CN111245365B - Photovoltaic module fault diagnosis method based on dynamic weighted depth forest - Google Patents
Photovoltaic module fault diagnosis method based on dynamic weighted depth forest Download PDFInfo
- Publication number
- CN111245365B CN111245365B CN202010021220.8A CN202010021220A CN111245365B CN 111245365 B CN111245365 B CN 111245365B CN 202010021220 A CN202010021220 A CN 202010021220A CN 111245365 B CN111245365 B CN 111245365B
- Authority
- CN
- China
- Prior art keywords
- forest
- fault
- photovoltaic module
- training
- cascade
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000012549 training Methods 0.000 claims abstract description 56
- 238000012360 testing method Methods 0.000 claims abstract description 8
- 238000003066 decision tree Methods 0.000 claims abstract description 7
- 238000012216 screening Methods 0.000 claims abstract description 4
- 238000007637 random forest analysis Methods 0.000 claims description 12
- 238000005070 sampling Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 5
- 238000001514 detection method Methods 0.000 claims description 4
- 230000032683 aging Effects 0.000 claims description 3
- 238000013145 classification model Methods 0.000 claims description 3
- OWZREIFADZCYQD-NSHGMRRFSA-N deltamethrin Chemical compound CC1(C)[C@@H](C=C(Br)Br)[C@H]1C(=O)O[C@H](C#N)C1=CC=CC(OC=2C=CC=CC=2)=C1 OWZREIFADZCYQD-NSHGMRRFSA-N 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 abstract description 12
- 230000008569 process Effects 0.000 description 7
- 238000013135 deep learning Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000007635 classification algorithm Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 238000010248 power generation Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 239000002800 charge carrier Substances 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000012946 outsourcing Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Images
Classifications
-
- 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
- H02S50/10—Testing of PV devices, e.g. of PV modules or single PV cells
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Pure & Applied Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Algebra (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a photovoltaic module fault diagnosis method based on dynamic weighted depth forest, which comprises the following steps: establishing an equivalent circuit model of the photovoltaic module, and screening out fault data representing a fault type; selecting three sliding windows with different sizes, and training a forest in a multi-granularity scanning stage; dynamically weighting the decision tree in the forest; calculating the prediction probability vector of the subtree, and selecting the optimal probability to obtain various prediction results; calculating a prediction result of the current cascade forest; judging whether the accuracy of the current cascade forest is improved or not until a training model with the highest accuracy is obtained or the number of layers of the cascade forest is increased to continue training; and inputting the test sample to obtain a classification result. The fault diagnosis method has the advantages that the super-parameter setting by utilizing the dynamic weighted depth forest algorithm is simple, the fault characteristics of the photovoltaic module can be automatically learned, the layer number of the cascade forests can be automatically determined, the diagnosis precision is high, and the diagnosis result is more direct.
Description
Technical Field
The invention relates to a photovoltaic module fault diagnosis method, in particular to a photovoltaic module fault diagnosis method based on dynamic weighted depth forest.
Background
Energy shortage and pollution problems have become important factors affecting industrial production and environmental protection. The development of clean energy has become a global issue. Solar energy is widely applied to a photovoltaic power generation system as clean energy, can avoid the problem of environmental pollution caused by the traditional petrochemical energy as renewable energy, and brings a new concept of revolution for industrial manufacturing and production and life.
Although solar energy is widely replacing the increasingly short conventional energy sources, since the solar panels for photovoltaic power generation need to be erected in the natural environment or on the top of a house, some photovoltaic modules are shielded by dark clouds, leaves and the like, and damage to the battery modules is caused. In order to avoid loss caused by faults of the photovoltaic module and improve the efficiency of photovoltaic power generation, various faults of the solar panel in the photovoltaic system need to be judged and processed. In the prior art, some intelligent algorithms are mainly adopted to model fault types, and the type of the fault of the photovoltaic module is judged through model training samples.
