CN111327271A - Photovoltaic array fault diagnosis method based on semi-supervised extreme learning machine - Google Patents
Photovoltaic array fault diagnosis method based on semi-supervised extreme learning machine Download PDFInfo
- Publication number
- CN111327271A CN111327271A CN202010064582.5A CN202010064582A CN111327271A CN 111327271 A CN111327271 A CN 111327271A CN 202010064582 A CN202010064582 A CN 202010064582A CN 111327271 A CN111327271 A CN 111327271A
- Authority
- CN
- China
- Prior art keywords
- characteristic
- semi
- photovoltaic
- equation
- learning machine
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000003745 diagnosis Methods 0.000 title claims abstract description 18
- 238000005457 optimization Methods 0.000 claims abstract description 25
- 239000002245 particle Substances 0.000 claims abstract description 18
- 238000004422 calculation algorithm Methods 0.000 claims description 29
- 235000013305 food Nutrition 0.000 claims description 22
- 238000012549 training Methods 0.000 claims description 20
- 230000006870 function Effects 0.000 claims description 15
- 241000257303 Hymenoptera Species 0.000 claims description 12
- 239000011159 matrix material Substances 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 4
- 238000005286 illumination Methods 0.000 claims description 3
- 230000001133 acceleration Effects 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 claims description 2
- 230000008859 change Effects 0.000 claims description 2
- 238000002939 conjugate gradient method Methods 0.000 claims description 2
- 230000003247 decreasing effect Effects 0.000 claims description 2
- 238000001507 sample dispersion Methods 0.000 claims description 2
- 230000032683 aging Effects 0.000 abstract description 8
- 238000010248 power generation Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 230000001965 increasing effect Effects 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 239000011449 brick Substances 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 210000003608 fece Anatomy 0.000 description 1
- 235000013312 flour Nutrition 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000010845 search algorithm Methods 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- 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
-
- 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
Abstract
The invention relates to a photovoltaic array fault diagnosis method based on a semi-supervised extreme learning machine. Firstly, acquiring an output voltage-current curve of a photovoltaic array through an acquisition device; then, extracting the characteristics of the current-voltage curve, and constructing a fitting characteristic output equation with an adjusting coefficient; secondly, solving an adjusting coefficient by adopting a nonlinear least square method based on particle swarm-confidence domain reflection optimization; obtaining a characteristic standardized equation by carrying out item shifting and standardization on the characteristic output equation; thirdly, a semi-supervised extreme learning machine based on artificial bee colony optimization is used as a classifier for photovoltaic array fault identification of a small number of labeled samples combined with a large number of unlabeled samples; and finally, regularly measuring a current-voltage curve of the normal operation of the photovoltaic array to update a standardized equation, so that the method can adapt to the natural aging of the photovoltaic array.
Description
Technical Field
The invention relates to the field of electric power and electrical equipment, in particular to a photovoltaic array fault diagnosis method based on a semi-supervised extreme learning machine.
Background
With the vigorous development of the photovoltaic industry, the construction cost of a photovoltaic system is gradually reduced, the installed capacity and the number are continuously increased, and the operation and maintenance cost is also continuously increased. Because the photovoltaic system needs to be installed in an outdoor environment with a lot of uncertain factors, the photovoltaic system is easily influenced by various environmental factors such as thermal cycle, humidity, ultraviolet rays and shadow in the operation process, various unknown faults are easily caused, the power generation efficiency is reduced, and when the faults are serious, equipment is even damaged, and serious hazards such as fire disasters are caused. The complicated outdoor environment makes the photovoltaic module easy to have electrical faults of short circuit and grounding, internal faults of photovoltaic cells such as abnormal aging and hot spots, and problems caused by external objects such as partial shading. Although the photovoltaic array can continuously operate in a fault state, the power generation efficiency is low, and the module can be irreversibly damaged even a fire disaster is caused by long-term failure removal. Therefore, the fault diagnosis technology is more important in the photovoltaic power generation system, and timely and reliable fault early warning can effectively prolong the service life and economic benefit of the solar photovoltaic power generation system. The I-V curve of the photovoltaic system contains rich information, and in recent years, the development of an embedded online I-V tracker of a photovoltaic inverter can acquire the I-V curve of each series without adding extra hardware, so that the method has important significance for fault diagnosis of a photovoltaic module. It is a hot spot of current research to extract corresponding diagnosis parameters from the I-V curve and analyze and identify faults. In an actual photovoltaic field, operators of photovoltaic systems often store a large amount of label-free historical data in a cloud, and a semi-supervised learning algorithm can establish a fault classification model by using the historical label-free data and combining a very small amount of label data, so that the photovoltaic fault diagnosis method has a good prospect. The current semi-supervised learning algorithm applied to photovoltaic faults still has defects and a great promotion space.
Disclosure of Invention
In view of the above, the present invention provides a photovoltaic array fault diagnosis method based on a semi-supervised extreme learning machine, which realizes photovoltaic array fault identification by combining a small number of labeled samples with a large number of unlabeled samples, and is suitable for natural aging of a photovoltaic array.
