CN109992911A - Photovoltaic module fast modeling method based on extreme learning machine and IV characteristic - Google Patents
Photovoltaic module fast modeling method based on extreme learning machine and IV characteristic Download PDFInfo
- Publication number
- CN109992911A CN109992911A CN201910278837.5A CN201910278837A CN109992911A CN 109992911 A CN109992911 A CN 109992911A CN 201910278837 A CN201910278837 A CN 201910278837A CN 109992911 A CN109992911 A CN 109992911A
- Authority
- CN
- China
- Prior art keywords
- curve
- point
- data
- voltage
- photovoltaic module
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Testing Of Individual Semiconductor Devices (AREA)
- Photovoltaic Devices (AREA)
Abstract
The present invention relates to the photovoltaic module fast modeling methods based on extreme learning machine and IV characteristic, comprising the following steps: step S1: the IV characteristic curve data of acquisition photovoltaic module under various environmental conditions;Step S2: according to IV characteristic curve Data Data, IV characteristic curve is obtained;Step S3: the IV curve data of rejecting abnormalities obtains normal IV curve data collection;Step S4: according to the irradiation level of IV characteristic curve data and component backboard temperature, mesh extraction is carried out to normal IV curve data collection, acquires the IV curve data collection of the various operating conditions of uniform fold, and be randomly divided into independent training curve collection and test curve collection;Step S5 obtains the training data point set and test data point set by each data point.Step S6: the point of the training data according to obtained in S5 establishes photovoltaic module model.The present invention can accurate and effective rapidly the IV characteristic of photovoltaic module is modeled, model training speed is fast, and have good accuracy and generalization ability.
Description
Technical field
The invention belongs to solar battery and the modeling techniques of photovoltaic power generation array, and in particular to one kind is learnt based on the limit
The photovoltaic module fast modeling method of machine and IV characteristic.
Background technique
The Accurate Model of photovoltaic module and array plays a significant role the generating efficiency for optimizing photovoltaic plant.However,
It due to photovoltaic array and power station installation and works in severe outdoor environment, while vulnerable to thermal cycle, humidity when its work,
Ultraviolet light, the influence of the various environmental factors such as wind exciting lead to the failures such as local ageing, performance decline, the crackle of material, greatly
Ground with influencing photovoltaic array electrical characteristic, and then the generating efficiency of influence photovoltaic plant.Therefore it provides a kind of precise and high efficiency is reliable
Photovoltaic module/array model maximum power point of photovoltaic array is tracked, fault detection, power prediction etc. all has very heavy
The effect wanted.In addition, the rapid growth with photovoltaic power generation installation amount in worldwide, for photovoltaic module/array
The reliable modeling method of precise and high efficiency has been subjected to the extensive research of domestic and foreign scholars.
Presently, there are photovoltage model modeling method can substantially be divided into two types, i.e. the whitepack model based on equivalent circuit
With the black-box model based on data-driven.Whitepack model based on equivalent circuit mainly uses mono-/bis-diode model, these
Model finds corresponding equivalent circuit equation by I-V characteristic curve, by finding the matched curve and actual measurement song that equation obtains
Between line when root-mean-square error minimum, internal parameter value, it is accurate to obtain it then to substitute into obtained optimized parameter
Photovoltage model.Although this method can have very high accuracy on standard set, the accuracy of its model is great
Dependent on the parameter values inside formula, and these parameters are easily influenced by environmental factor.Therefore, to each operating condition of acquisition
Under accurate model, to the I-V characteristic curve under each operating condition carry out parameter extraction, cause photovoltaic modeling it is cumbersome and low
Effect.But if the model parameter that the extraction of characteristic condition is used only can bring the inaccurate of model parameter as general model parameter
Really.In conclusion solving the problems, such as existing whitepack model, the black-box model of data-driven has received people and has widely closed
Note, common method have, BP neural network, multi-layer perception (MLP) (MLP), General Neural Network (GRNN) etc., although these algorithms
It can predict corresponding I-V characteristic curve, but all have that the network model training time is long, and accuracy is low mostly, Generalization Capability
The problems such as poor.For this purpose, this paper presents a kind of photovoltaic module fast modeling method based on extreme learning machine and IV characteristic, passes through
Experimental comparison, this method have shorter model training time, stronger Generalization Capability compared to traditional machine learning algorithm
Higher accuracy.
