CN111273131A - Photovoltaic grid-connected power generation island detection method based on energy characteristics and random forest - Google Patents
Photovoltaic grid-connected power generation island detection method based on energy characteristics and random forest Download PDFInfo
- Publication number
- CN111273131A CN111273131A CN202010188667.4A CN202010188667A CN111273131A CN 111273131 A CN111273131 A CN 111273131A CN 202010188667 A CN202010188667 A CN 202010188667A CN 111273131 A CN111273131 A CN 111273131A
- Authority
- CN
- China
- Prior art keywords
- power generation
- connected power
- photovoltaic grid
- wavelet packet
- random forest
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 59
- 238000010248 power generation Methods 0.000 title claims abstract description 47
- 238000007637 random forest analysis Methods 0.000 title claims abstract description 43
- 238000003066 decision tree Methods 0.000 claims abstract description 37
- 238000000034 method Methods 0.000 claims abstract description 30
- 238000005070 sampling Methods 0.000 claims abstract description 25
- 238000004088 simulation Methods 0.000 claims abstract description 22
- 239000013598 vector Substances 0.000 claims abstract description 19
- 238000012545 processing Methods 0.000 claims abstract description 9
- 238000004458 analytical method Methods 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 18
- 238000000354 decomposition reaction Methods 0.000 claims description 17
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 7
- 238000012795 verification Methods 0.000 claims description 4
- 101100257981 Arabidopsis thaliana S-ACP-DES3 gene Proteins 0.000 claims description 3
- 101150057677 DAD3 gene Proteins 0.000 claims description 3
- 102100034004 Gamma-adducin Human genes 0.000 claims description 3
- 101000799011 Homo sapiens Gamma-adducin Proteins 0.000 claims description 3
- 101001080624 Homo sapiens Proline/serine-rich coiled-coil protein 1 Proteins 0.000 claims description 3
- 101000652684 Homo sapiens Transcriptional adapter 3 Proteins 0.000 claims description 3
- 102100027427 Proline/serine-rich coiled-coil protein 1 Human genes 0.000 claims description 3
- 101100433696 Saccharomyces cerevisiae (strain ATCC 204508 / S288c) AAD3 gene Proteins 0.000 claims description 3
- 102100030836 Transcriptional adapter 3 Human genes 0.000 claims description 3
- 238000013145 classification model Methods 0.000 claims description 3
- 208000036552 dowling-degos disease 3 Diseases 0.000 claims description 3
- 208000002153 familial abdominal 3 aortic aneurysm Diseases 0.000 claims description 3
- 238000013138 pruning Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 230000007547 defect Effects 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 150000001875 compounds Chemical class 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000002955 isolation Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000002238 attenuated effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/081—Locating faults in cables, transmission lines, or networks according to type of conductors
- G01R31/086—Locating faults in cables, transmission lines, or networks according to type of conductors in power transmission or distribution networks, i.e. with interconnected conductors
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/088—Aspects of digital computing
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02S—GENERATION OF ELECTRIC POWER BY CONVERSION OF INFRARED RADIATION, VISIBLE LIGHT OR ULTRAVIOLET LIGHT, e.g. USING PHOTOVOLTAIC [PV] MODULES
- H02S50/00—Monitoring or testing of PV systems, e.g. load balancing or fault identification
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02B—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO BUILDINGS, e.g. HOUSING, HOUSE APPLIANCES OR RELATED END-USER APPLICATIONS
- Y02B10/00—Integration of renewable energy sources in buildings
- Y02B10/10—Photovoltaic [PV]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/50—Photovoltaic [PV] energy
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a photovoltaic grid-connected power generation island detection method based on energy characteristics and random forests, which comprises the following steps: step 1: the method comprises the steps that a sampling device is arranged in a photovoltaic grid-connected power generation system, and a PCC voltage signal is collected when the photovoltaic grid-connected power generation system is in an island operation state; step 2: carrying out wavelet packet analysis processing on the PCC voltage signal obtained by sampling; and step 3: extracting a wavelet packet characteristic energy value of the PCC voltage signal; and 4, step 4: combining the wavelet packet characteristic energy values of the extracted PCC voltage signals to construct input characteristic vectors of island detection; and 5: constructing a random forest decision tree classifier; step 6: constructing a random forest island identification system; and 7: and (3) building a photovoltaic grid-connected power generation simulation model by adopting Matlab/Simulink, and verifying the feasibility of the provided island detection method. The invention can effectively identify various interference states, is not easy to make wrong judgment, and obviously improves the accuracy of detection results.
Description
Technical Field
The invention belongs to the technical field of photovoltaic grid-connected power generation island detection, and particularly relates to a photovoltaic grid-connected power generation island detection method based on energy characteristics and random forests.