The patent application with the application number of CN201810907602.3 discloses a photovoltaic array fault diagnosis method based on an improved random forest algorithm, which reflects the operation state of each branch in a photovoltaic array through the parameters of a total trunk and each branch of the photovoltaic array, and reflects the operation state of each photovoltaic cell assembly in the branch through the voltage difference between arrays of different branches, thereby realizing the fault location of the photovoltaic array, and simultaneously, optimizing and improving the three parts of decision tree weight voting, tie processing and the importance measurement of fault characteristics by using an outsourcing sample; however, this patent requires a series of processes by random forest algorithm, decision tree weighting, voting tie processing, and variable importance measurement, and thus the training time is long.
The patent application with the application number of CN201810766919.X discloses an intelligent photovoltaic array fault diagnosis method based on an optimal rotating forest, which comprises the steps of firstly collecting photovoltaic electrical characteristic data under various working conditions, then utilizing a selection algorithm to carry out importance weight sorting on the photovoltaic electrical characteristic data to obtain the most important fault characteristics, then utilizing an improved rotating forest algorithm to process the photovoltaic electrical characteristic data to obtain input variables of a base classifier, adopting a traversal method to obtain optimal model parameters, and finally adopting a mixed algorithm combining the rotating forest algorithm and an extreme learning machine to train each sample in a training set to obtain an optimal rotating forest fault diagnosis training model; finally, carrying out fault detection and classification on the photovoltaic array by using a training model; however, this method requires collecting the photovoltaic electrical characteristic data under various working conditions, so the preparation conditions are harsh, and the weight sorting method may cause unexpected deviation of the diagnosis result.
Disclosure of Invention
The invention aims to: the photovoltaic module fault diagnosis method based on the dynamic weighted depth forest is provided, and the diagnosis efficiency and accuracy can be improved.
In order to achieve the purpose, the invention provides a photovoltaic module fault diagnosis method based on dynamic weighted depth forest, which comprises the following steps:
step 4, calculating the probability vector and the prediction result of the fault category n predicted by the subtree on the training sample set R;
step 5, calculating the prediction result of each level of the cascade forest on the training sample set R, and calculating the final result of the ith fault of the photovoltaic module;
and 7, inputting a test sample until the mapping from the fault data to the fault state is completed, and outputting a fault type result, thereby realizing the fault diagnosis of the photovoltaic module.
As a further limitation of the present invention, in step 1, a photovoltaic module with a plurality of cells connected in series is used as a detection target, and 8 parameters of a maximum power point, a maximum power point voltage, a maximum power point current, an open-circuit voltage, a short-circuit current, a fill factor, a temperature and irradiance in a working state are collected as input samples.
As a further limitation of the present invention, the multi-granularity scan of step 2 specifically includes: firstly, inputting a training sample set R with 8 acquisition parameters into a fault diagnosis system, and then acquiring three different dimensions dThe sample sliding window carries out unitized sliding sampling on the training sample set R to obtain a new feature vector set with a dimension d, and the number S of the vector setsiComprises the following steps: si9-d; wherein, i is 3, and d is 2, 3 and 4.
As a further limitation of the present invention, the dynamically weighted probability of class c of the e-th forest of the cascade forest in step 3 is:
Dec＝ω1eP1c+ω2eP2c+···+ωkePkc+···+ωtePtc；
in the formula, DecIs the probability, omega, of class c of the e-th forestkeIs the weight value, P, of the kth tree of the e-level forestkcThe probability of class c for the kth sub-tree.
As a further limitation of the present invention, in step 4, when the number of the training sample set R collected is i and the number of the diagnosis result categories is c, the kth sub-tree R of a certain forestkThen, the prediction probability vector of a certain fault category is: gx (R)k)＝[Pi1,Pi2,···,Pic]。
As a further limitation of the invention, said step 5 comprises a combination C of random forests and fully random forests in the forest of level e of the cascaded foresteAnd on the training sample set R, the prediction result of the ith fault of the photovoltaic module is as follows:
as a further limitation of the invention, the prediction probability of the final forest for the ith fault of the photovoltaic module is obtained through the forest F of the N level:
after dynamic weighting, the maximum value is taken for processing, and the final result of the ith fault is obtained as follows: qt (C)N)＝Max(Gx(Fi))。
As a further limitation of the present invention, in step 6, when the cascaded forest is not expanded any more, in the expanded levels, finding out the level number N corresponding to the highest value of the prediction accuracy on the training sample set R, combining the forest levels as a final classifier model, and in the test sample set E, using the classifier model obtained by training as the prediction model of the whole dynamic weighted depth forest.