Summary of the invention
In view of this, the purpose of the present invention is to provide a kind of photovoltaic module based on extreme learning machine and IV characteristic is quick
Modeling method, so that the model training time be greatly shortened, improves the essence of photovoltage model to overcome the defect of existing the relevant technologies
Degree and generalization ability.
To achieve the above object, the present invention adopts the following technical scheme:
A kind of photovoltaic module fast modeling method based on extreme learning machine and IV characteristic, comprising the following steps:
Step S1: IV characteristic curve tester is used, IV characteristic curve data, irradiation level and the component of photovoltaic module are acquired
Backboard temperature;
Step S2: to the IV characteristic curve data of photovoltaic module, the resampling of data point interpolation is carried out, it is bent to obtain IV characteristic
Line;
Step S3: it based on single diode five-parameter model, is carried out curve fitting using optimization algorithm and obtains fitting root mean square
Error judges that abnormal curve, the IV curve data of rejecting abnormalities obtain normal IV curve data according to fitting root-mean-square error
Collection;
Step S4: according to the characteristic irradiation level of IV and component backboard temperature, normal IV curve data collection is carried out
Mesh extraction, acquires the IV curve data collection of the various operating conditions of uniform fold, and is randomly divided into independent training curve collection and survey
Try curve set;
Step S5: training curve collection and test curve are concentrated to data point voltage, electric current and the irradiation level, temperature of every curve
Degree is combined, and obtains the training data point set and test number being made of the voltage of each data point, electric current, irradiation level and temperature
Strong point collection.
Step S6: being input, component electricity with irradiation level, component backboard temperature and component voltage according to training data point set
Stream is output, using extreme learning machine fitting algorithm training Single hidden layer feedforward neural networks, establishes photovoltaic module model;
Step S7: test data point set is input to the extreme learning machine photovoltaic module model that training obtains in S6, is obtained
The current value that model calculates, and be compared with measured current value, to test and assess extreme learning machine photovoltaic module model
Error.
Further, the step S2 specifically:
Step S21: from IV characteristic curve data, the open-circuit voltage V of the IV characteristic curve data of photovoltaic module is foundocWith
Short circuit current Isc;
Step S22: in [0 Voc] voltage range in uniformly obtain the voltage of N number of resampling point, adjacent resampling point
Interval voltage isAnd record the resampling voltage vector [V1, V2..., Vc..., VN], wherein VcIndicate c-th of voltage weight
The voltage of sampled point.
Step S23: it is directed to each resampling point Vc, c value range [1N], find original IV characteristic curve data in resampling
The voltage and current value of the left and right adjacent two data point of point, i.e. Vc-1And Vc+1And corresponding current value Ic-1And Ic+1;
Step S24: linear interpolation method is utilized, each electrical voltage point V is calculatedcCorresponding electric current Ic, specific to calculate
Method are as follows:To obtain N number of voltage resampling point (Vc,Ic)
Step S25: in [0Isc] electric current section in uniformly obtain the electric current of M resampling point, adjacent resampling point interval
Electric current isAnd record the resampling current vector [I1, I2..., Id..., IM], wherein IdIndicate d-th of electric current resampling
The electric current of point.
Step S26: it is directed to each electric current resampling point Id, d value range [1M], find original IV characteristic curve data in weigh
The current value I of the left and right adjacent two data point of sampled pointd-1And Id+1And corresponding voltage value Vd-1And Vd+1;
Step S27: linear interpolation method is utilized, each electric current resampling point I is calculateddCorresponding voltage value Vd, specific to count
Calculation method are as follows:To obtain M electric current resampling point (Vd,Id)。
Step S28: the M electric current resampling point that the S24 N number of voltage resampling point obtained and S27 are obtained, according to voltage
Ascending sequence merges, and obtains the resampling IV characteristic curve comprising M+N data point.
Further, the curve-fitting method is specifically, using based on a kind of ADAPTIVE MIXED list based on hawk strategy
The photovoltage model parameter extracting method of pure shape extracts five parameters of single diode photovoltage model of each curve under each operating condition
Value, and the root-mean-square error RMSE between the curve and measured curve being fitted under five Parameter Conditions is calculated, it is specific to count
Calculation method is as follows:
Wherein I is the current value of actual measurement,It is the current value of prediction, N is the number of whole I-V curve up-sampling point, and
RMSE is greater than acceptable threshold value RMSETCurve regard as abnormal curve and rejected, obtain normal IV curve data.
Further, the step S4 specifically:
Step S41: to IV characteristic curve Data Data, the extraction range and sampling interval that grid is sampled are set;
Step S42: grid is successively selected, and is counted to sample is distributed in it;
Step S43: if sample size is less than maximum sampling number in the grid, all samples in the grid are randomly selected
70% be used as training sample, 30% be used as test sample, wherein 90% in training sample be used for hands-on, 10% conduct
Verifying collection;
Step S44: if quantity is greater than maximum number of samples in grid, first from the maximum sampling of selection in sample in the grid
A several samples, then randomly select all samples in the grid 70% is used as training sample, wherein in training sample
90% be used for hands-on, 10% as verifying collection;
Step S45: repeat the above steps S42 to S44, is defeated by obtained sample process until taking all grids
Enter for (G, T, V), export the data set for I, wherein G and T is irradiation level and temperature, V and I be voltage and corresponding electric current to
Amount;
Step S46: the test set returned, verifying collection and training set.
Further, the step S5 specifically:
Step S51: the IV curved line number for including in training set and test set obtained in statistic procedure S4 respectively;
Step S52: to each training set IV curve, its irradiation level, component backboard temperature, each electrical voltage point and electricity are taken
Flow point is as training data point set;
Step S53: to each test set IV curve, its irradiation level, component backboard temperature, each electrical voltage point and electricity are taken
Flow point is as test data point set.
Further, the step S6 specifically:
Step S61: the training data point set for taking step S5 to obtain, with the irradiation level of each data point, component backboard temperature,
The voltage of component is the input of training data, using its corresponding electric current as the output of training data;
Step S62: Single hidden layer feedforward neural networks model is established;
Step S63: the input layer of random initializtion neural network and the connection weight w and hidden layer neuron of hidden layer
Threshold value b, connection weight between hidden layer and output layer is β, indicate that w and β are as follows with matrix:
Threshold value b=[the b of hidden layer neuron1, b2, b3..., bl]T
Step S64: assuming that hidden layer neuron activation function is g (x), then input matrix X is by after neural network
First output isIt is found that when neural network inputs j-th of sample, the prediction of network output
Value is
For the hidden layer output matrix of neural network, the predicted value of network output is T=[t1, t2..., tQ];
Step S65: operating limit learning machine fitting algorithm calculates the hidden layer and output layer of Single hidden layer feedforward neural networks
Weight β calculates to obtain β=H by least square method+T ' thus obtains network model, and wherein parameter is w, b and β.
Further, the step S7 specifically: test data point set obtained in step S5 is taken, by irradiation level, component
Backboard temperature and voltage value are input to the extreme learning machine photovoltaic module model that training obtains in step S6, obtain model prediction
Current value, and be compared with the current value of actual measurement.With between the current value of actual measurement and the current value of model prediction
Error assess the accuracy of extreme learning machine photovoltaic module model.
Compared with the prior art, the invention has the following beneficial effects:
When the present invention has shorter model training compared to traditional machine learning algorithm by using extreme learning machine
Between, higher accuracy and stronger generalization ability.And using the method for a kind of pair of Current Voltage while uniform sampling, sufficiently
The generality of sampled point is considered, so that model has higher precision of prediction.Compared with existing modeling method, the present invention is big
The model training time is shortened greatly, there is better Generalization Capability and improves the accuracy of model prediction.
Detailed description of the invention
Fig. 1 is overview flow chart of the invention;
Fig. 2 is the flow chart of the resampling and abnormity removing in one embodiment of the invention;
Fig. 3 is Single hidden layer feedforward neural networks structure chart proposed in one embodiment of the invention;
Fig. 4 is the modeling training flow chart in the present invention based on extreme learning machine;
Fig. 5 is the prediction result in one embodiment of the invention to CdTe75669 component.
Specific embodiment
The present invention will be further described with reference to the accompanying drawings and embodiments.
Fig. 1 is please referred to, the present invention provides a kind of photovoltaic module fast modeling method based on extreme learning machine and IV characteristic,
Flow diagram is as shown in Figure 1, it specifically comprises the following steps:
Step S1: IV characteristic curve tester is used, IV characteristic curve data, irradiation level and the component of photovoltaic module are acquired
Backboard temperature;
Step S2: to the IV characteristic curve data of photovoltaic module, the resampling of data point interpolation is carried out, it is bent to obtain IV characteristic
Line;
Step S3: it based on single diode five-parameter model, is carried out curve fitting using optimization algorithm and obtains fitting root mean square
Error judges that abnormal curve, the IV curve data of rejecting abnormalities obtain normal IV curve data according to fitting root-mean-square error
Collection;
Step S4: according to the characteristic irradiation level of IV and component backboard temperature, normal IV curve data collection is carried out
Mesh extraction, acquires the IV curve data collection of the various operating conditions of uniform fold, and is randomly divided into independent training curve collection and survey
Try curve set;
Step S5: training curve collection and test curve are concentrated to data point voltage, electric current and the irradiation level, temperature of every curve
Degree is combined, and obtains the training data point set and test number being made of the voltage of each data point, electric current, irradiation level and temperature
Strong point collection.
Step S6: being input, component electricity with irradiation level, component backboard temperature and component voltage according to training data point set
Stream is output, using extreme learning machine fitting algorithm training Single hidden layer feedforward neural networks, establishes photovoltaic module model;
Step S7: test data point set is input to the extreme learning machine photovoltaic module model that training obtains in S6, is obtained
The current value that model calculates, and be compared with measured current value, to test and assess extreme learning machine photovoltaic module model
Error.
Such as Fig. 2, for a kind of photovoltaic module fast modeling method embodiment based on extreme learning machine and IV characteristic of proposition
In the bilinearity method of sampling flow chart, detailed process is divided into two parts, obtains largely surveying IV characteristic curve first
Data carry out resampling, detailed process are as follows:
Step S21: from IV characteristic curve data, the open-circuit voltage V of the IV characteristic curve data of photovoltaic module is foundocWith
Short circuit current Isc;
Step S22: in [0 Voc] voltage range in uniformly obtain the voltage of N number of resampling point, adjacent resampling point
Interval voltage isAnd record the resampling voltage vector [V1, V2..., Vc..., VN], wherein VcIndicate c-th of voltage weight
The voltage of sampled point.
Step S23: it is directed to each resampling point Vc, c value range [1N], find original IV characteristic curve data in resampling
The voltage and current value of the left and right adjacent two data point of point, i.e. Vc-1And Vc+1And corresponding current value Ic-1And Ic+1;
Step S24: linear interpolation method is utilized, each electrical voltage point V is calculatedcCorresponding electric current Ic, specific to calculate
Method are as follows:To obtain N number of voltage resampling point (Vc,Ic)
Step S25: in [0Isc] electric current section in uniformly obtain the electric current of M resampling point, adjacent resampling point interval
Electric current isAnd record the resampling current vector [I1, I2..., Id..., IM], wherein IdIndicate d-th of electric current resampling
The electric current of point.
Step S26: it is directed to each electric current resampling point Id, d value range [1M], find original IV characteristic curve data in weigh
The current value I of the left and right adjacent two data point of sampled pointd-1And Id+1And corresponding voltage value Vd-1And Vd+1;
Step S27: linear interpolation method is utilized, each electric current resampling point I is calculateddCorresponding voltage value Vd, specific to count
Calculation method are as follows:To obtain M electric current resampling point (Vd,Id)。
Step S28: the M electric current resampling point that the S24 N number of voltage resampling point obtained and S27 are obtained, according to voltage
Ascending sequence merges, and obtains the resampling IV curve comprising M+N data point, models for subsequent photovoltaic module.
The present embodiment, the grid in the photovoltaic module fast modeling method based on extreme learning machine and IV characteristic of proposition are adopted
Quadrat method, detailed process are as follows:
Step S41: to the IV characteristic curve Data Data of component, the extraction range and sampling interval that grid is sampled are set.
It is [- 10,70] that its temperature, which extracts range, and it is [80,1000] that irradiation level, which extracts range, and the sampling interval of temperature is 1 DEG C, irradiation level
Extraction range is 50W/m2
Step S42: grid is successively selected, and is counted to sample is distributed in it;
Step S43: if sample size is less than maximum sampling number in the grid, all samples in the grid are randomly selected
70% be used as training sample, 30% be used as test sample, wherein 90% in training sample be used for hands-on, 10% conduct
Verifying collection;
Step S44: if quantity is greater than maximum number of samples in grid, first from the maximum sampling of selection in sample in the grid
A several samples, then randomly select all samples in the grid 70% is used as training sample, wherein in training sample
90% be used for hands-on, 10% as verifying collection;
Step S45: repeat the above steps S42 to S44, is defeated by obtained sample process until taking all grids
Enter for (G, T, V), export the data set for I, wherein G and T is irradiation level and temperature, V and I be voltage and corresponding electric current to
Amount;
Step S46: the test set passed back through, verifying collection and training set.
