.
And S3, performing feature extraction and combination on the voltage and current signals after the event occurs by adopting a V-I curve extraction method and a harmonic feature extraction method to obtain load features. S3 specifically includes the following steps:
s31, performing feature extraction on the voltage and current signals after the event occurs by adopting a V-I curve extraction method; firstly, smoothing and interpolating voltage and current waveforms within T seconds before and after an event; then, within T seconds, taking a period of voltage and current waveform every second, carrying out Fourier transform on the voltage waveform, and then taking a point with a fundamental wave voltage phase angle of 0 as an initial sampling point of the voltage and current waveform; averaging sampling points at the same positions of each period of the voltage and current waveform, and drawing a V-I curve by taking the voltage as an abscissa and the current as an ordinate; finally, the characteristics of the V-I curve are used as load marks.
There are different methods for shape characterization and classification of V-I curves, one of which is to use shape features to describe the shape, each shape feature on the V-I curve being shown in Table 1.
TABLE 1 characteristics of the respective shapes on the V-I curves
Figure BDA0002279024160000071
Figure BDA0002279024160000081
S32, performing feature extraction on the voltage and current signals after the event occurs by adopting a harmonic feature extraction method; the current signal in the time domain is converted into a frequency spectrum signal in the frequency domain through Fast Fourier Transform (FFT), as shown in formula (1), so that the characteristic information of the load equipment in the frequency domain is obtained, that is, the harmonic component is extracted.
Figure BDA0002279024160000082
In the formula (1), i0Is a direct current component, ikIs the kth harmonic current amplitude, k ω is the kth harmonic component angular frequency,
Figure BDA0002279024160000083
is the initial phase angle of the k harmonic component. Regarding the harmonic order and amplitude of the harmonic component, the existing research indicates that the amplitude of even harmonic generated when most loads operate is small, and the amplitude of odd harmonic is large, and the typical harmonic components are 2 th harmonic, 3 rd harmonic and 5 th harmonic. The invention selects harmonic amplitude X6Third harmonic X7Fifth harmonic X8As a load signature.
S33, because the characteristic of the V-I curve has the advantage of good differentiation, but the small loads cannot be well differentiated, and the harmonic characteristic has the advantages of simple measurement, anti-interference capability and capability of differentiating the small loads, the load marks obtained from S31 and S32 are used as load characteristics to identify the loads.
And S4, taking the load mark obtained in S3 as a load characteristic, carrying out SVM-based load recognition on the event obtained in S2, and recognizing the electric appliance in the working state in the family of the user.
The load characteristics can be classified through a linear classifier so as to obtain a recognition result. The invention selects the V-I curve and the harmonic wave characteristics as the input of a Support Vector Machine (SVM) classifier, and S4 comprises the following steps:
s41, given the input data and the learning objective: x ═ X1,X2,…,X8}，y＝{y1,y2,…,yNIn which yiIndicating the identification result (i is 1,2, …,8), wherein N is the number of the electrical appliances; if the feature space where the input data is located has a hyperplane H, omega as a decision boundary (decision boundary)TX + b is 0, and if the hyperplane is the learning target, the sum is positiveThe negative class separates the hyperplane H, and the distance from the point of any sample to the hyperplane H is greater than or equal to 1, then the classification problem for given input data and learning target is said to have linear separability, as shown in equation (2):
Figure BDA0002279024160000091
in the formula (2), ω and b are the normal vector and intercept of the hyperplane H, respectively.
S42, optimally classifying the hyperplane, namely obtaining the hyperplane with the largest classification interval, wherein the classification problem can be the following optimization problem:
Figure BDA0002279024160000092
solving the formula (3) by using a Lagrange multiplier method, and introducing a Lagrange multiplier α by using omega and b as variablesiNot less than 0, i-1, ·, l yields:
Figure BDA0002279024160000093