In order to achieve the purpose, the invention adopts the following technical scheme:
a photovoltaic array fault diagnosis method based on a semi-supervised extreme learning machine comprises the following steps:
step S1, acquiring voltage-current curves of the photovoltaic system in different fault states;
step S2, extracting characteristics according to the obtained voltage-current curves of different fault states, and constructing a fitting characteristic output equation with an adjusting coefficient;
step S3, calculating a characteristic coefficient based on a particle swarm-confidence domain reflection algorithm and a nonlinear least square method to obtain a complete photovoltaic parameter characteristic equation;
step S4, carrying out item shifting and standardization processing on the complete photovoltaic parameter characteristic equation to obtain a characteristic standardization equation;
step S5, acquiring standardized feature data of the photovoltaic system according to a feature standardized equation, and dividing the standardized feature data into a labeled training set and an unlabeled training set;
step S6, constructing and training a semi-supervised extreme learning machine for artificial bee colony optimization according to the obtained labeled training set and unlabeled training set;
and step S7, acquiring standardized characteristic data of the photovoltaic system to be tested, inputting the standardized characteristic data into a trained semi-supervised extreme learning machine for artificial bee colony optimization, and identifying faults.
Further, the step S2 is specifically:
step S21, extracting the open-circuit voltage V of the current-voltage curveocMaximum power point current ImMaximum power point voltage VmAnd an equivalent series resistance RsAs an identification feature quantity;
wherein, the equivalent series resistance RsIs defined as follows:
(I1,V1) Is the closest (0, V) in the I-V curveoc) The collection point of (1);
and step S22, setting unknown quantities a, b, c and d to represent the adjusting coefficients of the characteristic functions, wherein the characteristic formula of the photovoltaic system is as follows.
Wherein G is the measured irradiance; gstcIs 1000W/m2A constant of (d); dT is the temperature of the measured temperature minus STC; voc.f、Vm.f、Im.fAnd Rs.fRespectively representing the open-circuit voltage, the voltage and the current of a maximum power tracking point and the equivalent series resistance which are subjected to parameter fitting under different irradiance and temperature; voc.stc、Vm.stc、Im.stcAnd Rs.stcRepresenting the open circuit voltage at equivalent STC, the voltage and current at MPPT, and the series resistance, respectively.
Further, the step S3 is specifically:
in step S31, the characteristic coefficients are calculated by using a nonlinear least-squares method, which is actually to solve the problem of minimizing the sum of squares of the error vector functions, as shown in the following equation.
Wherein, F (x, G)i,Ti) For the photovoltaic signature equation to be calculated, y (G)i,Ti) As actual measured value, GiAnd TiRespectively representing the measured values of the ith irradiance and the temperature, wherein n is the number of the measured samples;
due to F (x, G)i,Ti) The method cannot obtain the parameter estimation value by a method of solving an extremum of a multivariate function like a linear least square method, and a complex optimization algorithm is needed to solve the parameter estimation value. The invention provides a nonlinear optimization search algorithm: hybrid particle swarm optimization algorithm and confidence domain reflection algorithm
And step S32, calculating a characteristic coefficient by adopting a hybrid particle swarm optimization algorithm and a trust domain reflection algorithm to obtain a complete photovoltaic parameter characteristic equation.
Further, the hybrid particle swarm optimization algorithm and the confidence domain reflection algorithm are specifically:
in each iteration of the trust domain algorithm, in a selected spherical trust domain N, a quadratic approximation model q (x) of a second-order Taylor expansion of an objective function at a current iteration point is selected, and the following formula is shown as follows:
wherein x is a group having upper and lower bounds (l)<xi<u) the vector of constraints is determined,
denotes the partial derivative, T denotes the transpose operation, and H denotes the Hessian matrix of f (x).
Then calculating an optimal trial step length s, if f (x + s) < f (x), the step is successful, replacing x with x + s, and updating the radius of the trust domain according to the following formula; otherwise, the step is unsuccessful, x is kept unchanged, and the radius of the confidence domain is updated according to the following formula; the pseudo code for trusted domain radius update is:
radius of trust domain
Let 0＜μ＜η＜1，0＜Λl＜Λu,γ1＜1＜γ2：
·Ifρk＜μ,Δk+1∈(0，γ1Δk]
·Ifρk∈(μ,η)，Δk+1∈(γ1Δk，Δk]
Ifρk＞η
IfΔk+1＞Λl,Δk+1∈[γ1Δk,Δk]orΔk+1∈[Δk,γ2Δk]
elseΔk+1∈[Δk,min(γ2Δk,Λu)].