As a further limitation of the present invention, in said step 7, the photovoltaic module fault diagnosis types include normal, open-circuit fault, short-circuit fault, shadow hotspot fault and aging fault.
The invention has the beneficial effects that:
1. according to the method, the deep learning algorithm and the fault diagnosis of the photovoltaic module are combined, the dynamic weighting optimization decision result is introduced, the difficulty in fault type diagnosis of the traditional method is improved, the advantages of the deep learning classification algorithm are introduced into the field of fault diagnosis, compared with the traditional deep learning algorithm for analyzing the fault type of the photovoltaic module, the dynamic weighting deep forest classification algorithm is simple in parameter setting, capable of automatically learning characteristics, capable of automatically determining the number of layers of cascade forests, high in diagnosis precision and more direct in diagnosis result.
2. The fault diagnosis method can utilize the prediction probability vector of the correction forest through the weighted depth forest, and the corrected probability vector is used as the input of the next stage, so that the next stage forest can be continuously optimized in the training process and the prediction precision of the next stage forest can be improved.
Drawings
FIG. 1 is a block flow diagram of a photovoltaic module fault diagnosis method based on dynamic weighted depth forest according to the present invention;
FIG. 2 is a photovoltaic module battery equivalent circuit model of the photovoltaic module fault diagnosis method based on the dynamic weighted depth forest of the invention;
FIG. 3 is a schematic diagram of dynamic weighting of a single forest in the photovoltaic module fault diagnosis method based on dynamic weighted deep forests;
FIG. 4 is a schematic diagram of a multi-granularity scanning process of the photovoltaic module fault diagnosis method based on the dynamic weighted deep forest;
fig. 5 is a schematic diagram of a cascade forest process of the photovoltaic module fault diagnosis method based on the dynamic weighted deep forest.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings. It should be understood that the detailed description and specific examples, while indicating the invention, are intended for purposes of illustration only and are not intended to limit the scope of the invention.
As shown in fig. 1, the invention provides a method for diagnosing faults of a photovoltaic module of a dynamic weighted depth forest, which comprises the following steps:
step 4, calculating the probability vector and the prediction result of the fault category n predicted by the subtree on the training sample set R;
step 5, calculating the prediction result of each level of the cascade forest on the training sample set R, and calculating the final result of the ith fault of the photovoltaic module;
and 7, inputting a test sample until the mapping from the fault data to the fault state is completed, and outputting a fault type result, thereby realizing the fault diagnosis of the photovoltaic module.
In order to collect important parameters of the photovoltaic module, reduce interference of other factors on electric parameter collection, and establish an equivalent circuit model of the photovoltaic module battery, as shown in fig. 2.
Considering the resistance loss, according to kirchhoff's law, the output characteristic equation of the photovoltaic cell assembly is expressed as:
Ipv＝Iph-Id-IRsh (1)
in the formula IpvOutputting current for the model load end; i isphIs a photosensitive current, representing charge carrier generation in the semiconductor layers of the PV cell caused by incident radiation; i isdIs a diode current; current I on loss resistanceRsh。
When a photovoltaic assembly with a plurality of batteries connected in series is taken as a detection target in the step 1, 8 parameters of a maximum power point, a maximum power point voltage, a maximum power point current, an open-circuit voltage, a short-circuit current, a filling factor, a temperature and irradiance in a working state are collected as input samples; that is, the training sample of step 1 and the test sample of step 7 are both input samples, and both use these 8 parameters.
Further, the multi-granularity scanning process in step 2 can be described as: the multi-granularity scanning in the step 2 specifically comprises the following steps: firstly, inputting a training sample set R with 8 acquisition parameters to a fault diagnosis system, and then performing unitized sliding sampling on the training sample set R through sampling sliding windows with three different dimensions d to obtain a new feature vector set with dimension d, wherein the number S of the vector setsiCan be expressed using equation (2):
Si＝9-d (2)
wherein, i is 3, and d is 2, 3 and 4; after scanning, S is generatediA subset of data.