because the function is convex and the points meeting the constraints also form a convex set, the quadratic programming problem is a convex quadratic programming problem and does not have a local minimum, so the idea of the optimal hyperplane of the SVM method can overcome the local minimum problem; converting the problem described by formula (3) to a dual form:
Figure BDA0002279024160000094
Figure BDA0002279024160000095
yi(ω·xi+b)-1≥0,i＝1,···,l (7)
αi≥0,i＝1,···,l (8)
αi(yi(ω·xi+b)-1)＝0,i＝1,···,l (9)
in formulae (5) and (6), LpAn objective function that is a dual form of the above optimization problem;
the following formulas (5), (6), (7), (8) and (9) are simplified:
Figure BDA0002279024160000101
s43, the dual form has the same optimal point as the original optimization problem, so the original optimization problem is converted into:
Figure BDA0002279024160000102
s44, only a small part of the solutions of the formula (11) are not 0, and x isj(i ∈ SV) is the support vector (SV for short), and each α is solved by equation (11)i *Corresponding to one sample, calculating to obtain the normal vector of the optimal classification hyperplane:
Figure BDA0002279024160000103
the final discriminant function is:
Figure BDA0002279024160000104
in formula (13), z is the distance from the sample point to the hyperplane; namely, obtaining the classification result of the SVM classifier through f (z).
The embodiment of the load identification method based on the support vector machine and the V-I curve characteristics of the invention is explained in detail so that the person skilled in the art can understand the invention more:
the invention records the electricity utilization condition of 4 families within three days. According to the load power change condition in the low-frequency apparent power data, Son1Taking 30VA, Son2Taking 100VA, 1553 events are detected in total.
According to the step S3, T is respectively 3 and 5, the initial sampling point of the voltage waveform and the current waveform in each period is determined by the point with the phase angle of the fundamental wave voltage closest to 0, a V-I curve graph is drawn by taking the voltage as the abscissa and the current as the ordinate, and harmonic features are extracted through Fourier transform to carry out combination of the two. As shown in fig. 2, a piece of data in a user's family is actually extracted and plotted as a V-I curve.
And extracting the V-I curve and the harmonic load mark of the acquired user data, then taking the extracted user data as the input of the SVM classifier, and outputting a recognition result. 1553 events are detected in the invention, 1000 samples are selected for training, and the rest samples are used for testing. Non-intrusive load identification is performed on 8 electrical appliances in each of the 4 monitored households, and the accuracy of the identification result is shown in table 2:
TABLE 2 accuracy of recognition results of non-intrusive load recognition System
Figure BDA0002279024160000111
As can be seen from the table 2, the accuracy of the obtained result is improved by training and identifying based on the SVM classifier by using the combination of the V-I curve and the harmonic load imprint, and particularly, the accuracy of the identification of high-power electrical appliances is high, the accuracy of the identification of certain power fluctuation electrical appliances is also high, and the objective engineering requirements can be met.
In summary, the load identification method based on the support vector machine and the V-I curve characteristics overcomes the defect of false identification possibly generated in non-invasive load identification by using the combination of the V-I curve and the harmonic wave as the load imprints, and increases the accuracy of load identification by forming a plurality of load imprints by using the shape characteristics of the V-I curve track. And the harmonic characteristic identification obtained by Fourier transform of the electrical quantity is used as an auxiliary factor, so that the defect that the V-I curve is not good in identifying the small-sized load is overcome. The invention uses the support vector machine as a classifier and has a plurality of advantages which are not available in the traditional method: such as non-linear classification problems that can be handled better than non-linear classifiers and that do not cause "dimension disasters"; the method can process machine learning of small samples, and can not fall into local minimum and over-learning and under-learning problems. In addition, the method takes non-invasive as a starting point, and has the characteristics of economy, practicability and easy implementation.
The above disclosure is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of changes or modifications within the technical scope of the present invention, and shall be covered by the scope of the present invention.