Background
The existing island detection method can be roughly divided into a remote detection method and a photovoltaic grid-connected inverter output end detection method according to different basic detection principles, and the photovoltaic grid-connected inverter output end island detection method can be divided into a passive island detection method and an active island detection method. Along with the continuous research of intelligent systems in various countries in the world, intelligent algorithms develop rapidly. The research on photovoltaic grid-connected power generation also explores an intelligent algorithm applied to island detection so as to make up for the defects of the traditional island detection method. The passive detection method is combined with an intelligent algorithm, so that no disturbance signal is added, the power quality is not interfered, and the operation result of the system is not influenced by the algorithm result, so that the method becomes a mainstream island detection method at present.
Although the island detection algorithm based on wavelet packet transformation is obviously improved in the aspects of island detection time, island detection blind areas and influence on electric energy quality compared with the traditional passive method and active method. However, human factors exist in the threshold setting of the algorithm, erroneous judgment and missing detection may be caused due to improper threshold setting, and particularly, due to the existence of error factors such as complexity of a power grid, variability of local load characteristics, signal acquisition and calculation and the like, the threshold setting is more difficult.
The island detection method based on wavelet packet transformation and BP neural network improves the island detection performance to a certain extent, but still has some defects, such as slow learning convergence speed and difficult network structure determination; and a large amount of processor resources are occupied during detection, and the detection process is complex. Meanwhile, the selection of the network structure, the initial connection weight and the threshold has randomness, so that the influence on network training is large, and the possibility of missed judgment and erroneous judgment exists in a certain interval.
The problems existing in the existing island detection method are as follows:
(1) due to the existence of human factors, the complexity of the power grid, the variability of local load characteristics, error factors such as signal acquisition and calculation, the setting of a threshold value is very difficult when the isolated island signal is analyzed by applying simple wavelet transformation.
(2) The detection algorithm with strong system identification capability is applied, the problems of low learning convergence speed and inconsistent network structure selection exist, and the selection of the initial connection weight and the threshold has randomness, so that the island detection result is greatly influenced.
(3) The traditional island detection method is easily interfered under the action of external disturbance, so that misjudgment is made, a large number of training samples are processed with certain difficulty, and an island detection result is influenced.
Disclosure of Invention
Aiming at the defects of the method, the invention provides the photovoltaic grid-connected power generation island detection method based on the energy characteristics and the random forest, which can effectively identify various interference states, is not easy to make wrong judgment and obviously improves the accuracy of the detection result.
In order to solve the technical problems, the invention is realized by the following technical scheme: the invention provides a photovoltaic grid-connected power generation island detection method based on energy characteristics and random forests, which comprises the following steps:
step 1: the method comprises the steps that a sampling device is arranged in a photovoltaic grid-connected power generation system, and a PCC voltage signal is collected when the photovoltaic grid-connected power generation system is in an island operation state;
step 2: carrying out wavelet packet analysis processing on the PCC voltage signal obtained by sampling;
and step 3: extracting a wavelet packet characteristic energy value of the PCC voltage signal;
and 4, step 4: combining the wavelet packet characteristic energy values of the extracted PCC voltage signals to construct input characteristic vectors of island detection;
and 5: constructing a random forest decision tree classifier;
step 6: constructing a random forest island identification system;
and 7: and (3) building a photovoltaic grid-connected power generation simulation model by adopting Matlab/Simulink, and verifying the feasibility of the provided island detection method.
In the above technical solution, the performing wavelet packet analysis processing on the PCC voltage signal obtained by sampling includes:
performing three-layer wavelet packet decomposition on the extracted PCC voltage signal, wherein the three-layer wavelet packet decomposition is performed on the PCC voltage signal, and an original signal meets the following formula:
S＝AAA3+DAA3+ADA3+DDA3+AAD3+DAD3+ADD3+DDD3
decomposing and reconstructing the PCC voltage signal
Then
Can be expressed as:
the wavelet packet decomposition process comprises the following steps:
in the formula (I), the compound is shown in the specification,
is the wavelet packet decomposition coefficient; a isk-2lIs a low pass filter coefficient; bk-2lIs a high pass filter coefficient;
wavelet packet reconstruction of PCC voltage signal can be performed by
And
the following are obtained:
in the above technical solution, the extracting a wavelet packet characteristic energy value of the PCC voltage signal includes:
the sum of squares of signals of different frequency bands in a certain sampling period, that is, the characteristic energy at this time is obtained as follows:
wherein E is(i，j)Representing the characteristic energy value of the jth node on the ith layer, m representing the mth sampling instant, S(i,j)To sample the signal, t1And t2Representing the start and end times of the sampling period.