The multi-granularity scanning process is illustrated in fig. 4, taking an 8-dimensional feature vector as an example. As can be seen from fig. 4, the original feature size is 8, and 6 sub-feature samples are generated after scanning and sampling through a sliding window with the dimension d being 3. When the subsamples are input into random forests, each forest generates a 2-dimensional class distribution vector (assuming two classes), and the original feature vector with the size of 8 is changed into a feature vector with the size of 24 dimensions and then used for training of the cascade forests. Of course, scanning windows of different dimensions d can be selected, so as to realize multi-granularity scanning in a true sense.
The class probability of the subtree prediction of each forest in the deep forest is different in size, and the classification result has different functions. Because the prediction precision of each subtree has certain difference, the prediction precision of the forest is reduced by directly carrying out arithmetic mean without data distribution processing, thereby influencing the prediction effect of the next-stage forest and increasing the stage number of the cascade forest. Therefore, the method is improved by combining dynamic weighting according to the prediction probability of each subtree, and a photovoltaic module fault diagnosis model of the dynamic weighted depth forest is established.
The structure of each forest in the step 3 is shown in figure 3; the dynamically weighted probability of class c of the e-th forest is:
Dec＝ω1eP1c+ω2eP2c+···+ωkePkc+···+ωtePtc (3)
in the formula, DecIs the probability, omega, of class c of the e-th forestkeIs the weight value, P, of the kth tree of the e-level forestkcThe probability of class c for the kth sub-tree.
In the above formula (3), the weight value ω of the kth tree of the e-level forestkeCalculated by the following way:
because the least square method can obtain the unknown data with the minimum error sum of squares with the real data to achieve the purpose of obtaining the optimal function, the weight value of a single forest is selected and the least square error sum is used as the target by the least square method:
wherein the content of the first and second substances,
is the probability of a single forest class c of the e-th level, and d is the window width of the adjacent historical samples.
Bringing formula (3) into formula (4):
calculating Q pair combination coefficient omegakePartial derivatives of (a):
let the partial derivative obtain 0, then calculate to obtain omegakeThe value of (c).
In the above formula (3), the probability P of class c of the kth sub-treekcCalculated by the following way: prediction probability PkcThe class C with the largest number of nodes of the current subtreejWith the current sample set RcThe ratio of (a) to (b), namely: pkc＝Cj/Rc。
And when the e +1 level forest is predicted, updating the prediction result of the next level forest, recalculating the combined weight value, and further realizing dynamic adjustment of the weight value. Substituting the obtained new combined weight value into the formula (3) to obtain the prediction result of the next forest.
Further, in step 4, when the number of the training sample set R collected is i and the number of the diagnosis result categories is c, assuming that a certain forest has t subtrees, the kth subtree is recorded as Rk(k∈[1,t]) The probability vector for predicting the nth failure class on the training sample set R is as follows:
Gx(Rk)＝[Pi1,Pi2,···,Pic] (7)
to obtain a better classification result, the prediction vector Gx (R) is obtained using a function Maxk) The value of the maximum value in the training samples is recorded as Qt (R)k) Then in the training sampleOn set R, subtree RkThe prediction result of a certain type of fault is shown as the following formula:
Qt(Rk)＝Max(Gx(Rk))＝pi (8)
in step 5, in order to extract the fault characteristics of the photovoltaic module, the fault characteristics are subjected to characterization learning by using a cascade forest. Each level of the cascade forest is composed of a random forest and a completely random forest in a cascade structure, and the random forest is formed by integrating decision trees of fault categories.