Claims (9)

1. A load identification method based on a support vector machine and V-I curve characteristics is characterized in that: the method comprises the following steps:
s1, acquiring power utilization data, and preprocessing the power utilization data;
s2, judging whether an event occurs in the preprocessed data of S1; if the event occurs, the step enters S3, otherwise, the step enters S1;
s3, performing feature extraction and combination on the voltage and current signals after the event occurs by adopting a V-I curve extraction method and a harmonic feature extraction method to obtain load features;
and S4, taking the load mark obtained in S3 as a load characteristic, carrying out load identification based on an SVM on the event obtained in S2, and identifying to obtain the electric appliance in the working state in the family of the user.
2. The load identification method based on the support vector machine and the V-I curve features as claimed in claim 1, wherein: the electricity consumption data includes: current, voltage, and power.
3. The load identification method based on the support vector machine and the V-I curve features as claimed in claim 1, wherein: in S2, the occurrence of the event is determined by the change of the effective value of the power, and when the change of the effective value of the power is greater than the threshold, the event occurs, otherwise, the event does not occur.
4. The load identification method based on the support vector machine and the V-I curve features as claimed in claim 3, wherein: the judging method comprises the following steps:
setting the apparent power resulting from the preprocessing of S1: s1，…，St，St+1…; event start time tonT seconds, event end time toffT + TL seconds; the step length of each movement of the event detection window is L;
the total apparent power change Δ St＝St+1-St，StTotal apparent power at t seconds;
when Δ St＞Son1At that time, the event detection window starts to move and Δ S is calculatedt+1，ΔSt+2…, up to Δ St+TL＜Son1，Son1Detecting a power change threshold for an event, Son2The minimum event power change value which can be detected;
if St+TL-St＜Son2If the load is changed in the t-t + TL seconds, no event occurs.
5. The load identification method based on the support vector machine and the V-I curve features as claimed in claim 1, wherein: the method for extracting the characteristics of the voltage and current signals after the event occurs by adopting the V-I curve extraction method comprises the following steps:
firstly, smoothing and interpolating voltage and current waveforms within T seconds before and after an event;
then, within T seconds, taking a period of voltage and current waveform every second, carrying out Fourier transform on the voltage waveform, and then taking a point with a fundamental wave voltage phase angle of 0 as an initial sampling point of the voltage and current waveform;
averaging sampling points at the same positions of each period of the voltage and current waveform, and drawing a V-I curve by taking the voltage as an abscissa and the current as an ordinate;
finally, the characteristics of the V-I curve are used as load marks.
6. The load identification method based on the support vector machine and the V-I curve features as claimed in claim 1, wherein: the method for extracting the characteristics of the voltage and current signals after the event occurs by adopting the harmonic characteristic extraction method comprises the following steps: converting the current signal in the time domain into a frequency spectrum signal in the frequency domain through fast Fourier transform, wherein the frequency spectrum signal is shown as a formula (1);
Figure FDA0002279024150000021
in the formula (1), i0Is a direct current component, ikIs the kth harmonic current amplitude, k ω is the kth harmonic component angular frequency,
Figure FDA0002279024150000022
is the initial phase angle of the kth harmonic component; and extracting harmonic components from the spectrum signal to obtain the characteristic information of the load equipment on the frequency domain, and taking the frequency domain characteristic as a load mark.
7. The load identification method based on the support vector machine and the V-I curve features of claim 6, wherein: the harmonic components include: harmonic order and amplitude.
8. The load identification method based on the support vector machine and the V-I curve features of claim 7, wherein: the third and fifth harmonic orders, the amplitude of the harmonic being the amplitude of the sixth harmonic.
9. The load identification method based on the support vector machine and the V-I curve features as claimed in claim 1, wherein: the S4 includes the steps of:
s41, given the input data and the learning objective: x ═ X1,X2,…,X8}，y＝{y1,y2,…,yNIn which yiIndicating that the identification result i is 1,2, …,8, and N is the number of the electrical appliances; if the feature space where the input data is located has a hyperplane H as a decision boundary, the hyperplane H separates the input data according to the positive class and the negative class, and makes the distance from the point of any sample to the hyperplane H largeIf 1, the classification problem is shown in formula (2):
Figure FDA0002279024150000031
in the formula (2), omega and b are respectively a normal vector and an intercept of the hyperplane;
s42, optimally classifying the hyperplanes to obtain the hyperplane with the largest classification interval, and simplifying the classification problem into the following optimization problem:
Figure FDA0002279024150000032
solving the formula (3) by using a Lagrange multiplier method, and introducing a Lagrange multiplier α by using omega and b as variablesiNot less than 0, i-1, ·, l yields:
Figure FDA0002279024150000033
converting the problem described by formula (3) to a dual form:
Figure FDA0002279024150000034
Figure FDA0002279024150000035
yi(ω·xi+b)-1≥0,i＝1,···,l (7)
αi≥0,i＝1,···,l (8)
αi(yi(ω·xi+b)-1)＝0,i＝1,···,l (9)
in formulae (5) and (6), LpAn objective function that is a dual form of the above optimization problem;
the following formulas (5), (6), (7), (8) and (9) are simplified:
Figure FDA0002279024150000041
s43, the dual form has the same optimal point as the original optimization problem, so the original optimization problem is converted into:
Figure FDA0002279024150000042
s44, calculating according to the formula (11) to obtain the normal vector of the optimal classification hyperplane:
Figure FDA0002279024150000043
the final discriminant function is:
Figure FDA0002279024150000044
in formula (13), z is the distance from the sample point to the hyperplane; namely, obtaining the classification result of the SVM classifier through f (z).
CN201911133730.8A 2019-11-19 2019-11-19 A Load Identification Method Based on Support Vector Machine and V-I Curve Features Pending CN111027408A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911133730.8A CN111027408A (en) 2019-11-19 2019-11-19 A Load Identification Method Based on Support Vector Machine and V-I Curve Features

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911133730.8A CN111027408A (en) 2019-11-19 2019-11-19 A Load Identification Method Based on Support Vector Machine and V-I Curve Features

Publications (1)

Publication Number Publication Date
CN111027408A true CN111027408A (en) 2020-04-17

Family

ID=70200565

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911133730.8A Pending CN111027408A (en) 2019-11-19 2019-11-19 A Load Identification Method Based on Support Vector Machine and V-I Curve Features

Country Status (1)

Country Link
CN (1) CN111027408A (en)

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111830347A (en) * 2020-07-17 2020-10-27 四川大学 An event-based two-stage non-intrusive load monitoring method
CN112067043A (en) * 2020-08-14 2020-12-11 常州机电职业技术学院 Defective degree detecting system of timber structure ancient building
CN112101110A (en) * 2020-08-13 2020-12-18 西安理工大学 Non-invasive load identification method for user side of power system
CN112180193A (en) * 2020-09-28 2021-01-05 华中科技大学 Non-invasive load identification system and method based on track image identification
CN112924779A (en) * 2021-01-25 2021-06-08 河北科技大学 Non-invasive load identification method and device and terminal equipment
CN113191253A (en) * 2021-04-28 2021-07-30 浙江大学 Non-invasive load identification method based on feature fusion under edge machine learning
CN113408341A (en) * 2021-05-10 2021-09-17 深圳供电局有限公司 Load identification method and device, computer equipment and storage medium
CN113901973A (en) * 2021-09-29 2022-01-07 北京智芯微电子科技有限公司 Power load identification method and device, storage medium and chip equipment
CN114236234A (en) * 2021-11-17 2022-03-25 国网天津市电力公司 Electrical appliance characteristic identification method based on fundamental wave and harmonic wave mixed criterion
CN114325081A (en) * 2021-12-29 2022-04-12 润建股份有限公司 Non-invasive load identification method based on multi-modal characteristics
CN114399098A (en) * 2021-12-30 2022-04-26 昆明能讯科技有限责任公司 High-applicability and high-precision industrial power utilization user classification peak-shifting power utilization method
CN114527344A (en) * 2022-02-25 2022-05-24 重庆和航科技股份有限公司 Intelligent identification method for load characteristics of electrical equipment
CN116559575A (en) * 2023-07-07 2023-08-08 国网江苏省电力有限公司常州供电分公司 Load event detection method and device
CN117272018A (en) * 2023-08-29 2023-12-22 广西电网有限责任公司 A load identification method based on spectrum decomposition
CN118340461A (en) * 2024-05-14 2024-07-16 苏州精源创智能科技有限公司 Hard floor and carpet recognition method for sweeping robot based on binary classification algorithm

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107273920A (en) * 2017-05-27 2017-10-20 西安交通大学 A kind of non-intrusion type household electrical appliance recognition methods based on random forest
CN107330517A (en) * 2017-06-14 2017-11-07 华北电力大学 One kind is based on S_Kohonen non-intrusion type resident load recognition methods

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107273920A (en) * 2017-05-27 2017-10-20 西安交通大学 A kind of non-intrusion type household electrical appliance recognition methods based on random forest
CN107330517A (en) * 2017-06-14 2017-11-07 华北电力大学 One kind is based on S_Kohonen non-intrusion type resident load recognition methods

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
丁世飞: "《高级人工智能》", 31 January 2015, 中国矿业大学出版社 *
娄藕蝶: "非侵入式负荷特征提取与识别研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 *
陈小闽: "非侵入式负荷分解技术及应用研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *

Cited By (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111830347A (en) * 2020-07-17 2020-10-27 四川大学 An event-based two-stage non-intrusive load monitoring method
CN112101110B (en) * 2020-08-13 2024-02-23 河北经纬九方科技有限公司 A method for non-intrusive load identification on the user side of power systems
CN112101110A (en) * 2020-08-13 2020-12-18 西安理工大学 Non-invasive load identification method for user side of power system
CN112067043A (en) * 2020-08-14 2020-12-11 常州机电职业技术学院 Defective degree detecting system of timber structure ancient building
CN112180193A (en) * 2020-09-28 2021-01-05 华中科技大学 Non-invasive load identification system and method based on track image identification
CN112180193B (en) * 2020-09-28 2021-11-02 华中科技大学 A non-invasive load recognition system and method based on trajectory image recognition
CN112924779A (en) * 2021-01-25 2021-06-08 河北科技大学 Non-invasive load identification method and device and terminal equipment
CN113191253A (en) * 2021-04-28 2021-07-30 浙江大学 Non-invasive load identification method based on feature fusion under edge machine learning
CN113408341A (en) * 2021-05-10 2021-09-17 深圳供电局有限公司 Load identification method and device, computer equipment and storage medium
CN113901973A (en) * 2021-09-29 2022-01-07 北京智芯微电子科技有限公司 Power load identification method and device, storage medium and chip equipment
CN114236234A (en) * 2021-11-17 2022-03-25 国网天津市电力公司 Electrical appliance characteristic identification method based on fundamental wave and harmonic wave mixed criterion
CN114325081A (en) * 2021-12-29 2022-04-12 润建股份有限公司 Non-invasive load identification method based on multi-modal characteristics
CN114399098A (en) * 2021-12-30 2022-04-26 昆明能讯科技有限责任公司 High-applicability and high-precision industrial power utilization user classification peak-shifting power utilization method
CN114527344A (en) * 2022-02-25 2022-05-24 重庆和航科技股份有限公司 Intelligent identification method for load characteristics of electrical equipment
CN116559575A (en) * 2023-07-07 2023-08-08 国网江苏省电力有限公司常州供电分公司 Load event detection method and device
CN116559575B (en) * 2023-07-07 2023-11-24 国网江苏省电力有限公司常州供电分公司 Load event detection method and device
CN117272018A (en) * 2023-08-29 2023-12-22 广西电网有限责任公司 A load identification method based on spectrum decomposition
CN118340461A (en) * 2024-05-14 2024-07-16 苏州精源创智能科技有限公司 Hard floor and carpet recognition method for sweeping robot based on binary classification algorithm

Similar Documents

Publication Publication Date Title
CN111027408A (en) A Load Identification Method Based on Support Vector Machine and V-I Curve Features
CN111830347B (en) Two-stage non-invasive load monitoring method based on event
Liu et al. Dynamic time warping based non-intrusive load transient identification
Yadav et al. Real-time multiple event detection and classification in power system using signal energy transformations
CN106786534B (en) Non-invasive power load transient process identification method and system
CN109598451B (en) Non-invasive load identification method based on PCA (principal component analysis) and LSTM (least Square TM) neural network
CN109387712A (en) Non-intrusion type cutting load testing and decomposition method based on state matrix decision tree
Zhu et al. Networked time series shapelet learning for power system transient stability assessment
CN114236234A (en) Electrical appliance characteristic identification method based on fundamental wave and harmonic wave mixed criterion
CN105447502A (en) Transient power disturbance identification method based on S conversion and improved SVM algorithm
CN112101110A (en) Non-invasive load identification method for user side of power system
CN106908671A (en) A kind of non-intrusion type household loads intelligent detecting method and system
Wu et al. Nonintrusive on-site load-monitoring method with self-adaption
Fan et al. Post-fault transient stability assessment based on k-nearest neighbor algorithm with Mahalanobis distance
Himeur et al. Appliance identification using a histogram post-processing of 2d local binary patterns for smart grid applications
CN117688426A (en) A method for extracting power quality fluctuation periods based on probability statistics
Li et al. Stacking ensemble learning-based load identification considering feature fusion by cyber-physical approach
Gao et al. Non-intrusive multi-label load monitoring via transfer and contrastive learning architecture
CN119557765A (en) A non-intrusive load identification method, system, medium and device for electrical equipment
Gurbuz et al. A brief review of non-intrusive load monitoring and its impact on social life
CN111639583A (en) A method and system for identifying power quality disturbances in power grids
CN114325081A (en) Non-invasive load identification method based on multi-modal characteristics
Li et al. The investigation of residential load identification technologies
Zhang et al. Improved YOLO algorithm for identifying abnormal states in electric energy metering devices
CN116628620A (en) A Non-Intrusive Load Identification Calculation Method

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20200417