Wherein, DeltakRepresenting confidence domain radius for the kth iteration Λl，Λl，μ，η，γ1，γ2Is a threshold degree of radius change; matrix CkAs follows:
Ck＝Dkdiag(g(xk))Jv(xk)Ck
Jv(xk)＝diag(sgn(g(xk)))
vector function v (x) ═ v1(x),…,vn(x))TIs defined as follows:
·
·
·
·
in processing the sub-problem of the confidence domain, limiting the sub-problem of the confidence domain in a 2-D sub-space S by using a subspace approximation method; the 2-D subspace S is defined by S1Sum of directions s2And the linear space formed by the direction stretch is calculated by a preconditioned conjugate gradient method. Wherein s is1The direction coincides with the gradient g, s2The direction can be approximated by Newton's direction as solved by the equation:
Hs2＝-g
or from the direction of negative curvature:
searching iterative initial values of characteristic coefficients by using a particle swarm algorithm, global variations of the particle swarm algorithm, individual optimal positions P obtained from the whole swarm so fariAnd optimum position PgMiddle learning;
let position x (t) be the parameter to be optimized and velocity v (t) be the search step; the position x (t) and velocity v (t) of each generation of particles are updated according to the following equation:
v(t+1)＝wv(t)+c1r1(Pi(t)-x(t))+c2r2(Pg(t)-x(t))
x(t+1)＝x(t)+v(t+1)
wherein t is the current iteration; c. C1And c2Is a normal number as an acceleration coefficient; r is1And r2Is [0, 1 ]]Two evenly distributed random numbers in (a); w is an inertial weight for balancing global and local search capability, decreasing linearly with evolution algebra
Wherein T is the total number of evolutions, wmaxIs set to 1, wminIs set to 0.8, therebyAnd enhancing the global searching capability.
The fitness function of particle swarm optimization is defined as:
the root mean square error objective function is minimized as:
further, the step S4 is specifically: the characteristic normalization equation obtained by shifting the formula is shown as follows,
wherein, Voc、Vm、ImAnd RsRespectively representing the output open-circuit voltage, the maximum power point voltage and current and the equivalent series resistance value of the photovoltaic system under different illumination and temperature; voc.ref、Vm.ref、Im.refAnd Rs.refRespectively representing the output open-circuit voltage, the maximum power point voltage and current and the equivalent series resistance value of the photovoltaic system under the STC; voc.norm、Vm.norm、Im.normAnd Rs.normRespectively representing the output open circuit voltage, the maximum power point voltage and current and the equivalent series resistance of the unitized photovoltaic system.
Further, the semi-supervised extreme learning machine for artificial bee colony optimization specifically comprises:
(1) the ELM is improved to SS-ELM by introducing a manifold regularization term through the Laplace operand, wherein the manifold regularization term is as follows:
wherein, wijIs pairwise similarity, L ∈ R(l+u)×(l+u)The graph Laplace matrix composed of the label-free data and the label data is shown as the formula, wherein l and u respectively represent the number of the labeled data and the label-free data; tr (×) is a trace representing the matrix;
(3) and optimizing the penalty factor by adopting an artificial bee colony algorithm to obtain the artificial bee colony optimized semi-supervised extreme learning machine.
Further, the artificial bee colony algorithm specifically comprises:
initializing the number of bee populations, evolution algebra of the bee populations, and optimizing parameters of lambda and C in SS-ELM0Real number encoding is used to reduce the dimensionality:
the location of each food source of ABC is represented by a 2-dimensional vector:
xi＝[xi1xi2]|i＝1，...，S
wherein S is the set bee population.
The initial position of the food source is randomly generated, and the upper and lower limits of the solution space are set as:
UB＝[-10,-10],LB＝[10,10]
the initial food source positions were:
xid＝LBd+(UBd-LBd)×rand(0，1)
d＝1，2 i＝1，...，S
each food source xiCorresponding to a hiring bee, the hiring bee randomly searches a neighbor for the domain to update the food source and obtain new onesPosition v ofi. If v isiLess than x of the food sourceiWhere, then viWill substitute xiOtherwise, hold xiIs not changed
vid＝xid+φid×(xid-xkd)
Wherein phi isidIs [ -1,1 [ ]]Random numbers uniformly distributed thereon; k is a randomly selected food source different from i; the fitness function for the food source is as follows:
in the formula, I (×) represents 1 when the condition (×) is satisfied, and 0 when the condition (×) is not satisfied;
and ykRespectively representing the prediction category and the original category of the label sample;
representing unlabeled sample xiThe prediction category of (1);
a prediction category representing the closest 10 sample points; coefficient afAnd bfIs a pair of constraint weights, which means that the less labeled samples are, the greater the training error weight of the labeled samples is. Coefficient cfMuch less than afAnd bfRepresenting the penalty coefficients lambda and C of the SS-ELM model with the best generalization capability in the penalty coefficients with the minimum training error and the minimum label-free sample dispersion0。
After optimization of the employed bees, the following bees select a food source according to the fitness of the bee-collecting bees and a certain probability, the specific probability calculation formula is as follows, and neighborhood search is expanded according to the corresponding reference model combination
After the optimization of the hiring bees and the following bees is finished, checking whether the food sources corresponding to the bees are unchanged within the maximum search times, namely, the solution is not updated after the iteration; for solutions that are search times out of limit, the old solution will be discarded and converted into a scout bee looking for a new food source to replace it.
Compared with the prior art, the invention has the following beneficial effects:
the invention has higher fitting precision; parameter standardization is adopted, so that the influence caused by different installation environments of a photovoltaic system and the deviation of sensor placement can be effectively avoided; and the data fitting is carried out regularly, so that the phenomenon that the parameter standardization has larger deviation due to the natural aging of the photovoltaic system can be avoided.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a best fit quantity of normal operation data in an embodiment of the present invention;
FIG. 3 shows the performance of the ABC-SSELM as mentioned in the experimental case of one embodiment of the present invention.
Detailed Description
The invention is further explained below with reference to the drawings and the embodiments.
Referring to fig. 1, the invention provides a photovoltaic array fault diagnosis method based on a semi-supervised extreme learning machine, which includes the following steps:
step S1, acquiring voltage-current curves of the photovoltaic system in different fault states;
step S2, extracting characteristics according to the obtained voltage-current curves of different fault states, and constructing a fitting characteristic output equation with an adjusting coefficient;
step S3, calculating a characteristic coefficient based on a particle swarm-confidence domain reflection algorithm and a nonlinear least square method to obtain a complete photovoltaic parameter characteristic equation;
step S4, carrying out item shifting and standardization processing on the complete photovoltaic parameter characteristic equation to obtain a characteristic standardization equation;
step S5, acquiring standardized feature data of the photovoltaic system according to a feature standardized equation, and dividing the standardized feature data into a labeled training set and an unlabeled training set;
step S6, constructing and training a semi-supervised extreme learning machine for artificial bee colony optimization according to the obtained labeled training set and unlabeled training set;
and step S7, acquiring standardized characteristic data of the photovoltaic system to be tested, inputting the standardized characteristic data into a trained semi-supervised extreme learning machine for artificial bee colony optimization, and identifying faults.
The present invention will be further described with reference to the following examples.
In the embodiment, a 3.51kW photovoltaic system with two times of series connection of 13 polycrystalline photovoltaic modules of GTEC-260G6M6A is selected; selecting a photovoltaic array analysis system with the model of Prova 1011; the computer software was programmed by MATLAB:
in this embodiment, the photovoltaic system conditions studied include: a normal state; short-circuit failure: the taps of the two photovoltaic module outgoing lines MC4 are adopted for short circuit to form a short circuit loop, and the module in the middle of the two taps is short-circuited; incomplete shading: broken small sundries such as small bricks are adopted to shield a certain photovoltaic module, flour is adopted to simulate bird droppings to shield, the shielding area is small, and bypass diodes of the photovoltaic module substrings are not enough to conduct; completely shading the shade: adopt film and paper piece to shelter from a certain photovoltaic module, simulate comparatively serious shade, shade this moment is comparatively serious, and this photovoltaic module's bypass diode is all in the conducting state. Abnormal aging: the sliding rheostat is adopted as an aging resistor to be connected in series in the photovoltaic sub-assembly, so that abnormal aging faults of the photovoltaic module are simulated to be used as the aging resistor to be connected in series in the photovoltaic sub-assembly. The experimental model is mainly embodied on the shielding degree of the whole assembly for the simulation of complete shading, and is mainly embodied on the shielding degree of the substring for the simulation of incomplete shading. Irradiance range in experimental environment is about 150W/m2To 1000W/m2。
In this embodiment, the specific implementation process of identification includes the following steps:
acquiring an I-V curve signal of a photovoltaic system: the current I-V curves were set to be automatically collected every minute, with 150 data points per curve. The collection time per day was 8 in the morning: 00 to 5 in the afternoon: 00.
and (3) fitting a characteristic output equation: and selecting photovoltaic characteristic parameters with different illumination intensities and temperatures under normal operation to construct a photovoltaic output characteristic equation. Calling an lsqcurvefit function of Matlab, setting the attribute of the algorithm as confidence domain reflection, optimizing an initial value input into the confidence domain algorithm through a particle swarm optimization algorithm, and fitting and solving the characteristic coefficient of the characteristic equation through a nonlinear least square method. And then shifting the parameter characteristic equation to obtain a parameter standardization equation, and converting the characteristic parameters under different irradiances and temperatures into parameter values under equivalent STC. And finally, normalizing by using the characteristic parameter value obtained under the STC as a reference value to obtain the final 4 photovoltaic identification characteristic quantities.
Due to the complex environmental factors, the signal conversion of the acquisition equipment has deviation, and the noise in the measured data is inevitable. In the data preprocessing of fitting the feature coefficients, the result of the fitting depends on the given number of samples, and the fitting result is influenced by insufficient data amount. In order to find the minimum number of samples required for normal operation of the photovoltaic system, the total measured normal data is divided into four levels of irradiance, G<300W/m2、300W/m2<G<500W/m2、500W/m2<G<700W/m2、G>700W/m2. Then, the same number of samples n are randomly selected from each level in turn, fitting is performed 100 times, respectively, and the fitting result is evaluated using the remaining normal samples. The fit results for each parameter were evaluated as the mean of 100 RMSEs, with the largest mean being the reference for the fit evaluation, as shown in fig. 2. It can be seen from fig. 2 that the fitting result tends to be stable after n is 7. Thus, at different irradiance, only 28 normal samples of the photovoltaic system are required for parameter normalization. In order to make the fitting result more reliable, n is 10, i.e. 40 normal samples are used for parameter normalization.
Semi-supervised extreme learning machine for artificial bee colony optimization: and randomly acquiring different quantities of historical label-free numbers from the data set every time, randomly acquiring labeled data in the rest data sets, increasing the quantity of each type of label samples from 1 one by one, and taking the rest data as a test sample set. And, each experiment was run 50 times separately, and the results were taken as the average accuracy. The total number of samples measured was 1615. The classification result of the measured data of the photovoltaic module is shown in fig. 2, and "UL" represents the amount of non-label data. With the increase of the non-label data, the accuracy and the stability of fault identification are improved. And as the number of labeled samples increases, the accuracy of the test increases rapidly. When the number of the labeled samples reaches 25, the classification result tends to be stable, and the average accuracy of the judgment of the operation simulation of the 5 photovoltaic systems is more than 98%.
The above description is only a preferred embodiment of the present invention, and all equivalent changes and modifications made in accordance with the claims of the present invention should be covered by the present invention.
Claims (7)
1. A photovoltaic array fault diagnosis method based on a semi-supervised extreme learning machine is characterized by comprising the following steps:
step S1, acquiring voltage-current curves of the photovoltaic system in different fault states;
step S2, extracting characteristics according to the obtained voltage-current curves of different fault states, and constructing a fitting characteristic output equation with an adjusting coefficient;
step S3, calculating a characteristic coefficient based on a particle swarm-confidence domain reflection algorithm and a nonlinear least square method to obtain a complete photovoltaic parameter characteristic equation;
step S4, carrying out item shifting and standardization processing on the complete photovoltaic parameter characteristic equation to obtain a characteristic standardization equation;
step S5, acquiring standardized feature data of the photovoltaic system according to a feature standardized equation, and dividing the standardized feature data into a labeled training set and an unlabeled training set;
step S6, constructing and training a semi-supervised extreme learning machine for artificial bee colony optimization according to the obtained labeled training set and unlabeled training set;
and step S7, acquiring standardized characteristic data of the photovoltaic system to be tested, inputting the standardized characteristic data into a trained semi-supervised extreme learning machine for artificial bee colony optimization, and identifying faults.
2. The photovoltaic array fault diagnosis method based on the semi-supervised extreme learning machine as claimed in claim 1, wherein the step S2 is specifically as follows:
step S21, extracting the open-circuit voltage V of the current-voltage curveocMaximum power point current ImMaximum power point voltage VmAnd an equivalent series resistance RsAs an identification feature quantity;
wherein, the equivalent series resistance RsIs defined as follows:
(I1,V1) Is the closest (0, V) in the I-V curveoc) The collection point of (1);
step S22, setting unknown quantity a1、a2、a3、b1、b2、b3、c1、c2、c3And d1、d2、d3The adjustment coefficient of the characteristic function is represented, and the photovoltaic system characteristic formula is shown as follows.
Wherein G is the measured irradiance; gstcIs 1000W/m2A constant of (d); dT is the temperature of the measured temperature minus STC; voc.f、Vm.f、Im.fAnd Rs.fRespectively representing the open-circuit voltage, the voltage and the current of a maximum power tracking point and the equivalent series resistance which are subjected to parameter fitting under different irradiance and temperature; voc.stc、Vm.stc、Im.stcAnd Rs.stcRepresenting the open circuit voltage at equivalent STC, the voltage and current at MPPT, and the series resistance, respectively.
3. The photovoltaic array fault diagnosis method based on the semi-supervised extreme learning machine as claimed in claim 1, wherein the step S3 is specifically as follows:
step S31, calculating characteristic coefficient by using nonlinear least square method as shown in the following formula
Wherein, F (x, G)i,Ti) For the photovoltaic signature equation to be calculated, y (G)i,Ti) As actual measured value, GiAnd TiRespectively representing the measured values of the ith irradiance and the temperature, wherein n is the number of the measured samples;
and step S32, calculating a characteristic coefficient by adopting a hybrid particle swarm optimization algorithm and a trust domain reflection algorithm to obtain a complete photovoltaic parameter characteristic equation.
4. The photovoltaic array fault diagnosis method based on the semi-supervised extreme learning machine as recited in claim 3, wherein the hybrid particle swarm optimization algorithm and the confidence domain reflection algorithm are specifically as follows:
in each iteration of the trust domain algorithm, in a selected spherical trust domain N, a quadratic approximation model q (x) of a second-order Taylor expansion of an objective function at a current iteration point is selected, and the following formula is shown as follows:
wherein x is a group having upper and lower bounds (l)<xi<u) the vector of constraints is determined,
denotes the partial derivative, T denotes the transpose operation, H denotes the Hessian matrix of f (x);
then calculating an optimal trial step length s, if f (x + s) < f (x), the step is successful, replacing x with x + s, and updating the radius of the trust domain according to the following formula; otherwise, the step is unsuccessful, x is kept unchanged, and the radius of the confidence domain is updated according to the following formula;
radius of trust domain:
setting 0 & mu & lt η & lt 1, 0 & lt Λl＜Λu,γ1＜1＜γ2：
Ifρk＜μ,Δk+1∈(0，γ1Δk]
Ifρk∈(μ,η)，Δk+1∈(γ1Δk，Δk]
Ifρk＞η
IfΔk+1＞Λl,Δk+1∈[γ1Δk,Δk]orΔk+1∈[Δk,γ2Δk]
elseΔk+1∈[Δk,min(γ2Δk,Λu)].
Wherein, DeltakRepresenting confidence domain radius for the kth iteration Λl，Λl，μ，η，γ1，γ2Is a threshold degree of radius change; matrix CkAs follows:
Ck＝Dkdiag(g(xk))Jv(xk)Ck
Jv(xk)＝diag(sgn(g(xk)))
vector function v (x) ═ v1(x),…,vn(x))TIs defined as follows:
in processing the sub-problem of the confidence domain, limiting the sub-problem of the confidence domain in a 2-D sub-space S by using a subspace approximation method; the 2-D subspace S is defined by S1Sum of directions s2The linear space formed by the direction stretch is calculated by a preconditioned conjugate gradient method; wherein s is1The direction coincides with the gradient g, s2The direction can be approximated by Newton's direction as solved by the equation:
Hs2＝-g
or from the direction of negative curvature:
searching iterative initial values of characteristic coefficients by using a particle swarm algorithm, global variations of the particle swarm algorithm, individual optimal positions P obtained from the whole swarm so fariAnd optimum position PgMiddle learning;
let position x (t) be the parameter to be optimized and velocity v (t) be the search step; the position x (t) and velocity v (t) of each generation of particles are updated according to the following equation:
v(t+1)＝wv(t)+c1r1(Pi(t)-x(t))+c2r2(Pg(t)-x(t))
x(t+1)＝x(t)+v(t+1)
wherein t is the current iteration; c. C1And c2Is a normal number as an acceleration coefficient; r is1And r2Is [0, 1 ]]Two evenly distributed random numbers in (a); w is an inertial weight for balancing global and local search capability, decreasing linearly with evolution algebra
Wherein T is the total number of evolutions, wmaxIs set to 1, wminSet to 0.8 to enhance global search capability.
The fitness function of particle swarm optimization is defined as:
the root mean square error objective function is minimized as:
5. the photovoltaic array fault diagnosis method based on the semi-supervised extreme learning machine as claimed in claim 1, wherein the step S4 is specifically as follows: the characteristic normalization equation obtained by shifting the formula is shown as follows,
wherein, Voc、Vm、ImAnd RsRespectively representing the output open-circuit voltage, the maximum power point voltage and current and the equivalent series resistance value of the photovoltaic system under different illumination and temperature; voc.ref、Vm.ref、Im.refAnd Rs.refRespectively representing the output open-circuit voltage, the maximum power point voltage and current and the equivalent series resistance value of the photovoltaic system under the STC; voc.norm、Vm.norm、Im.normAnd Rs.normRespectively representing the output open circuit voltage, the maximum power point voltage and current and the equivalent series resistance of the unitized photovoltaic system.
6. The photovoltaic array fault diagnosis method based on the semi-supervised extreme learning machine as claimed in claim 1, wherein the semi-supervised extreme learning machine for artificial bee colony optimization is specifically:
(1) the ELM is improved to SS-ELM by introducing a manifold regularization term through the Laplace operand, wherein the manifold regularization term is as follows:
wherein, wijIs pairwise similarity, L ∈ R(l+u)×(l+u)The graph Laplace matrix composed of the label-free data and the label data is shown as the formula, wherein l and u respectively represent the number of the labeled data and the label-free data; tr (×) is a trace representing the matrix;
(2) and optimizing the penalty factor by adopting an artificial bee colony algorithm to obtain the artificial bee colony optimized semi-supervised extreme learning machine.
7. The photovoltaic array fault diagnosis method based on the semi-supervised extreme learning machine as recited in claim 6, wherein the artificial bee colony algorithm is specifically as follows:
initializing the number of bee populations, evolution algebra of the bee populations, and optimizing parameters of lambda and C in SS-ELM0Real number encoding is used to reduce the dimensionality:
the location of each food source of ABC is represented by a 2-dimensional vector:
xi＝[xi1xi2]|i＝1，...，S
in the formula, S is the set bee population number;
the initial position of the food source is randomly generated, and the upper and lower limits of the solution space are set as:
UB＝[-10,-10],LB＝[10,10]
the initial food source positions were:
xid＝LBd+(UBd-LBd)×rand(0，1)
d＝1，2 i＝1，...，S
each food source xiFor a bee, the bee randomly searches a neighborhood for updating the food source to obtain a new location vi. If v isiLess than x of the food sourceiWhere, then viWill substitute xiOtherwise, hold xiIs not changed
vid＝xid+φid×(xid-xkd)
Wherein phi isidIs [ -1,1 [ ]]Random numbers uniformly distributed thereon; k is a randomly selected food source different from i; the fitness function for the food source is as follows:
in the formula, I (×) represents 1 when the condition (×) is satisfied, and 0 when the condition (×) is not satisfied;
and ykRespectively representing the prediction category and the original category of the label sample;
representing unlabeled sample xiThe prediction category of (1);
a prediction category representing the closest 10 sample points; coefficient afAnd bfThe training error weight of the labeled sample is a pair of constraint weights, which represents that the training error weight of the labeled sample is larger when the number of labeled samples is less; coefficient cfMuch less than afAnd bfRepresenting the penalty coefficients lambda and C of the SS-ELM model with the best generalization capability in the penalty coefficients with the minimum training error and the minimum label-free sample dispersion0；
After optimization of the employed bees, the following bees select a food source according to the fitness of the bee-collecting bees and a certain probability, the specific probability calculation formula is as follows, and neighborhood search is expanded according to the corresponding reference model combination
After the optimization of the hiring bees and the following bees is finished, checking whether the food sources corresponding to the bees are unchanged within the maximum search times, namely, the solution is not updated after the iteration; for solutions that are search times out of limit, the old solution will be discarded and converted into a scout bee looking for a new food source to replace it.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010064582.5A CN111327271B (en) | 2020-01-20 | 2020-01-20 | Photovoltaic array fault diagnosis method based on semi-supervised extreme learning machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010064582.5A CN111327271B (en) | 2020-01-20 | 2020-01-20 | Photovoltaic array fault diagnosis method based on semi-supervised extreme learning machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111327271A true CN111327271A (en) | 2020-06-23 |
| CN111327271B CN111327271B (en) | 2021-11-26 |
Family
ID=71170929
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010064582.5A Active CN111327271B (en) | 2020-01-20 | 2020-01-20 | Photovoltaic array fault diagnosis method based on semi-supervised extreme learning machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111327271B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113203953A (en) * | 2021-04-02 | 2021-08-03 | 中国人民解放军92578部队 | Lithium battery residual service life prediction method based on improved extreme learning machine |
| CN113392936A (en) * | 2021-07-09 | 2021-09-14 | 四川英创力电子科技股份有限公司 | Oven fault diagnosis method based on machine learning |
| CN113515851A (en) * | 2021-06-07 | 2021-10-19 | 淮阴工学院 | Photovoltaic fault detection characteristic quantity extraction method |
| WO2022054954A1 (en) * | 2020-09-14 | 2022-03-17 | ヒラソルエナジー株式会社 | Method, system, and program for estimating power generation characteristics of solar cell. |
| CN115238472A (en) * | 2022-06-30 | 2022-10-25 | 华能安源发电有限责任公司 | Fan state overhauling method and device, electronic equipment and computer readable medium |
| CN115455730A (en) * | 2022-09-30 | 2022-12-09 | 南京工业大学 | Photovoltaic module hot spot fault diagnosis method based on complete neighborhood preserving embedding |
| CN115955193A (en) * | 2023-03-13 | 2023-04-11 | 国网江西省电力有限公司电力科学研究院 | Photovoltaic cell fault judgment method and system based on I-V curve reverse thrust |
| CN116627847A (en) * | 2023-07-21 | 2023-08-22 | 山东云科汉威软件有限公司 | Browser compatibility testing method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN109766952A (en) * | 2019-01-21 | 2019-05-17 | 福州大学 | Photovoltaic array fault detection method based on Partial Least Squares and extreme learning machine |
| US20190312547A1 (en) * | 2018-04-09 | 2019-10-10 | International Business Machines Corporation | Internet of Things-Enabled Solar Photovoltaic Health Monitoring and Advising Related Thereto |
| CN110555474A (en) * | 2019-08-28 | 2019-12-10 | 上海电力大学 | photovoltaic panel fault detection method based on semi-supervised learning |
-
2020
- 2020-01-20 CN CN202010064582.5A patent/CN111327271B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190312547A1 (en) * | 2018-04-09 | 2019-10-10 | International Business Machines Corporation | Internet of Things-Enabled Solar Photovoltaic Health Monitoring and Advising Related Thereto |
| 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 |
| CN109766952A (en) * | 2019-01-21 | 2019-05-17 | 福州大学 | Photovoltaic array fault detection method based on Partial Least Squares and extreme learning machine |
| CN110555474A (en) * | 2019-08-28 | 2019-12-10 | 上海电力大学 | photovoltaic panel fault detection method based on semi-supervised learning |
Non-Patent Citations (1)
| Title |
|---|
| 张振良等: "基于半监督迁移学习的轴承故障诊断方法", 《北京航空航天大学学报》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022054954A1 (en) * | 2020-09-14 | 2022-03-17 | ヒラソルエナジー株式会社 | Method, system, and program for estimating power generation characteristics of solar cell. |
| CN113203953A (en) * | 2021-04-02 | 2021-08-03 | 中国人民解放军92578部队 | Lithium battery residual service life prediction method based on improved extreme learning machine |
| CN113515851A (en) * | 2021-06-07 | 2021-10-19 | 淮阴工学院 | Photovoltaic fault detection characteristic quantity extraction method |
| CN113515851B (en) * | 2021-06-07 | 2023-12-01 | 淮阴工学院 | Photovoltaic fault detection characteristic quantity extraction method |
| CN113392936A (en) * | 2021-07-09 | 2021-09-14 | 四川英创力电子科技股份有限公司 | Oven fault diagnosis method based on machine learning |
| CN115238472A (en) * | 2022-06-30 | 2022-10-25 | 华能安源发电有限责任公司 | Fan state overhauling method and device, electronic equipment and computer readable medium |
| CN115455730A (en) * | 2022-09-30 | 2022-12-09 | 南京工业大学 | Photovoltaic module hot spot fault diagnosis method based on complete neighborhood preserving embedding |
| CN115455730B (en) * | 2022-09-30 | 2023-06-20 | 南京工业大学 | Photovoltaic module hot spot fault diagnosis method based on complete neighborhood preserving embedding |
| CN115955193A (en) * | 2023-03-13 | 2023-04-11 | 国网江西省电力有限公司电力科学研究院 | Photovoltaic cell fault judgment method and system based on I-V curve reverse thrust |
| CN115955193B (en) * | 2023-03-13 | 2023-08-15 | 国网江西省电力有限公司电力科学研究院 | Photovoltaic cell fault judging method and system based on I-V curve reverse push |
| CN116627847A (en) * | 2023-07-21 | 2023-08-22 | 山东云科汉威软件有限公司 | Browser compatibility testing method and system |
| CN116627847B (en) * | 2023-07-21 | 2023-10-10 | 山东云科汉威软件有限公司 | Browser compatibility testing method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111327271B (en) | 2021-11-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111327271B (en) | Photovoltaic array fault diagnosis method based on semi-supervised extreme learning machine | |
| Abbassi et al. | Identification of unknown parameters of solar cell models: A comprehensive overview of available approaches | |
| Zhu et al. | Fault diagnosis approach for photovoltaic arrays based on unsupervised sample clustering and probabilistic neural network model | |
| Basnet et al. | An intelligent fault detection model for fault detection in photovoltaic systems | |
| Abbassi et al. | An improved single-diode model parameters extraction at different operating conditions with a view to modeling a photovoltaic generator: A comparative study | |
| Liu et al. | Fault diagnosis approach for photovoltaic array based on the stacked auto-encoder and clustering with IV curves | |
| Ge et al. | A hybrid model for short-term PV output forecasting based on PCA-GWO-GRNN | |
| Huang et al. | Newly-designed fault diagnostic method for solar photovoltaic generation system based on IV-curve measurement | |
| Ibrahim et al. | An optimized offline random forests-based model for ultra-short-term prediction of PV characteristics | |
| Chen et al. | Rapid and accurate modeling of PV modules based on extreme learning machine and large datasets of IV curves | |
| CN109672406B (en) | Photovoltaic power generation array fault diagnosis and classification method based on sparse representation and SVM | |
| CN111444615B (en) | Photovoltaic array fault diagnosis method based on K nearest neighbor and IV curve | |
| CN110929953A (en) | Photovoltaic power station ultra-short term output prediction method based on cluster analysis | |
| CN110503153B (en) | Photovoltaic system fault diagnosis method based on differential evolution algorithm and support vector machine | |
| Ma et al. | Novel field-support vector regression-based soft sensor for accurate estimation of solar irradiance | |
| Liu et al. | Intelligent fault diagnosis of photovoltaic array based on variable predictive models and I–V curves | |
| Omer et al. | Comparison of decision tree based ensemble methods for prediction of photovoltaic maximum current | |
| CN111985678A (en) | Photovoltaic power short-term prediction method | |
| Laudani et al. | An analytical approach for maximum power point calculation for photovoltaic system | |
| Mellit et al. | Handbook of Artificial Intelligence Techniques in Photovoltaic Systems: Modeling, Control, Optimization, Forecasting and Fault Diagnosis | |
| CN114117921A (en) | Intelligent diagnosis method for faults of photovoltaic array | |
| Cotfas et al. | Analytical versus metaheuristic methods to extract the photovoltaic cells and panel parameters | |
| CN113595132A (en) | Photovoltaic online parameter identification method based on global maximum power point tracking and hybrid optimization algorithm | |
| Vinod | Estimation of parameters for one diode solar PV cell using grey wolf optimizer to obtain exact VI characteristics | |
| CN112487347B (en) | Photovoltaic module model parameterization method considering environment and time-varying factors |
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 |