The structural diagram of the cascading forest is shown in FIG. 5, and if h forests are arranged at the level e in the cascading forest, on the training sample set R, a combination C of a random forest and a completely random forest in the forest is formedeThe prediction result of the ith fault of the photovoltaic module is as follows:
then, through the forest F of the N-level, the prediction probability of the final forest to the ith fault of the photovoltaic module is obtained as follows:
finally, after dynamic weighting, the maximum value is taken to obtain the final result of the ith fault as follows:
Qt(CN)＝Max(Gx(Fi)) (11)
and when the cascade forest is not expanded any more in the step 6, finding out the stage number N corresponding to the highest value of the prediction accuracy on the training sample set R in the expanded stage, and combining the stage forest as a final classifier model. On the test sample set E, the classifier model obtained by training is used as a prediction model of the whole dynamic weighted depth forest.
Further, the photovoltaic module fault diagnosis types in the step 7 include normal, open-circuit fault, short-circuit fault, shadow hot spot fault and aging fault.
By adopting the technical scheme disclosed by the invention, the deep learning algorithm is combined with the fault diagnosis of the photovoltaic module, the dynamic weighting optimization decision result is introduced, the difficulty in fault type diagnosis of the traditional method is improved, the advantages of the deep learning classification algorithm are introduced into the field of fault diagnosis, and compared with the traditional deep learning algorithm for analyzing the fault type of the photovoltaic module, the dynamic weighting deep forest classification algorithm has the advantages of simple parameter setting, capability of automatically learning characteristics, capability of automatically determining the number of layers of cascade forests, high diagnosis precision and more direct diagnosis result.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and improvements can be made without departing from the principle of the present invention, and such modifications and improvements should also be considered within the scope of the present invention.
Claims (5)
1. A photovoltaic module fault diagnosis method based on dynamic weighted depth forest is characterized by comprising the following steps:
step 1, establishing an equivalent circuit model of a photovoltaic module, collecting various data output by the photovoltaic module model, screening out fault data representing fault types, and setting part of fault data as a training sample set R;
step 2, performing multi-granularity scanning on the training samples by adopting sampling sliding windows with different dimensions, and performing step-by-step training on the training samples by adopting a cascade forest structure to obtain a multi-granularity cascade forest diagnosis model;
step 3, dynamically weighting each sub-tree in the multi-granularity cascading forest of the step 2;
step 4, calculating the probability vector and the prediction result of the fault category n predicted by the subtree on the training sample set R;
step 5, calculating the prediction result of each level of the cascade forest on the training sample set R, and calculating the final result of the ith fault of the photovoltaic module;
step 6, judging whether the accuracy is improved or not until a training model with the highest accuracy is obtained or the number of layers of a cascade forest is increased to continue training a decision tree of the forest, and finally obtaining a photovoltaic module fault classification model of the dynamic weighted depth forest;
step 7, inputting a test sample until the mapping from the fault data to the fault state is completed, outputting a fault type result, thereby realizing the fault diagnosis of the photovoltaic module,
wherein, the probability of the dynamic weighting of the class c of the e-th forest of the cascade forest in the step 3 is as follows:
Dec＝ω1eP1c+ω2eP2c+···+ωkePkc+···+ωtePtc；
in the formula, ωkeIs the weight value, P, of the kth tree of the e-level forestkcProbability of class c for the kth sub-tree, DecIs the probability of class c in the e-th forest,
in the step 4, when the collection number of the training sample set R is i and the number of the diagnosis result categories is c, the kth sub-tree R of a certain forestkThen, the prediction probability vector of a certain fault category is: gx (R)k)＝[Pi1,Pi2,···,Pic]，
And C, a combination of a random forest and a completely random forest in the forest of the e-th level of the cascade forest in the step 5eAnd on the training sample set R, the prediction result of the ith fault of the photovoltaic module is as follows:
and obtaining the prediction probability of the final forest to the ith fault of the photovoltaic module through the forest F of the N levels as follows:
after dynamic weighting, the maximum value is taken for processing, and the final result of the ith fault is obtained as follows: qt (C)N)＝Max(Gx(Fi))。
2. The photovoltaic module fault diagnosis method based on the dynamic weighted depth forest as claimed in claim 1, wherein: in the step 1, a photovoltaic assembly with a plurality of batteries connected in series is taken as a detection target, and 8 parameters of a maximum power point, a maximum power point voltage, a maximum power point current, an open-circuit voltage, a short-circuit current, a filling factor, a temperature and irradiance in a working state are collected to be taken as input samples.