Such as Fig. 3, for single hidden layer in the photovoltaic module fast modeling method based on extreme learning machine and IV characteristic of proposition
The structure chart of feedforward neural network, it is characterised in that network structure is simple, there is three layers, is input layer, hidden layer and output respectively
Layer.Different from other complicated neural network structures, it contains only a hidden layer, and between input layer and hidden layer, it is hidden
Containing being all to connect entirely, and the connection between every two layers can all assign weight between layer and output layer, hidden neuron is also set up
There are activation primitive and threshold value to achieve the purpose that the adjustment to input data.The working principle of Single hidden layer feedforward neural networks is such as
Under: all parameters are initialized first, then input layer receives input data, and input data reaches output layer by hidden layer,
Centre can be endowed two layers of weight and activation primitive and threshold value.Output layer exports the calculated value of network, by output valve and very
The comparison of real value, output layer feedforward predicts error, finally, network adjusts weight and threshold according to the error that output layer is fed back
Value, i.e. update network parameter.By successive ignition, network can reach higher precision of prediction.
It is according to claim 1 that photovoltaic module is modeled by the method based on extreme learning machine such as Fig. 4,
And IV characteristic is predicted.Its modeling process is specific as follows:
Step S61: the training data point set for taking step S5 to obtain, with the irradiation level of each data point, component backboard temperature,
The voltage of component is the input of training data, using its corresponding electric current as the output of training data.
Step S62: Single hidden layer feedforward neural networks, the input layer of random initializtion neural network and the company of hidden layer are established
The threshold value b for connecing weight w and hidden layer neuron, the connection weight between hidden layer and output layer is β, indicates w and β with matrix
It is as follows:
Threshold value b=[the b of hidden layer neuron1, b2, b3..., bl]T, the number of hidden layer neuron takes in the present embodiment
150.
Step S63: in the present embodiment, activation primitive takes g (x)=sigmoid (x), then input matrix X by neural network it
First output afterwards isIt is found that when neural network inputs j-th of sample, network output
Predicted value is
For the hidden layer output matrix of neural network, the predicted value of network output is T=[t1, t2..., tQ]。
Step S64: operating limit learning machine fitting algorithm calculates the hidden layer and output layer of Single hidden layer feedforward neural networks
Weight β calculates to obtain β=H by least square method+T ' thus obtains network model, and wherein parameter is w, b, β.
Such as table 1, for proposition the photovoltaic module fast modeling method based on extreme learning machine and IV characteristic actual measurement light
Lie prostrate the application result on component CdTe75669.From the comparison of statistical method as it can be seen that and BP neural network, general regression neural
Network (GRNN), the conventional machines learning algorithm such as support vector machines (SVM) are compared, and what the present embodiment was proposed is learnt based on the limit
The photovoltaic module fast modeling method of machine, is greatly shortened on the training speed of network model, and in training set, verifying
Collect, the precision and robustness of model are obviously improved on test set.In addition, still there is good performance on test set, say
Bright this method has good Generalization Capability.
Table 1: each algorithm performance comparison
Such as Fig. 5, for the photovoltaic module fast modeling method and generalized regression network GRNN based on extreme learning machine of proposition
The characteristic comparison of the IV modeled under various operating conditions, it is upper from figure as it can be seen that proposed by the invention learnt based on the limit
The IV characteristic curve of the photovoltaic module fast modeling method prediction of machine is more close to the IV characteristic curve of actual measurement.
The above are preferred embodiments of the present invention, all any changes made according to the technical solution of the present invention, and generated function is made
When with range without departing from technical solution of the present invention, all belong to the scope of protection of the present invention.
Claims (7)
1. a kind of photovoltaic module fast modeling method based on extreme learning machine and IV characteristic, which is characterized in that including following step
It is rapid:
Step S1: IV characteristic curve tester is used, IV characteristic curve data, irradiation level and the component backboard of photovoltaic module are acquired
Temperature;
Step S2: to the IV characteristic curve data of photovoltaic module, the resampling of data point interpolation is carried out, obtains IV characteristic curve;
Step S3: based on single diode five-parameter model, being carried out curve fitting using optimization algorithm and obtain fitting root-mean-square error,
Judge that abnormal curve, the IV curve data of rejecting abnormalities obtain normal IV curve data collection according to fitting root-mean-square error;
Step S4: according to the characteristic irradiation level of IV and component backboard temperature, grid is carried out to normal IV curve data collection
It extracts, acquires the IV curve data collection of the various operating conditions of uniform fold, and be randomly divided into independent training curve collection and test song
Line collection;
Step S5: by training curve collection and test curve concentrate the data point voltage, electric current and irradiation level of every curve, temperature into
Row combination, obtains the training data point set being made of the voltage of each data point, electric current, irradiation level and temperature and test data point
Collection.
Step S6: according to training data point set, it is by input, component electric current of irradiation level, component backboard temperature and component voltage
Output establishes photovoltaic module model using extreme learning machine fitting algorithm training Single hidden layer feedforward neural networks;
Step S7: test data point set is input to the extreme learning machine photovoltaic module model that training obtains in S6, obtains model
The current value of calculating, and be compared with measured current value, to test and assess the error of extreme learning machine photovoltaic module model.
2. the photovoltaic module fast modeling method according to claim 1 based on extreme learning machine and IV characteristic, feature
It is: the step S2 specifically:
Step S21: from IV characteristic curve data, the open-circuit voltage V of the IV characteristic curve data of photovoltaic module is foundocAnd short circuit
Electric current Isc;
Step S22: in [0 Voc] voltage range in uniformly obtain the voltage of N number of resampling point, the interval of adjacent resampling point
Voltage isAnd record the resampling voltage vector [V1, V2..., Vc..., VN], wherein VcIndicate c-th of voltage resampling
The voltage of point;
Step S23: it is directed to each resampling point Vc, c value range [1 N] finds resampling point in original IV characteristic curve data
The voltage and current value of left and right adjacent two data point, i.e. Vc-1And Vc+1And corresponding current value Ic-1And Ic+1;
Step S24: linear interpolation method is utilized, each electrical voltage point V is calculatedcCorresponding electric current Ic, specific calculation method
Are as follows:To obtain N number of voltage resampling point (Vc, Ic);
Step S25: in [0 Isc] electric current section in uniformly obtain the electric current of M resampling point, adjacent resampling point interval electricity
Stream isAnd record the resampling current vector [I1, I2..., Id... IM], wherein IdIndicate d-th of electric current resampling point
Electric current;
Step S26: it is directed to each electric current resampling point Id, d value range [1 M], find original IV characteristic curve data in resampling
The current value I of the left and right adjacent two data point of pointd-1And Id+1And corresponding voltage value Vd-1And Vd+1;
Step S27: linear interpolation method is utilized, each electric current resampling point I is calculateddCorresponding voltage value Vd, specific calculating side
Method are as follows:
To obtain M electric current resampling point (Vd, Id);
Step S28: the M electric current resampling point that the S24 N number of voltage resampling point obtained and S27 are obtained, according to voltage by small
Merge to big sequence, obtains the resampling IV characteristic curve comprising M+N data point.
3. the photovoltaic module fast modeling method according to claim 1 based on extreme learning machine and IV characteristic, feature
Be: the curve-fitting method is specifically, using the photovoltaic mould based on a kind of ADAPTIVE MIXED simplex based on hawk strategy
Shape parameter extracting method extracts five parameter values of single diode photovoltage model of each curve under each operating condition, and calculates five
The root-mean-square error RMSE between curve and measured curve being fitted under Parameter Conditions, circular are as follows:
Wherein I is the current value of actual measurement,It is the current value of prediction, N is the number of whole I-V curve up-sampling point, and by RMSE
Greater than acceptable threshold value RMSETCurve regard as abnormal curve and rejected, obtain normal IV curve data.
4. the photovoltaic module fast modeling method according to claim 1 based on extreme learning machine and IV characteristic, feature
It is: the step S4 specifically:
Step S41: to IV characteristic curve Data Data, the extraction range and sampling interval that grid is sampled are set;
Step S42: grid is successively selected, and is counted to sample is distributed in it;
Step S43: if sample size is less than maximum sampling number in the grid, all samples in the grid are randomly selected
70% is used as training sample, and 30% is used as test sample, and wherein 90% in training sample is used for hands-on, and 10% conduct is tested
Card collection;
Step S44: if quantity is greater than maximum number of samples in grid, first from choosing maximum number of samples in the grid in sample
A sample, then randomly select all samples in the grid 70% is used as training sample, wherein in training sample
90% is used for hands-on, and 10% as verifying collection;
Step S45: repeat the above steps S42 to S44, is that input is by obtained sample process until taking all grids
(G, T, V) exports the data set for I, and wherein G and T is irradiation level and temperature, and V and I are voltage and corresponding current vector;
Step S46: the test set returned, verifying collection and training set.
5. the photovoltaic module fast modeling method according to claim 1 based on extreme learning machine and IV characteristic, feature
It is: the step S5 specifically:
Step S51: the IV curved line number for including in training set and test set obtained in statistic procedure S4 respectively;
Step S52: to each training set IV curve, its irradiation level, component backboard temperature, each electrical voltage point and current point are taken
As training data point set;
Step S53: to each test set IV curve, its irradiation level, component backboard temperature, each electrical voltage point and current point are taken
As test data point set.
6. the photovoltaic module fast modeling method according to claim 1 based on extreme learning machine and IV characteristic, feature
It is: the step S6 specifically:
Step S61: the training data point set for taking step S5 to obtain, with the irradiation level of each data point, component backboard temperature, component
Voltage be training data input, using its corresponding electric current as the output of training data;
Step S62: Single hidden layer feedforward neural networks model is established;
Step S63: the threshold of the input layer of random initializtion neural network and the connection weight w of hidden layer and hidden layer neuron
Value b, the connection weight between hidden layer and output layer is β, indicates that w and β are as follows with matrix:
Threshold value b=[the b of hidden layer neuron1, b2, b3..., bl]T
Step S64: assuming that hidden layer neuron activation function is g (x), then input matrix X is by the after neural network
One output isIt is found that when neural network inputs j-th of sample, the prediction of network output
Value is
For the hidden layer output matrix of neural network, the predicted value of network output is T=[t1, t2..., tQ];
Step S65: operating limit learning machine fitting algorithm calculates the hidden layer and output layer weight of Single hidden layer feedforward neural networks
β calculates to obtain β=H by least square method+T ' thus obtains network model, and wherein parameter is w, b and β.
7. the photovoltaic module fast modeling method according to claim 1 based on extreme learning machine and IV characteristic, feature
It is: the step S6 specifically: test data point set obtained in step S5 is taken, by irradiation level, component backboard temperature and electricity
Pressure value is input in step S6 the extreme learning machine photovoltaic module model that training obtains, and obtains the current value of model prediction, and with
The current value of actual measurement is compared with the error between the current value of actual measurement and the current value of model prediction and assesses
The accuracy of extreme learning machine photovoltaic module model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910278837.5A CN109992911B (en) | 2019-05-06 | 2019-05-06 | Photovoltaic module rapid modeling method based on extreme learning machine and IV characteristics |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910278837.5A CN109992911B (en) | 2019-05-06 | 2019-05-06 | Photovoltaic module rapid modeling method based on extreme learning machine and IV characteristics |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109992911A true CN109992911A (en) | 2019-07-09 |
CN109992911B CN109992911B (en) | 2023-04-07 |
Family
ID=67131021
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910278837.5A Active CN109992911B (en) | 2019-05-06 | 2019-05-06 | Photovoltaic module rapid modeling method based on extreme learning machine and IV characteristics |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109992911B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113037213A (en) * | 2021-03-04 | 2021-06-25 | 国网新疆电力有限公司信息通信公司 | Photovoltaic cell model parameter identification method and device based on goblet sea squirt group algorithm |
WO2022095265A1 (en) * | 2020-11-09 | 2022-05-12 | 贵州电网有限责任公司 | Photovoltaic cell parameter identification method based on improved equilibrium optimizer algorithm |
CN114660531A (en) * | 2022-05-24 | 2022-06-24 | 江西西平计量检测有限公司 | Detection method, system and device based on ammeter measurement error compensation |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160358099A1 (en) * | 2015-06-04 | 2016-12-08 | The Boeing Company | Advanced analytical infrastructure for machine learning |
CN109409024A (en) * | 2018-12-25 | 2019-03-01 | 福州大学 | Photovoltaic module voltage-current characteristic modeling method based on one-dimensional depth residual error network |
-
2019
- 2019-05-06 CN CN201910278837.5A patent/CN109992911B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160358099A1 (en) * | 2015-06-04 | 2016-12-08 | The Boeing Company | Advanced analytical infrastructure for machine learning |
CN109409024A (en) * | 2018-12-25 | 2019-03-01 | 福州大学 | Photovoltaic module voltage-current characteristic modeling method based on one-dimensional depth residual error network |
Non-Patent Citations (1)
Title |
---|
CHEN, ZC (CHEN, ZHICONG) 等: "Intelligent fault diagnosis of photovoltaic arrays based on optimized kernel extreme learning machine and I-V characteristics.", 《APPLIED ENERGY》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2022095265A1 (en) * | 2020-11-09 | 2022-05-12 | 贵州电网有限责任公司 | Photovoltaic cell parameter identification method based on improved equilibrium optimizer algorithm |
CN113037213A (en) * | 2021-03-04 | 2021-06-25 | 国网新疆电力有限公司信息通信公司 | Photovoltaic cell model parameter identification method and device based on goblet sea squirt group algorithm |
CN114660531A (en) * | 2022-05-24 | 2022-06-24 | 江西西平计量检测有限公司 | Detection method, system and device based on ammeter measurement error compensation |
CN114660531B (en) * | 2022-05-24 | 2022-08-26 | 江西西平计量检测有限公司 | Detection method, system and device based on ammeter measurement error compensation |
Also Published As
Publication number | Publication date |
---|---|
CN109992911B (en) | 2023-04-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109873610B (en) | Photovoltaic array fault diagnosis method based on IV characteristic and depth residual error network | |
CN109842373B (en) | Photovoltaic array fault diagnosis method and device based on space-time distribution characteristics | |
CN106777984B (en) | A method of photovoltaic array Working state analysis and fault diagnosis are realized based on density clustering algorithm | |
CN109660206B (en) | Wasserstein GAN-based photovoltaic array fault diagnosis method | |
CN106059496B (en) | A kind of photovoltaic module array performance monitoring and the method and system of Fault Identification | |
CN109409024B (en) | Photovoltaic module voltage and current characteristic modeling method based on one-dimensional depth residual error network | |
CN109766952B (en) | Photovoltaic array fault detection method based on partial least square method and extreme learning machine | |
CN108983749B (en) | Photovoltaic array fault diagnosis method based on K-SVD training sparse dictionary | |
CN109992911A (en) | Photovoltaic module fast modeling method based on extreme learning machine and IV characteristic | |
CN106021806A (en) | Photovoltaic string fault diagnosis method based on kernel function limit learning machine | |
CN111245365B (en) | Photovoltaic module fault diagnosis method based on dynamic weighted depth forest | |
CN109670553A (en) | Diagnosing failure of photovoltaic array method based on Adaptive Neuro-fuzzy Inference | |
CN110503153B (en) | Photovoltaic system fault diagnosis method based on differential evolution algorithm and support vector machine | |
WO2024051524A1 (en) | Joint prediction method and apparatus for hydraulic, wind and photovoltaic generation power | |
Djalab et al. | Robust method for diagnosis and detection of faults in photovoltaic systems using artificial neural networks | |
CN113595132B (en) | Photovoltaic online parameter identification method based on maximum power point and hybrid optimization algorithm | |
CN113131515A (en) | Intelligent microgrid AI fault detection and network optimization reconstruction system and method | |
CN114091329A (en) | Regional photovoltaic power generation amount calculation method based on inverse distance interpolation and extreme learning machine | |
CN111177973B (en) | Photovoltaic array online modeling method based on reinforcement learning | |
Gong et al. | Fault diagnosis method of photovoltaic array based on random forest algorithm | |
Liu et al. | Fault diagnosis of photovoltaic array based on XGBoost method | |
Jiurong et al. | A data-driven fault diagnosis method for photovoltaic modules | |
CN114819382B (en) | LSTM-based photovoltaic power prediction method | |
CN113221465B (en) | Photovoltaic power generation operation and maintenance oriented component model parameter identification method | |
Seuret-Jiménez et al. | Non-coherent detection of dust in photovoltaic systems in series configuration using Lipschitz exponent |
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 |