CN113887705B - Photovoltaic panel running state monitoring method based on sparse RBF neural network - Google Patents
Photovoltaic panel running state monitoring method based on sparse RBF neural network Download PDFInfo
- Publication number
- CN113887705B CN113887705B CN202111156324.0A CN202111156324A CN113887705B CN 113887705 B CN113887705 B CN 113887705B CN 202111156324 A CN202111156324 A CN 202111156324A CN 113887705 B CN113887705 B CN 113887705B
- Authority
- CN
- China
- Prior art keywords
- vector
- photovoltaic panel
- neural network
- sparse
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 47
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000012544 monitoring process Methods 0.000 title claims abstract description 23
- 238000005070 sampling Methods 0.000 claims abstract description 36
- 230000002159 abnormal effect Effects 0.000 claims abstract description 8
- 239000013598 vector Substances 0.000 claims description 116
- 239000011159 matrix material Substances 0.000 claims description 36
- 210000002569 neuron Anatomy 0.000 claims description 30
- 239000010410 layer Substances 0.000 claims description 27
- 238000005286 illumination Methods 0.000 claims description 13
- 238000010606 normalization Methods 0.000 claims description 8
- 230000005856 abnormality Effects 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 3
- 239000011229 interlayer Substances 0.000 claims description 3
- 238000003064 k means clustering Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 101100001675 Emericella variicolor andJ gene Proteins 0.000 claims 1
- 238000004364 calculation method Methods 0.000 claims 1
- 238000000605 extraction Methods 0.000 abstract description 4
- 238000005259 measurement Methods 0.000 abstract description 3
- 238000010248 power generation Methods 0.000 description 5
- 238000011161 development Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000003062 neural network model Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 101100272279 Beauveria bassiana Beas gene Proteins 0.000 description 1
- 101100001674 Emericella variicolor andI gene Proteins 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a photovoltaic panel running state monitoring method based on a sparse RBF neural network, which aims to effectively mine nonlinear relation characteristics among measurement data of a photovoltaic panel and monitor the running state of the photovoltaic panel on the basis. Specifically, the method of the invention designs a sparse RBF neural network structure, takes real-time sampling data of the photovoltaic panel as input and output at the same time, and reflects whether the running state of the photovoltaic panel is abnormal or not through the generated error. The method of the invention has the advantages that: firstly, the method utilizes the nonlinear fitting capability of the RBF neural network, and realizes the nonlinear characteristic extraction of sample data under the normal running state of the photovoltaic panel through the built sparse RBF neural network; secondly, the method generates errors through the sparse RBF neural network to judge whether the photovoltaic panel is abnormal in real time, which is different from the traditional method for monitoring and extracting nonlinear characteristic components.
Description
Technical Field
The invention relates to a photovoltaic panel running state monitoring method, in particular to a photovoltaic panel running state monitoring method based on a sparse RBF neural network.
Background
The application range of solar energy is very wide, and the solar energy is currently applied to photovoltaic and photochemical technologies and the like. Compared with the traditional non-renewable energy sources, the solar energy has the natural advantages of zero pollution, convenient installation, small regional limitation, safety, reliability and permanent use. The main solar energy utilization modes include solar energy heat utilization, solar energy thermal power generation, solar energy photovoltaic power generation and the like. The photovoltaic power generation technology is mature, has the overall conditions of rapid development and realization of energy radical transformation, and becomes a main choice for the development of new energy in the future. With the continuous progress of the photovoltaic power generation technology and the simplification of the manufacturing process, the development prospect and development potential of the photovoltaic industry are wider.
Photovoltaic devices are generally installed outdoors or in a relatively harsh environment, and long-time illumination, rain erosion and dust accumulation can cause various faults of the photovoltaic devices, so that the power generation efficiency is reduced or the photovoltaic devices are in an abnormal working state. In order to solve the problem, the photovoltaic equipment needs to be monitored in real time, and whether the photovoltaic equipment is abnormal or not is judged according to the monitored data so as to timely solve the problem. In the current background of the widespread use of artificial intelligence and data mining, monitoring the operational status of photovoltaic panels by sampling data in real time has received increasing attention from engineering technicians. In general, sample data of a photovoltaic panel in a normal working state is abundant, and sample data of the photovoltaic panel in different fault states is deficient. Therefore, the data-driven photovoltaic panel operation state monitoring also needs to find suitable solutions and technologies from the single-classification feature extraction point of view.