3. The photovoltaic module fault diagnosis method based on the dynamic weighted depth forest as claimed in claim 1, wherein: the multi-granularity scanning in the step 2 specifically comprises the following steps: firstly, inputting a training sample set R with 8 acquisition parameters to a fault diagnosis system, and then performing unitized sliding sampling on the training sample set R through sampling sliding windows with three different dimensions d to obtain a new feature vector set with dimension d, wherein the number S of the vector setsiComprises the following steps: si9-d; wherein, i is 3, and d is 2, 3 and 4.
4. The photovoltaic module fault diagnosis method based on the dynamic weighted depth forest as claimed in claim 1, wherein: in the step 6, when the cascade forest is not expanded any more, finding out the stage number N corresponding to the highest prediction accuracy on the training sample set R in the expanded stage, combining the forest of the stage as a final classifier model, and using the classifier model obtained by training as a prediction model of the whole dynamic weighted depth forest on the test sample set E.
5. The photovoltaic module fault diagnosis method based on the dynamic weighted depth forest as claimed in claim 1, wherein: in the step 7, the photovoltaic module fault diagnosis types comprise normal, open-circuit fault, short-circuit fault, shadow hot spot fault and aging fault.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010021220.8A CN111245365B (en) | 2020-01-09 | 2020-01-09 | Photovoltaic module fault diagnosis method based on dynamic weighted depth forest |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010021220.8A CN111245365B (en) | 2020-01-09 | 2020-01-09 | Photovoltaic module fault diagnosis method based on dynamic weighted depth forest |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111245365A CN111245365A (en) | 2020-06-05 |
| CN111245365B true CN111245365B (en) | 2020-12-15 |
Family
ID=70876091
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010021220.8A Active CN111245365B (en) | 2020-01-09 | 2020-01-09 | Photovoltaic module fault diagnosis method based on dynamic weighted depth forest |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111245365B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111884587B (en) * | 2020-07-17 | 2021-09-14 | 特变电工新疆新能源股份有限公司 | Photovoltaic module fault diagnosis and prediction method |
| CN112116058B (en) * | 2020-09-16 | 2022-05-31 | 昆明理工大学 | Transformer fault diagnosis method for optimizing multi-granularity cascade forest model based on particle swarm algorithm |
| CN112633368B (en) * | 2020-12-21 | 2022-07-01 | 四川大学 | Flat vibration motor defect detection system and method based on improved multi-granularity cascade forest |
| CN113076975A (en) * | 2021-03-17 | 2021-07-06 | 长江水利委员会长江科学院 | Dam safety monitoring data anomaly detection method based on unsupervised learning |
| CN113591911A (en) * | 2021-06-25 | 2021-11-02 | 南京财经大学 | Cascade multi-class abnormity identification method in logistics transportation process of bulk grain container |
| CN115859196A (en) * | 2023-01-31 | 2023-03-28 | 北京智芯微电子科技有限公司 | SCGWO-DF (substation configuration bus-general purpose) -based intelligent electric meter fault diagnosis method and system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108062571A (en) * | 2017-12-27 | 2018-05-22 | 福州大学 | Diagnosing failure of photovoltaic array method based on differential evolution random forest grader |
| CN109034220A (en) * | 2018-07-13 | 2018-12-18 | 福州大学 | A kind of intelligent photovoltaic array method for diagnosing faults based on optimal rotation forest |
| CN109039281A (en) * | 2018-08-10 | 2018-12-18 | 江南大学 | A kind of diagnosing failure of photovoltaic array method based on improvement random forests algorithm |
| CN109150104A (en) * | 2018-08-10 | 2019-01-04 | 江南大学 | A kind of diagnosing failure of photovoltaic array method based on random forests algorithm |
| CN109241987A (en) * | 2018-06-29 | 2019-01-18 | 南京邮电大学 | The machine learning method of depth forest based on weighting |
| CN110224673A (en) * | 2019-05-14 | 2019-09-10 | 太原理工大学 | A kind of solar photovoltaic cell panel fault detection method based on deep learning |