In the above technical solution, the combining the extracted wavelet packet characteristic energy values of the PCC voltage signal, and constructing an input characteristic vector for island detection includes:
combining the characteristic energy of 8 nodes of the third layer to construct a characteristic vector T:
T＝[E(3,0),E(3,1),E(3,2),E(3,3),E(3,4),E(3,5),E(3,6),E(3,7)」
normalizing the characteristic vector T to obtain a characteristic vector T', wherein E0,E1,…E7Respectively representing the characteristic energy value of the wavelet packet of each frequency band when the grid-connected system normally operates:
in the above technical solution, the constructing a random forest decision tree classifier includes:
assume that a random forest is classified by a set of decision trees (h (X, theta))k) 1,2, …, n, where the sequence { theta } is a linear classifierkK is 1,2, … K, independent of each other and equally distributed, K being the number of decision trees; when the original sample set X is determined, the final result is determined by voting of all decision tree classifiers together:
wherein H (x) is a combined classification model; i (·) is an illustrative function; h isi(x) Classifying the models for a single decision tree; y is an output variable;
the island identification algorithm based on the random forest has certain convergence, and a group of decision trees { h } are determined1(x),h2(x),…,hk(x) Randomly sampling from an original sample set (X, Y) subject to random distribution, and defining a margin function as follows:
the generalization error and convergence expression for a random forest can be expressed as:
in the above technical solution, constructing a random forest island identification system includes:
(1) randomly extracting K samples from a training sample set X by using a bootstrap method to construct K decision tree classifiers;
(2) training the extracted K sample sets, and optimizing the decision tree classifier to avoid the problem of too many branches of the decision tree, wherein the degree of reduction of the overall loss function after pruning each internal node t of the decision tree is calculated as shown in the following formula:
α＝min(α,g(t))
wherein, TtRepresenting a tree with t as a source node; c (T)t) Representing the classification error of the selected sample; i TtIs T |tParameter α ≧ 0 measures the fitness of the training sample;
sequentially accessing an internal node t from top to bottom, judging whether g (t) is true or not at α, and finally selecting an optimal subtree by a cross verification method;
(3) and forming the generated optimal subtrees into a random forest identification system, classifying the test data, and voting and determining a final classification result by each decision tree classifier.
In the technical scheme, the Matlab/Simulink is adopted to build a photovoltaic grid-connected power generation simulation model, and the verification of the feasibility of the provided island detection method comprises the following steps:
step 7.1: setting simulation circuit parameters according to the operation standard of the photovoltaic grid-connected power generation system;
step 7.2: setting different operation states of the photovoltaic grid-connected simulation system to obtain an operation simulation result of the photovoltaic grid-connected power generation system;
step 7.3: the method comprises the steps of collecting PCC voltage signals of the photovoltaic grid-connected power generation system, extracting training samples and testing samples, testing under the condition of different training sample numbers, and calculating the accuracy of identification results.
Compared with the prior art, the technical scheme of the invention has the following beneficial effects: the invention provides a photovoltaic grid-connected power generation island detection method based on energy characteristics and random forests on the basis of passive detection. The wavelet packet energy characteristic can better reflect the change information of the island state, the problems of difficult threshold setting and the like do not exist, and the stability of island detection is improved; the random forest is used as one of classic machine learning algorithms, so that the defects that a BP neural network is low in convergence speed, not easy to determine in structure, easy to influence on selection of initial connection weight and threshold and the like are overcome, a large-scale training sample can be processed, and meanwhile, various feature vectors can be contained; the island detection method can effectively identify various interference states, is not easy to make wrong judgment, and obviously improves the accuracy of detection results.
The foregoing description is only an overview of the technical solutions of the present invention, and in order to make the technical means of the present invention more clearly understood, the present invention may be implemented in accordance with the content of the description, and in order to make the above and other objects, features, and advantages of the present invention more clearly understood, the following detailed description is given in conjunction with the preferred embodiments, together with the accompanying drawings.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings of the embodiments will be briefly described below.
FIG. 1 is a diagram of Matlab simulation results of the present invention;
FIG. 2 is a diagram of a three-layer wavelet packet decomposition process according to the present invention;
FIG. 3 is a diagram of an island identification model construction process according to the present invention;
fig. 4 is a simulation circuit diagram of the photovoltaic grid-connected power generation system of the invention.
Detailed Description
Other aspects, features and advantages of the present invention will become apparent from the following detailed description, taken in conjunction with the accompanying drawings, which form a part of this specification, and which illustrate, by way of example, the principles of the invention. In the referenced drawings, the same or similar components in different drawings are denoted by the same reference numerals.
The method for detecting the islanding of the photovoltaic grid-connected power generation based on the energy characteristics and the random forest comprises the following specific steps:
The sampling device is purchased from a cis-sourced science and technology limited company, and is a signal A/D converter with the model of ISO4021, and comprises power isolation, signal isolation, linearization, A/D conversion and RS-485 serial communication, so that signal acquisition between a sensor and a host can be realized, and voltage signals can be acquired.
And 2, carrying out wavelet packet analysis processing on the PCC voltage signal obtained by sampling.
After the photovoltaic grid-connected power generation system enters into the island operation, system parameters are changed, and island identification signals fluctuate. When the output power of the inverter is matched with the load power, the PCC voltage and frequency change is small, and a large blind area can occur by using a traditional island detection method. The PCC voltage signals are processed by applying wavelet packet transformation, so that more comprehensive operation information can be reserved. And carrying out three-layer wavelet packet decomposition on the extracted PCC voltage signal, wherein the decomposition schematic diagram is shown in FIG. 2.
As can be seen from the three-layer wavelet packet decomposition diagram, the original signal satisfies the following formula:
S＝AAA3+DAA3+ADA3+DDA3+AAD3+DAD3+ADD3+DDD3
decomposing and reconstructing the PCC voltage signal
Then
Can be expressed as:
the wavelet packet decomposition process comprises the following steps:
in the formula (I), the compound is shown in the specification,
is the wavelet packet decomposition coefficient; a isk-2lIs a low pass filter coefficient; bk-2lAre high pass filter coefficients.
Wavelet packet reconstruction of PCC voltage signal can be performed by
And
the following are obtained:
The wavelet packet characteristic energy is obtained by taking the sum of squares of signals of different frequency bands in a certain sampling period at a certain sampling moment, namely the characteristic energy at the moment is represented as follows.
Wherein E is(i，j)Representing the characteristic energy value of the jth node on the ith layer, m representing the mth sampling instant, S(i,j)To sample the signal, t1And t2Representing the start and end times of the sampling period.
Taking the case of performing 3-layer wavelet packet processing on the voltage signal of the common coupling point, performing 3-layer wavelet packet decomposition on the PCC voltage signal within a certain period of time to obtain wavelet packet decomposition coefficients of high and low frequency bands. Then, the obtained decomposition coefficients are subjected to reconstruction processing to finally obtain reconstruction signals of different frequency bands, S(3,0)，S(3,1)，S(3,2)，S(3,3)，S(3,4)，S(3,5)，S(3,6)，S(3,7)By the sequence S(3,n)Where n is 0,1, … 7.
Extracting the characteristic energy value of the wavelet packet as shown in the following formula, wherein pjmRepresenting the magnitude of the reconstruction coefficients.
And 4, combining the wavelet packet characteristic energy values of the extracted PCC voltage signals to construct an input characteristic vector of the island detection. Combining the characteristic energy of 8 nodes of the third layer to construct a characteristic vector T:
T＝[E(3,0),E(3,1),E(3,2),E(3,3),E(3,4),E(3,5),E(3,6),E(3,7)」
the detected island signal change has randomness, a certain sampling time period may appear, the energy after reconstruction of some frequency bands is larger, and some energy is too small, when the energy difference is larger, the energy value with important effect can weaken the influence due to the interaction of the two, and the energy value with smaller opposite influence can increase the effect, thereby greatly influencing the application effect of the characteristic vectorAnd (5) fruit. Aiming at the problem, the normalization processing is carried out on the obtained feature vector, so that the function of weakening the feature vector is avoided. Normalizing the characteristic vector T to obtain a characteristic vector T', wherein E0,E1,…E7Respectively representing the wavelet packet characteristic energy values of each frequency band when the grid-connected system normally operates.
And 5, constructing a random forest decision tree classifier.
Assume that a random forest is classified by a set of decision trees (h (X, theta))k) 1,2, …, n, where the sequence { theta } is a linear classifierkK is 1,2, … K, independent of each other and equally distributed, K being the number of decision trees. When the original sample set X is determined, the final result is determined by voting of all decision tree classifiers together:
wherein H (x) is a combined classification model; i (·) is an illustrative function; h isi(x) Classifying the models for a single decision tree; and Y is an output variable.
The island identification algorithm based on the random forest has certain convergence, and a group of decision trees { h } are determined1(x),h2(x),…,hk(x) And randomly sampling from an original sample set (X, Y) subjected to random distribution, and defining a margin function as the following formula. And (3) reflecting the classification accuracy by using a margin function, wherein the classification accuracy is higher along with the increase of the margin value.
The generalization error and convergence expression for a random forest can be expressed as:
therefore, the island identification algorithm based on the random forest does not have the over-fitting problem caused by the increase of the number of the decision tree classifiers, and the accuracy of the classification result is improved.
And 6, constructing a random forest island identification system.
The construction and classification process of the random forest island identification system is shown in fig. 3:
(1) and randomly extracting K samples from the training sample set X by using a bootstrap method to construct K decision tree classifiers.
(2) Training the extracted K sample sets, and optimizing the decision tree classifier to avoid the problem of too many branches of the decision tree, as shown in the following formula. Calculating the reduction degree of the overall loss function after the pruning of each internal node t of the decision tree:
α＝min(α,g(t))
wherein, TtRepresenting a tree with t as a source node; c (T)t) Representing the classification error of the selected sample; i TtIs T |tAnd parameter α ≧ 0 measures the fitness of the training sample.
And sequentially accessing the internal node t from top to bottom, judging whether g (t) is true or not α, and finally selecting the optimal subtree by a cross-validation method.
(3) And forming the generated optimal subtrees into a random forest identification system, classifying the test data, and voting and determining a final classification result by each decision tree classifier.
And 7, building a photovoltaic grid-connected power generation simulation model by adopting Matlab/Simulink, and verifying the feasibility of the provided island detection method.
The island identification simulation circuit of the photovoltaic grid-connected power generation system is shown in figure 4, wherein U1、U2、i1、i2PCC voltage, grid voltage, inverter output current and grid current, respectively.
And 7.1, setting simulation circuit parameters.
According to the operation standard of the photovoltaic grid-connected power generation system, simulation circuit parameters are set as shown in table 1.
TABLE 1 Square-rule Circuit parameters
| Simulation parameters | Numerical value |
| |
400V |
| Fundamental frequency | 50Hz |
| Output power of inverter | 1500W |
| Local load power | P＝1500W,QL＝QC＝1500var |
| Switching load power | P＝200W,QL＝QC＝200var |
And 7.2, setting different operation states of the photovoltaic grid-connected simulation system to obtain an operation simulation result of the photovoltaic grid-connected power generation system.
Aiming at the problem that the traditional island detection method is easily influenced by external disturbance and is easy to have misjudgment, in order to improve the accuracy of island identification, two running states of an island and disturbance are specially designed to be used for generating a sample set of characteristic vectors. Considering the influence of external disturbance on the identification result, normal and island operation states and 3 external disturbance operation states are designed.
The variation of the PCC voltage, the grid voltage, the inverter output current and the grid current is analyzed, and the corresponding Matlab simulation result is shown in fig. 1.
(1) Fig. 1 (a) shows an operation waveform of the photovoltaic grid-connected power generation system in a normal operating state, at this time, waveforms of the PCC voltage, the grid voltage, the inverter output current, and the grid current are all unchanged, and the amplitude value is kept stable.
(2) Fig. 1 (b) shows an island detection waveform when a load is suddenly switched on. Due to the input of the load, the grid current rapidly increases at 0.1s, and at the moment, the inverter is still in a grid-connected state, so that the PCC voltage, the grid voltage and the inverter output current are kept unchanged.
(3) Fig. 1 (c) shows an islanding detection waveform in the case of harmonic disturbance. The grid current has certain distortion at 0.1s, which shows that harmonic disturbance is added to the grid voltage at 0.1s, but the PCC voltage and the inverter output current are kept unchanged at the moment, and the state belongs to an island disturbance state.
(4) Fig. 1 (d) shows an island detection waveform with a short-circuited local load. When the grid current suddenly increases in 0.1s, the voltage of the PCC point is greatly attenuated, a strong interference state is formed, the output current of the inverter is still kept constant, and the state is a non-isolated island state.
(5) Fig. 1 (e) shows a detection waveform when an island is generated. The grid current is rapidly reduced to 0 at 0.1s, which indicates that an island is generated at the moment, the photovoltaic power generation system is in an off-grid state, the PCC voltage amplitude value is kept stable, and the inverter output current has small fluctuation, which is caused by the fact that the photovoltaic power generation system cannot rapidly reach a complete stable state at the moment of grid disconnection.
Step 7.3, collecting PCC voltage signals of the photovoltaic grid-connected power generation system, extracting 600 groups of training samples and 800 groups of test samples in total, and displaying in a table 2:
TABLE 2 sample set
In order to verify the classification capability of the island identification system, the island identification system is designed to be tested under the condition of different training sample numbers, and the accuracy of the identification result is calculated. Simulation identification results show that the island operation state can be accurately judged by the photovoltaic grid-connected power generation island identification method based on the energy characteristics and the random forest.
The above description is only the most basic embodiment of the present invention, but the scope of the present invention is not limited thereto, and any alternative, which can be understood by those skilled in the art within the technical scope of the present invention, should be covered by the present invention, such as other distributed generation island detection methods based on the method of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (7)
1. A photovoltaic grid-connected power generation island detection method based on energy characteristics and random forests is characterized by comprising the following steps:
step 1: the method comprises the steps that a sampling device is arranged in a photovoltaic grid-connected power generation system, and a PCC voltage signal is collected when the photovoltaic grid-connected power generation system is in an island operation state;
step 2: carrying out wavelet packet analysis processing on the PCC voltage signal obtained by sampling;
and step 3: extracting a wavelet packet characteristic energy value of the PCC voltage signal;
and 4, step 4: combining the wavelet packet characteristic energy values of the extracted PCC voltage signals to construct input characteristic vectors of island detection;
and 5: constructing a random forest decision tree classifier;
step 6: constructing a random forest island identification system;
and 7: and (3) building a photovoltaic grid-connected power generation simulation model by adopting Matlab/Simulink, and verifying the feasibility of the provided island detection method.
2. The method for detecting the islanding of the photovoltaic grid-connected power generation based on the energy characteristics and the random forest according to claim 1, wherein the step of performing wavelet packet analysis processing on the PCC voltage signals obtained by sampling comprises the steps of:
performing three-layer wavelet packet decomposition on the extracted PCC voltage signal, wherein the three-layer wavelet packet decomposition is performed on the PCC voltage signal, and an original signal meets the following formula:
S＝AAA3+DAA3+ADA3+DDA3+AAD3+DAD3+ADD3+DDD3
decomposing and reconstructing the PCC voltage signal
Then
Can be expressed as:
the wavelet packet decomposition process comprises the following steps:
in the formula (d)l j+1,n，dl j,2n，dl j,2n+1Is the wavelet packet decomposition coefficient; a isk-2lIs a low pass filter coefficient; bk-2lIs a high pass filter coefficient;
wavelet packet reconstruction of PCC voltage signal can be performed by
And
the following are obtained:
3. the method for detecting the islanding of photovoltaic grid-connected power generation based on the energy characteristics and the random forest according to claim 1, wherein the extracting the wavelet packet characteristic energy value of the PCC voltage signal comprises:
the sum of squares of signals of different frequency bands in a certain sampling period, that is, the characteristic energy at this time is obtained as follows:
wherein E is(i，j)Representing the characteristic energy value of the jth node on the ith layer, m representing the mth sampling instant, S(i,j)To sample the signal, t1And t2Representing the start and end times of the sampling period.
4. The method for detecting islanding on grid-connected photovoltaic power generation based on energy characteristics and random forests according to claim 1, wherein the step of combining the wavelet packet characteristic energy values of the extracted PCC voltage signals to construct an input characteristic vector for islanding detection comprises the steps of:
combining the characteristic energy of 8 nodes of the third layer to construct a characteristic vector T:
T＝[E(3,0),E(3,1),E(3,2),E(3,3),E(3,4),E(3,5),E(3,6),E(3,7)」
normalizing the characteristic vector T to obtain a characteristic vector T', wherein E0,E1,…E7Respectively representing the characteristic energy value of the wavelet packet of each frequency band when the grid-connected system normally operates:
5. the method for detecting the islanding of the photovoltaic grid-connected power generation based on the energy characteristics and the random forest as claimed in claim 1, wherein the constructing of the random forest decision tree classifier comprises:
assume that a random forest is classified by a set of decision trees (h (X, theta))k) 1,2, …, n, where the sequence { theta } is a linear classifierkK is 1,2, … K, independent of each other and equally distributed, K being the number of decision trees; when the original sample set X is determined, the final result is determined by voting of all decision tree classifiers together:
wherein H (x) is a combined classification model; i (·) is an illustrative function; h isi(x) Classifying the models for a single decision tree; y is an output variable;
the island identification algorithm based on the random forest has certain convergence, and a group of decision trees { h } are determined1(x),h2(x),…,hk(x) Randomly sampling from an original sample set (X, Y) subject to random distribution, and defining a margin function as follows:
the generalization error and convergence expression for a random forest can be expressed as:
6. the grid-connected photovoltaic power generation island detection method based on energy characteristics and random forests as claimed in claim 1, wherein constructing a random forest island identification system comprises:
(1) randomly extracting K samples from a training sample set X by using a bootstrap method to construct K decision tree classifiers;
(2) training the extracted K sample sets, and optimizing the decision tree classifier to avoid the problem of too many branches of the decision tree, wherein the degree of reduction of the overall loss function after pruning each internal node t of the decision tree is calculated as shown in the following formula:
α＝min(α,g(t))
wherein, TtRepresenting a tree with t as a source node; c (T)t) Representing the classification error of the selected sample; i TtIs T |tParameter α ≧ 0 measures the fitness of the training sample;
sequentially accessing an internal node t from top to bottom, judging whether g (t) is true or not at α, and finally selecting an optimal subtree by a cross verification method;
(3) and forming the generated optimal subtrees into a random forest identification system, classifying the test data, and voting and determining a final classification result by each decision tree classifier.
7. The method for detecting the islanding of the photovoltaic grid-connected power generation based on the energy characteristics and the random forest according to claim 1, wherein the Matlab/Simulink is adopted to build a photovoltaic grid-connected power generation simulation model, and the verification of the feasibility of the islanding detection method comprises the following steps:
step 7.1: setting simulation circuit parameters according to the operation standard of the photovoltaic grid-connected power generation system;
step 7.2: setting different operation states of the photovoltaic grid-connected simulation system to obtain an operation simulation result of the photovoltaic grid-connected power generation system;
step 7.3: the method comprises the steps of collecting PCC voltage signals of the photovoltaic grid-connected power generation system, extracting training samples and testing samples, testing under the condition of different training sample numbers, and calculating the accuracy of identification results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010188667.4A CN111273131A (en) | 2020-03-17 | 2020-03-17 | Photovoltaic grid-connected power generation island detection method based on energy characteristics and random forest |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010188667.4A CN111273131A (en) | 2020-03-17 | 2020-03-17 | Photovoltaic grid-connected power generation island detection method based on energy characteristics and random forest |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111273131A true CN111273131A (en) | 2020-06-12 |
Family
ID=70999785
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010188667.4A Pending CN111273131A (en) | 2020-03-17 | 2020-03-17 | Photovoltaic grid-connected power generation island detection method based on energy characteristics and random forest |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111273131A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112181055A (en) * | 2020-09-28 | 2021-01-05 | 广东小天才科技有限公司 | Indoor and outdoor state judgment method, wearable device and computer readable storage medium |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102253283A (en) * | 2011-06-20 | 2011-11-23 | 山东电力集团公司临沂供电公司 | Island detection method based on wavelet packet energy spectrum |
| CN105759177A (en) * | 2016-04-26 | 2016-07-13 | 浙江大学城市学院 | Classified-multi-mode-fusion-based distributed grid island detection method |
| CN107194216A (en) * | 2017-05-05 | 2017-09-22 | 中南大学 | A kind of mobile identity identifying method and system of the custom that swiped based on user |
| CN108303630A (en) * | 2018-02-08 | 2018-07-20 | 国电南瑞科技股份有限公司 | A kind of alternating current-direct current charging equipment power device method for diagnosing faults based on wavelet packet analysis |
| CN109490838A (en) * | 2018-09-20 | 2019-03-19 | 中国人民解放军战略支援部队航天工程大学 | A kind of Recognition Method of Radar Emitters of data base-oriented incompleteness |
| CN109934089A (en) * | 2018-10-31 | 2019-06-25 | 北京航空航天大学 | Multistage epileptic EEG Signal automatic identifying method based on supervision gradient lifter |
| CN110377927A (en) * | 2019-05-06 | 2019-10-25 | 河海大学 | A kind of pumping plant unit rotor method for monitoring state based on MATLAB emulation |
| CN110488152A (en) * | 2019-09-27 | 2019-11-22 | 国网河南省电力公司电力科学研究院 | A kind of distribution network fault line selection method based on Adaptive Neuro-fuzzy Inference |
| CN110547807A (en) * | 2019-09-17 | 2019-12-10 | 深圳市赛梅斯凯科技有限公司 | driving behavior analysis method, device, equipment and computer readable storage medium |
| CN110786851A (en) * | 2019-10-31 | 2020-02-14 | 长春理工大学 | Method for improving wavelet packet decomposition speed based on Mallat algorithm |
-
2020
- 2020-03-17 CN CN202010188667.4A patent/CN111273131A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102253283A (en) * | 2011-06-20 | 2011-11-23 | 山东电力集团公司临沂供电公司 | Island detection method based on wavelet packet energy spectrum |
| CN105759177A (en) * | 2016-04-26 | 2016-07-13 | 浙江大学城市学院 | Classified-multi-mode-fusion-based distributed grid island detection method |
| CN107194216A (en) * | 2017-05-05 | 2017-09-22 | 中南大学 | A kind of mobile identity identifying method and system of the custom that swiped based on user |
| CN108303630A (en) * | 2018-02-08 | 2018-07-20 | 国电南瑞科技股份有限公司 | A kind of alternating current-direct current charging equipment power device method for diagnosing faults based on wavelet packet analysis |
| CN109490838A (en) * | 2018-09-20 | 2019-03-19 | 中国人民解放军战略支援部队航天工程大学 | A kind of Recognition Method of Radar Emitters of data base-oriented incompleteness |
| CN109934089A (en) * | 2018-10-31 | 2019-06-25 | 北京航空航天大学 | Multistage epileptic EEG Signal automatic identifying method based on supervision gradient lifter |
| CN110377927A (en) * | 2019-05-06 | 2019-10-25 | 河海大学 | A kind of pumping plant unit rotor method for monitoring state based on MATLAB emulation |
| CN110547807A (en) * | 2019-09-17 | 2019-12-10 | 深圳市赛梅斯凯科技有限公司 | driving behavior analysis method, device, equipment and computer readable storage medium |
| CN110488152A (en) * | 2019-09-27 | 2019-11-22 | 国网河南省电力公司电力科学研究院 | A kind of distribution network fault line selection method based on Adaptive Neuro-fuzzy Inference |
| CN110786851A (en) * | 2019-10-31 | 2020-02-14 | 长春理工大学 | Method for improving wavelet packet decomposition speed based on Mallat algorithm |
Non-Patent Citations (2)
| Title |
|---|
| SANCHAY A,BHAVESH R.: "Islanding detection of distributed generation using random forest technique", 《IEEE INTERNATIONAL CONFERENCE ON POWER SYSTEM NEW DELHI: INSTITUTE OF ELECTRICAL AND ELECTRONICS》 * |
| 马文忠 等: "基于能量特性的分布式电源孤岛检测技术", 《电气应用》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112181055A (en) * | 2020-09-28 | 2021-01-05 | 广东小天才科技有限公司 | Indoor and outdoor state judgment method, wearable device and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Khokhar et al. | A new optimal feature selection algorithm for classification of power quality disturbances using discrete wavelet transform and probabilistic neural network | |
| Zhu et al. | Wavelet-based fuzzy reasoning approach to power-quality disturbance recognition | |
| Saini et al. | Detection and classification of power quality disturbances in wind‐grid integrated system using fast time‐time transform and small residual‐extreme learning machine | |
| CN110287762B (en) | Non-invasive load identification method and device based on data mining technology | |
| CN109672221B (en) | Direct-drive wind power plant dynamic equivalence method for subsynchronous oscillation analysis | |
| CN111160241B (en) | Power distribution network fault classification method, system and medium based on deep learning | |
| CN110516788A (en) | A kind of non-intrusion type load testing method and system based on Bi-LSTM algorithm | |
| CN109948726B (en) | Power quality disturbance classification method based on deep forest | |
| Song et al. | A negative selection algorithm-based identification framework for distribution network faults with high resistance | |
| CN110247420B (en) | Intelligent fault identification method for HVDC transmission line | |
| CN113496440B (en) | User abnormal electricity consumption detection method and system | |
| CN111553112A (en) | Power system fault identification method and device based on deep belief network | |
| Xu et al. | Novel fault location for high permeability active distribution networks based on improved VMD and S-transform | |
| Zhu et al. | Complex disturbances identification: A novel PQDs decomposition and modeling method | |
| CN111273131A (en) | Photovoltaic grid-connected power generation island detection method based on energy characteristics and random forest | |
| CN116712082A (en) | Electrocardiogram data enhancement method, system, medium, equipment and terminal | |
| CN114859169A (en) | Intelligent identification method and system for distribution transformer outgoing line load and storage medium | |
| CN109239585A (en) | A kind of method for diagnosing faults based on the preferred wavelet packet of improvement | |
| CN110245617B (en) | Artificial intelligence analysis method based on transient recording waveform | |
| CN114415056A (en) | Distributed power supply island fault detection method | |
| CN113902136A (en) | Load identification method based on electric power fingerprint features and integrated learning mechanism | |
| Golla et al. | Cross wavelet-SVM for detection and classification of power quality problems in distribution system | |
| Behzadi et al. | Identification of combined power quality disturbances in the presence of distributed generations using variational mode decomposition and K-nearest neighbors classifier | |
| Li et al. | Fault diagnosis for single-phase grounding in distribution network based on hilbert-huang transform and siamese convolution neural network | |
| CN116541686B (en) | Electric energy quality disturbance classification method based on multi-domain feature fusion extreme learning machine |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20200612 |