Furthermore, it is considered that the photovoltaic panel is intermittently operated and is directly affected by the sun illumination. In addition, the change of illumination intensity is not manually and accurately predicted and controlled, and the problems add great difficulty to the operation state monitoring of the photovoltaic panel. From the aspect of single-classification feature extraction, the extracted features should be the change features which can better represent the sampled data set in the normal running state. Taking a photovoltaic panel as an example, the extracted characteristics should reflect a complex nonlinear relationship between measurement data such as temperature, illumination intensity, current and voltage. An artificial neural network model represented by a radial basis function (Radial Basis Function, abbreviated as RBF) neural network can better process nonlinear classification or regression problems, but the characteristic problem of solving single classification by applying the RBF neural network is rarely solved.
Disclosure of Invention
The main technical problems to be solved by the invention are as follows: and (3) effectively excavating the nonlinear relation characteristics among the measurement data of the photovoltaic panel, and monitoring the operation state of the photovoltaic panel on the basis. Specifically, the method of the invention designs a sparse RBF neural network structure, takes real-time sampling data of the photovoltaic panel as input and output at the same time, and reflects whether the running state of the photovoltaic panel is abnormal or not through the generated error.
The technical scheme adopted by the method for solving the problems is as follows: a photovoltaic panel running state monitoring method based on a sparse RBF neural network comprises the following steps:
Step (1): after the data which can be measured in real time by the photovoltaic panel are determined, sample data vectors of all sampling moments are collected and stored according to fixed sampling time intervals in a normal working state of the photovoltaic panel; wherein, 8 data in the sample data vector of each sampling moment are in turn: illumination intensity, panel temperature, maximum dynamic dc power, dc current, dc voltage, ac power, ac voltage, and ac current.
Step (2): after the sample data vectors X 1,x2,…,xN of N sampling moments with the illumination intensity larger than zero form a training data matrix X= [ X 1,x2,…,xN ], normalization processing is carried out on each row vector in X epsilon R 8×N, thereby obtaining a new matrixWherein x i∈R8×1 represents a sample data vector at the i-th sampling time, R 8×N represents a real matrix of 8×n dimensions, R represents a real set, R 8×1 represents a real vector of 8×1 dimensions, i e {1,2, …, N }, and the normalization processing is performed in a manner specifically shown in steps (2.1) to (2.2).
Step (2.1): let z j∈R1×N denote the row vector of the j-th row in matrix X; where j ε {1,2, …,8}.
Step (2.2): after finding the minimum value M j and the maximum value M j in the row vector z j, according toCalculating to obtain new matrixLine vector of j-th line in (b)
Step (3): determining a center point vector of the sparse RBF neural network inter-layer neurons according to steps (3.1) through (3.5) as followsAnd a width parameter delta 1,δ2,…,δ8.
Step (3.1): the total number of clusters is determined to be C, and j=1 is initialized again.
Step (3.2): new matrix is to be formedAfter 7 row vectors except for the jth row vector form a matrix X j∈R7×N, N column vectors in the matrix X j are clustered into C clusters by using a k-means clustering (k-means clustering) algorithm, and the central point vector of each cluster is recorded as the jth group central point vector
Step (3.3): according to the formulaCalculate the a-th center point vectorAnd the b-th center point vectorSquare distance betweenWhere a e {1,2, …, C }, b e {1,2, …, C }, the superscript T represents the transpose of the matrix or vector.
Step (3.4): the width parameter delta j is calculated according to equation ① as follows:
Step (3.5): judging whether j is less than 8; if yes, setting j=j+1 and returning to the step (3.2); if not, obtaining the central point vector and the width parameter of the middle layer neuron of the sparse RBF neural network.
Step (4): respectively sequentiallyColumn vectors in (a)As an input vector xi epsilon R 8×1, calculating to obtain an output vector y 1,y2,…,yN of the middle layer neuron of the sparse RBF neural network; wherein the ith column vector is calculatedThe specific implementation process of the output vector y i of the corresponding sparse RBF neural network middle layer neuron is shown in the steps (4.1) to (4.4).
Step (4.1): setting input vectorsAfter that, j=1 is reinitialized.
Step (4.2): after forming 7 data except the jth data in the input vector ζ into a column vector ζ j, calculating a jth sub-output vector θ j∈R1×C by using a formula ② as follows:
In the above equation, e represents a natural constant, i.e., e≡2.71828, c∈ {1,2, …, C }, θ j (C) represents the C-th data in θ j.
Step (4.3): judging whether j is less than 8; if yes, setting j=j+1 and returning to the step (4.2); if not, the method comprises the steps of; then 8 sub-output vectors theta 1,θ2,…,θ8 are obtained.
Step (4.4): θ 1,θ2,…,θ8 is combined into an output vector y i＝[θ1,θ2,…,θ8]T.
Step (5): optimizing by using an error Back Propagation (BP) algorithm to obtain a weight vector w 1,w2,…,w8 of a neuron of an output layer of the sparse RBF neural network and a threshold d 1,d2,…,d 8; the activation function of the output layer neurons of the sparse RBF neural network is f (u) =1/(1+e -u), and u is an activation function argument.
As can be seen from the step (4) and the steps (4.1) to (4.3) attached to the step, the number of neurons in the middle layer of the sparse RBF neural network built by the method is equal to 8 XC. According to the step (5), the number of the neurons of the output layer of the sparse RBF neural network built by the method is equal to 8, namely the number of the photovoltaic panel which is determined in the step (1) and can be used for measuring data in real time. The number of neurons of the input layer of the built sparse RBF neural network is equal to 8, so the method of the invention is actually thatAnd also serves as the input and output of the sparse RBF neural network.
It can be seen that the objective function J that the error back propagation algorithm needs to minimize in step (5) is as follows:
In the above, F i∈R8×1 represents that the sparse RBF neural network output layer neurons correspond to The j-th element F i (j) in F i is calculated as follows:
In the above equation, w j and d j represent the jth weight vector and the jth threshold, respectively. After initializing w j and d j, the error back propagation algorithm is based on the idea of gradient descent, continuously updating the weight vector w j and the threshold d j until the target J is very close to 0.
Step (6): calculating the output vector F 1,F2,…,FN of the output layer neuron of the sparse RBF neural network, and then constructing an output matrix G= [ F 1,F2,…,FN ] and calculating an error matrix
Step (7): after calculating the mean vector μ of all column vectors in the error matrix E, the covariance matrix Λ= (E-U) T/(N-1) is calculated; wherein all column vectors in the mean matrix U.epsilon.R 8×N are equal to μ.
Step (8): after calculating the monitoring index vector Q E R N×1 according to the formula q=diag { (E-U) TΛ-1 (E-U) }, the average value of the maximum 10 data in Q is recorded as Q lim; wherein diag { } represents an operation of converting matrix diagonal elements in curly brackets into column vectors.
Step (9): collecting data of the latest sampling moment of the photovoltaic panel and forming a sample data vector x t∈R8×1; wherein, the reference symbol t indicates the latest sampling time, and 8 data in x t are sequentially arranged according to the sequence listed in step (1).
Step (10): judging whether the first data (namely, illumination intensity) in the column vector x t is larger than 0; if yes, executing the step (11); if not, the photovoltaic panel is in a standby state, and the step (9) is returned.
Step (11): according to the formulaNormalization is carried out on each data in x t to obtain a new data vectorWherein x t (j) andRespectively x t andIs the j-th data in (a).
Step (12): the output vector F t∈R8×1 of the sparse RBF neural network output layer neurons is calculated according to steps (12.1) through (12.3) as follows.
Step (12.1): setting input vectorsAfter that, j=1 is reinitialized.
Step (12.2): steps (4.2) to (4.3) are performed to obtain 8 sub-output vectors θ 1,θ2,…,θ8, which are then combined into an output vector y t＝[θ1,θ2,…,θ8]T.
Step (12.3): according to the formulaThe j-th data F t (j) in F t is calculated, resulting in an output vector F t∈R8×1.
Step (13): according to the formulaAfter the monitoring index Q t is calculated, judging whether Q t is smaller than Q lim or not; if yes, the photovoltaic panel normally operates at the current sampling moment, and the step (9) is returned to continue to monitor the operation state of the photovoltaic panel by using the sample data vector at the latest sampling moment; if not, then step (14) is performed to determine whether to trigger an photovoltaic panel abnormality warning.
Step (14): returning to the step (9), and continuously using the sample data vector at the latest sampling moment to implement photovoltaic panel fault detection; if the monitoring indexes of the continuous 6 sampling moments are all larger than Q lim, triggering the photovoltaic panel abnormality warning; otherwise, the photovoltaic panel normally operates, and the step (9) is returned to continue to monitor the operation state of the photovoltaic panel by using the sample data at the latest sampling time.
By carrying out the steps described above, the advantages of the method according to the invention are described below.
Firstly, the method of the invention utilizes the nonlinear fitting capability of the RBF neural network, and realizes the nonlinear characteristic extraction of sample data under the normal running state of the photovoltaic panel through the built sparse RBF neural network. Secondly, the method generates errors through the sparse RBF neural network for judging whether the photovoltaic panel is abnormal in real time, and the errors are different from the traditional method for monitoring the extracted nonlinear characteristic components, so that the errors can reflect abnormal changes of complex nonlinear relations among measured data such as temperature, illumination intensity, current and voltage.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention.
Fig. 2 is a schematic structural diagram of a sparse RBF neural network constructed by the method of the invention.
Detailed Description
The invention will be described in detail below with reference to the drawings and the detailed description.
The invention discloses a photovoltaic panel running state monitoring method based on a sparse RBF neural network, and a specific implementation mode of the method is described below by combining an implementation flow diagram shown in FIG. 1.
Step (1): after the data which can be measured in real time by the photovoltaic panel are determined, sample data vectors of all sampling moments are collected and stored according to fixed sampling intervals in a normal working state of the photovoltaic panel; wherein, 8 data that every sampling moment photovoltaic electricity board can measure are in proper order: illumination intensity, panel temperature, maximum dynamic dc power, dc current, dc voltage, ac power, ac voltage, and ac current.
Step (2): after N sample data vectors X 1,x2,…,xN with the illumination intensity larger than zero form a training data matrix X= [ X 1,x2,…,xN ], normalization processing is carried out on each row vector in X epsilon R 8×N, so as to obtain a new matrix
Step (3): determining a center point vector of the sparse RBF neural network inter-layer neurons according to the steps (3.1) to (3.5)And a width parameter delta 1,δ2,…,δ8.
Step (4): respectively sequentiallyColumn vectors in (a)As the input vector ζ e R 8×1, the output vector y 1,y2,…,yN of the middle layer neurons of the sparse RBF neural network is calculated sequentially according to the steps (4.1) to (4.4).
Step (5): and optimizing by using an error back propagation algorithm to obtain a weight vector w 1,w2,…,w8 of the neuron of the output layer of the sparse RBF neural network and a threshold d 1,d2,…,d8.
And (3) in the step (4) and the step (5), a neural network model with the number of neurons of an input layer equal to 8, the number of neurons of an intermediate layer equal to 8 multiplied by C and the number of neurons of an output layer equal to 8 is actually built. In addition, referring to the above steps (4.1) to (4.4), it can be found that the network model is not a fully connected neural network model, and a part of the input layer neurons and the middle layer neurons are not connected, so that a sparse connection structure as shown in fig. 2 is presented. Therefore, the method of the invention names the brand new neural network model as: sparse RBF neural networks.
Step (6): calculating the output vector F 1,F2,…,FN of the output layer neuron of the sparse RBF neural network, and then constructing an output matrix G= [ F 1,F2,…,FN ] and calculating an error matrix
Step (7): after calculating the mean vector μ of all column vectors in the error matrix E, the covariance matrix Λ= (E-U) T/(N-1) is calculated.
Step (8): after the monitor index vector Q E R N×1 is calculated according to the formula q=diag { (E-U) TΛ-1 (E-U) }, the average value of the maximum 10 data in Q is recorded as Q lim.
Step (9): and (3) collecting sample data of the latest sampling moment of the photovoltaic panel, wherein the sample data specifically comprises 8 data listed in the step (1), and the 8 data are combined into a sample data vector x t∈R8×1.
Step (10): judging whether the first data (namely, illumination intensity) in the column vector x t is larger than 0; if yes, executing the step (11); if not, the photovoltaic panel is in a standby state, and the step (9) is returned.
Step (11): according to the formulaNormalization is carried out on each data in x t to obtain a new data vector
Step (12): will beAs an input vector of the sparse RBF neural network, an output vector F t∈R8×1 of the output layer neurons is calculated.
Step (13): according to the formulaAfter the monitoring index Q t is calculated, judging whether Q t is smaller than Q lim or not; if yes, the photovoltaic panel normally operates at the current sampling moment, and the step (9) is returned to continue to monitor the operation state of the photovoltaic panel by using the sample data at the latest sampling moment; if not, then step (14) is performed to determine whether to trigger an photovoltaic panel abnormality warning.
Step (14): returning to the step (9), and continuously using the sample data at the latest sampling moment to implement photovoltaic panel fault detection; if the monitoring indexes of the continuous 6 sampling moments are all larger than Q lim, triggering the photovoltaic panel abnormality warning; otherwise, the photovoltaic panel normally operates, and the step (9) is returned to continue to monitor the operation state of the photovoltaic panel by using the sample data at the latest sampling time.
Claims (2)
1. The photovoltaic panel running state monitoring method based on the sparse RBF neural network is characterized by comprising the following steps of:
Step (1): after the data which can be measured in real time by the photovoltaic panel are determined, sample data vectors of all sampling moments are collected and stored according to fixed sampling time intervals in a normal working state of the photovoltaic panel; wherein, 8 data in the sample data vector of each sampling moment are in turn: illumination intensity, panel temperature, maximum dynamic DC power, DC current, DC voltage, AC power, AC voltage and AC current;
Step (2): after the sample data vectors X 1,x2,…,xN of N sampling moments with the illumination intensity larger than zero form a training data matrix X= [ X 1,x2,…,xN ], normalization processing is carried out on each row vector in X epsilon R 8×N, thereby obtaining a new matrix Wherein x i∈R8×1 represents a sample data vector at the ith sampling time, R 8×N represents a real matrix of 8×n dimensions, R represents a real set, R 8×1 represents a real vector of 8×1 dimensions, i e {1,2, …, N }, and the normalization processing is performed in a manner specifically shown in steps (2.1) to (2.2);
Step (2.1): let z j∈R1×N denote the row vector of the j-th row in matrix X; wherein j ε {1,2, …,8};
Step (2.2): after finding the minimum value M j and the maximum value M j in the row vector z j, according to Calculating to obtain new matrixLine vector of j-th line in (b)
Step (3): determining a center point vector of the sparse RBF neural network inter-layer neurons according to steps (3.1) through (3.5) as followsAnd a width parameter δ 1,δ2,…,δ8;
Step (3.1): determining the total number of cluster clusters as C, and initializing j=1;
Step (3.2): new matrix is to be formed After 7 rows of vectors except for the jth row of vectors form a matrix X j∈R7×N, N columns of vectors in the matrix X j are clustered into C clusters by using a k-means clustering algorithm, and the central point vector of each cluster is recorded as
Step (3.3): according to the formulaCalculate the a-th center point vectorAnd the b-th center point vectorSquare distance betweenWherein a e {1,2, …, C }, b e {1,2, …, C }, the upper label T representing the transpose of the matrix or vector;
Step (3.4): the width parameter delta j is calculated according to equation ① as follows:
step (3.5): judging whether j is less than 8; if yes, setting j=j+1 and returning to the step (3.2); if not, obtaining a central point vector and a width parameter of the middle layer neuron of the sparse RBF neural network;
Step (4): respectively sequentially Column vectors in (a)As an input vector xi epsilon R 8×1, calculating to obtain an output vector y 1,y2,…,yN of the middle layer neuron of the sparse RBF neural network; wherein the ith column vector is calculatedThe specific implementation process of the output vector y i of the corresponding sparse RBF neural network middle layer neuron is shown in the steps (4.1) to (4.4);
step (4.1): setting input vectors After that, j=1 is initialized again;
Step (4.2): after forming 7 data except the jth data in the input vector ζ into a column vector ζ j, calculating a jth sub-output vector θ j∈R1×C by using a formula ② as follows:
In the above formula, e represents a natural constant, c∈ {1,2, …, C }, θ j (C) represents the C-th data in θ j;
Step (4.3): judging whether j is less than 8; if yes, setting j=j+1 and returning to the step (4.2); if not, the method comprises the steps of; then 8 sub-output vectors θ 1,θ2,…,θ8 are obtained;
step (4.4): combining θ 1,θ2,…,θ8 into an output vector y i＝[θ1,θ2,…,θ8]T;
Step (5): optimizing by using an error back propagation algorithm to obtain a weight vector w 1,w2,…,w8 of a neuron of an output layer of the sparse RBF neural network and a threshold d 1,d2,…,d8;
Step (6): calculating the output vector F 1,F2,…,FN of the output layer neuron of the sparse RBF neural network, and then constructing an output matrix G= [ F 1,F2,…,FN ] and calculating an error matrix The calculation method of the ith output vector F i∈R8×1 is as follows:
In the above formula, F i (j) represents the j-th data in F i, w j and d j represent the j-th weight vector and the j-th threshold, F (u) =1/(1+e -u) represents the activation function of the sparse RBF neural network output layer neuron, To activate function arguments;
Step (7): after calculating the mean vector μ of all column vectors in the error matrix E, the covariance matrix Λ= (E-U) T/(N-1) is calculated; wherein, the mean matrix U epsilon R 8×N is composed of N mean vectors mu;
Step (8): after calculating the monitoring index vector Q E R N×1 according to the formula q=diag { (E-U) TΛ-1 (E-U) }, the average value of the maximum 10 data in Q is recorded as Q lim; wherein diag { } represents an operation of converting matrix diagonal elements in curly brackets into column vectors;
Step (9): collecting data of the latest sampling moment of the photovoltaic panel and forming a sample data vector x t∈R8×1; wherein, the reference symbol t represents the latest sampling time, 8 data in x t are sequentially arranged according to the sequence listed in the step (1);
step (10): judging whether the first data in the sample data vector x t is greater than 0; if yes, executing the step (11); if not, the photovoltaic panel is in a standby state, and the step (9) is returned;
step (11): according to the formula Normalization is carried out on each data in x t to obtain a new data vectorWherein x t (j) andRespectively x t andJ e {1,2, …,8};
Step (12): will be As an input vector of the sparse RBF neural network, calculating to obtain an output vector F t∈R8×1 of the output layer neurons;
Step (13): according to the formula After the monitoring index Q t is calculated, judging whether Q t is smaller than Q lim or not; if yes, the photovoltaic panel normally operates at the current sampling moment, and the step (9) is returned to continue to monitor the operation state of the photovoltaic panel by using the sample data vector at the latest sampling moment; if not, executing the step (14) to determine whether to trigger an abnormal warning of the photovoltaic panel;
Step (14): returning to the step (9), and continuously using the sample data vector at the latest sampling moment to implement photovoltaic panel fault detection; if the monitoring indexes of the continuous 6 sampling moments are all larger than Q lim, triggering the photovoltaic panel abnormality warning; otherwise, the photovoltaic panel normally operates, and the step (9) is returned to continue to monitor the operation state of the photovoltaic panel by using the sample data at the latest sampling time.
2. The photovoltaic panel operation state monitoring method based on the sparse RBF neural network according to claim 1, wherein the specific implementation process of calculating F t in the step (12) is as follows:
step (12.1): setting input vectors After that, j=1 is initialized again;
Step (12.2): executing the steps (4.2) to (4.3), thereby obtaining 8 sub-output vectors theta 1,θ2,…,θ8, and combining the sub-output vectors theta 1,θ2,…,θ8 into an output vector y t＝[θ1,θ2,…,θ8]T;
Step (12.3): according to the formula The j-th data F t (j) in F t is calculated, resulting in an output vector F t∈R8×1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111156324.0A CN113887705B (en) | 2021-09-22 | 2021-09-22 | Photovoltaic panel running state monitoring method based on sparse RBF neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111156324.0A CN113887705B (en) | 2021-09-22 | 2021-09-22 | Photovoltaic panel running state monitoring method based on sparse RBF neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113887705A CN113887705A (en) | 2022-01-04 |
| CN113887705B true CN113887705B (en) | 2024-07-05 |
Family
ID=79004603
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111156324.0A Active CN113887705B (en) | 2021-09-22 | 2021-09-22 | Photovoltaic panel running state monitoring method based on sparse RBF neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113887705B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118040910B (en) * | 2024-04-12 | 2024-06-28 | 国网山东省电力公司菏泽供电公司 | Online monitoring method for energy storage abnormal state of micro-grid in remote area |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104484833A (en) * | 2014-12-02 | 2015-04-01 | 常州大学 | Photovoltaic power generation output power tracking algorithm based on genetics algorithm improved RBF-BP neural network |
| CN110675278A (en) * | 2019-09-18 | 2020-01-10 | 上海电机学院 | Photovoltaic power short-term prediction method based on RBF neural network |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR102065120B1 (en) * | 2018-09-27 | 2020-02-11 | 경북대학교 산학협력단 | Battery charging state estimation method based on neural network |
| CN112231982B (en) * | 2020-10-13 | 2024-02-02 | 广东光美能源科技有限公司 | Photovoltaic panel fault detection method based on distributed soft measurement model |
-
2021
- 2021-09-22 CN CN202111156324.0A patent/CN113887705B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104484833A (en) * | 2014-12-02 | 2015-04-01 | 常州大学 | Photovoltaic power generation output power tracking algorithm based on genetics algorithm improved RBF-BP neural network |
| CN110675278A (en) * | 2019-09-18 | 2020-01-10 | 上海电机学院 | Photovoltaic power short-term prediction method based on RBF neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113887705A (en) | 2022-01-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108847686B (en) | Photovoltaic inverter fault prediction method | |
| Liu et al. | Intelligent fault diagnosis of photovoltaic array based on variable predictive models and I–V curves | |
| KR102230548B1 (en) | Power generation prediction and efficiency diagnosis system of solar power generation facilities using FRBFNN model | |
| CN109002781B (en) | Fault prediction method for energy storage converter | |
| CN112883634B (en) | DC measurement system state prediction method and system based on multi-dimensional analysis | |
| CN113887705B (en) | Photovoltaic panel running state monitoring method based on sparse RBF neural network | |
| CN110544003A (en) | temperature prediction-based wind power plant wind turbine generator set frequency converter state evaluation method | |
| Xu et al. | Short-term wind speed prediction based on GRU | |
| Zare et al. | Fault diagnosis in photovoltaic modules using a straightforward voltage-current characteristics evaluation | |
| Lahiouel et al. | Photovoltaic fault diagnosis algorithm using fuzzy logic controller based on calculating distortion ratio of values | |
| Bendale et al. | Deep Learning for Solar Panel Maintenance: Detecting Faults and Improving Performance | |
| Sobahi et al. | Data-Driven Approach for Condition Monitoring and Improving Power Output of Photovoltaic Systems. | |
| Liu et al. | Fault diagnosis of photovoltaic array based on XGBoost method | |
| Tan et al. | Ultra-short-term photovoltaic generation forecasting model based on weather clustering and markov chain | |
| CN113033088A (en) | Wind driven generator system fault diagnosis method based on depth subdomain adaptive migration network | |
| CN112270081A (en) | Parallel Elman-NN-based wind driven generator fault detection method | |
| CN112231982A (en) | Photovoltaic panel fault detection method based on distributed soft measurement model | |
| Zhou et al. | Day-ahead power forecasting model for a photovoltaic plant in Macao based on weather classification using SVM/PCC/LM-ANN | |
| Xu | Research on fault diagnosis of electromechanical actuator based on convolution neural network | |
| CN113935408B (en) | Wind power generation equipment state monitoring method based on simplified core principal component network | |
| CN113946608B (en) | Wind generating set running state monitoring method based on online RBF neural network | |
| Zhou et al. | Power Grid transient stability prediction method based on improved CNN under big data background | |
| Jiurong et al. | A data-driven fault diagnosis method for photovoltaic modules | |
| Hu et al. | A fault diagnosis method for photovoltaic modules based on transfer long short-term memory neural network | |
| Xiong et al. | Research on Icing Status Forecasting of Wind Turbine Blades Based on Machine Learning |
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 |