| CN110568327A (en) * | 2019-09-03 | 2019-12-13 | 复旦大学 | Photovoltaic system direct current fault arc detection method based on machine learning |
-
2020
- 2020-01-09 CN CN202010021220.8A patent/CN111245365B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108062571A (en) * | 2017-12-27 | 2018-05-22 | 福州大学 | Diagnosing failure of photovoltaic array method based on differential evolution random forest grader |
| CN109241987A (en) * | 2018-06-29 | 2019-01-18 | 南京邮电大学 | The machine learning method of depth forest based on weighting |
| CN109034220A (en) * | 2018-07-13 | 2018-12-18 | 福州大学 | A kind of intelligent photovoltaic array method for diagnosing faults based on optimal rotation forest |
| CN109039281A (en) * | 2018-08-10 | 2018-12-18 | 江南大学 | A kind of diagnosing failure of photovoltaic array method based on improvement random forests algorithm |
| CN109150104A (en) * | 2018-08-10 | 2019-01-04 | 江南大学 | A kind of diagnosing failure of photovoltaic array method based on random forests algorithm |
| CN110224673A (en) * | 2019-05-14 | 2019-09-10 | 太原理工大学 | A kind of solar photovoltaic cell panel fault detection method based on deep learning |
| CN110568327A (en) * | 2019-09-03 | 2019-12-13 | 复旦大学 | Photovoltaic system direct current fault arc detection method based on machine learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111245365A (en) | 2020-06-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111245365B (en) | Photovoltaic module fault diagnosis method based on dynamic weighted depth forest | |
| CN109873610B (en) | Photovoltaic array fault diagnosis method based on IV characteristic and depth residual error network | |
| Abbassi et al. | Identification of unknown parameters of solar cell models: A comprehensive overview of available approaches | |
| Dhimish et al. | Comparing Mamdani Sugeno fuzzy logic and RBF ANN network for PV fault detection | |
| Liu et al. | Fault diagnosis approach for photovoltaic array based on the stacked auto-encoder and clustering with IV curves | |
| Zaki et al. | Deep‐learning–based method for faults classification of PV system | |
| Reddy et al. | A MATLAB based PV module models analysis under conditions of nonuniform irradiance | |
| CN110503153B (en) | Photovoltaic system fault diagnosis method based on differential evolution algorithm and support vector machine | |
| Liu et al. | On the methodology of energy yield assessment for one-Sun tandem solar cells | |
| CN113283580A (en) | Automatic fault detection method for solar cell panel | |
| CN109992911A (en) | Photovoltaic module fast modeling method based on extreme learning machine and IV characteristic | |
| Al Mansur et al. | Determination of module rearrangement techniques for non-uniformly aged PV arrays with SP, TCT, BL and HC configurations for maximum power output | |
| Elyaqouti et al. | A novel hybrid numerical with analytical approach for parameter extraction of photovoltaic modules | |
| CN116800172B (en) | Photoelectric conversion method and energy storage system | |
| Benahmida et al. | An efficient iterative method for extracting parameters of photovoltaic panels with single diode model | |
| Zhang et al. | Correntropy‐based parameter estimation for photovoltaic array model considering partial shading condition | |
| CN115833747A (en) | Photovoltaic aging detection method and device based on maximum power point tracking | |
| Kebir et al. | An overview of solar cells parameters extraction methods | |
| CN115600503A (en) | Multi-junction compound battery model parameter optimization method based on simplex-simulated annealing method | |
| CN112487347B (en) | Photovoltaic module model parameterization method considering environment and time-varying factors | |
| CN115208315A (en) | Photovoltaic intelligent fault diagnosis method based on multiple fault characteristic values | |
| Gong et al. | Fault diagnosis method of photovoltaic array based on random forest algorithm | |
| Lodhi et al. | Modelling and experimental characteristics of photovoltaic modules in typical days at an actual photovoltaic power station | |
| Jha et al. | Effect of varying temperature on different PV array configurations under partial shading condition | |
| Elgamal et al. | A Python Module for Modelling and Design of Solar Panels with Unidentical Cells |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |