CN116317943A - Photovoltaic array hot spot detection method - Google Patents
Photovoltaic array hot spot detection method Download PDFInfo
- Publication number
- CN116317943A CN116317943A CN202310227297.4A CN202310227297A CN116317943A CN 116317943 A CN116317943 A CN 116317943A CN 202310227297 A CN202310227297 A CN 202310227297A CN 116317943 A CN116317943 A CN 116317943A
- Authority
- CN
- China
- Prior art keywords
- data
- photovoltaic
- output
- hot spot
- photovoltaic array
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 29
- 238000000034 method Methods 0.000 claims abstract description 43
- 238000012549 training Methods 0.000 claims abstract description 26
- 238000012360 testing method Methods 0.000 claims abstract description 16
- 238000013528 artificial neural network Methods 0.000 claims abstract description 14
- 230000006870 function Effects 0.000 claims description 32
- 238000007726 management method Methods 0.000 claims description 21
- 238000004891 communication Methods 0.000 claims description 17
- 230000000694 effects Effects 0.000 claims description 13
- 238000012544 monitoring process Methods 0.000 claims description 12
- 210000002569 neuron Anatomy 0.000 claims description 12
- 230000009466 transformation Effects 0.000 claims description 9
- 230000005540 biological transmission Effects 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000005516 engineering process Methods 0.000 claims description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 4
- 238000011478 gradient descent method Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 238000007405 data analysis Methods 0.000 claims description 3
- 238000013480 data collection Methods 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 claims description 3
- 238000012423 maintenance Methods 0.000 claims description 3
- 230000006855 networking Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000010248 power generation Methods 0.000 claims description 3
- 230000009286 beneficial effect Effects 0.000 abstract description 3
- 230000007547 defect Effects 0.000 abstract description 2
- 238000003745 diagnosis Methods 0.000 abstract description 2
- 230000006872 improvement Effects 0.000 description 7
- 238000001931 thermography Methods 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 240000007651 Rubus glaucus Species 0.000 description 1
- 235000011034 Rubus glaucus Nutrition 0.000 description 1
- 235000009122 Rubus idaeus Nutrition 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000013021 overheating Methods 0.000 description 1
- 239000005022 packaging material Substances 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000003466 welding 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
- H02S50/10—Testing of PV devices, e.g. of PV modules or single PV cells
- H02S50/15—Testing of PV devices, e.g. of PV modules or single PV cells using optical means, e.g. using electroluminescence
-
- 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
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
- H02J13/00002—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network characterised by monitoring
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
- H02J13/00006—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network characterised by information or instructions transport means between the monitoring, controlling or managing units and monitored, controlled or operated power network element or electrical equipment
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
- H02J13/00006—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network characterised by information or instructions transport means between the monitoring, controlling or managing units and monitored, controlled or operated power network element or electrical equipment
- H02J13/00022—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network characterised by information or instructions transport means between the monitoring, controlling or managing units and monitored, controlled or operated power network element or electrical equipment using wireless data transmission
-
- 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)
- Power Engineering (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Photovoltaic Devices (AREA)
Abstract
The invention belongs to the technical field of fault diagnosis of photovoltaic systems, and particularly relates to a photovoltaic array hot spot detection method. Aiming at the defect that the existing photovoltaic system hot spot detection method only depends on a single parameter and has low accuracy, the invention adopts the following technical scheme: a photovoltaic array hot spot detection method comprises the following steps: s1, constructing a data set comprising a training set and a testing set by taking current, voltage, power and temperature as characteristics and taking the fault degree of a photovoltaic panel as a label; step S2, building a neural network classifier, optimizing by adopting an optimizer, and training by using a training set; s3, collecting data and uploading the data to a management system, and judging whether the photovoltaic array meets the early warning rule according to the preset early warning rule; and S4, pushing the early warning information to the related party by the management system. The beneficial effects of the invention are as follows: compared with the existing method for analyzing and judging only according to current or temperature, the hot spot fault degree of the photovoltaic panel can be judged more accurately.
Description
Technical Field
The invention belongs to the technical field of fault diagnosis of photovoltaic systems, and particularly relates to a photovoltaic array hot spot detection method.
Background
Under certain conditions, the shielded or defective area in the series branch of the photovoltaic module in the power generation state is taken as a load, and energy generated by other areas is consumed, so that local overheating is caused, and the phenomenon is called a hot spot effect of the photovoltaic module. The hot spot effect can reduce the output power of the component, and causes the photovoltaic component to be burnt locally to form dark spots, the welding spots are melted, the packaging materials are aged and the like to be permanently damaged, so that the output power of the photovoltaic component is directly influenced, and the service life of the photovoltaic component is directly prolonged.
The existing photovoltaic system hot spot detection mainly comprises an infrared thermal imager detection method and a photovoltaic module I-V curve scanning method. According to the principle that the photovoltaic cells have obvious temperature differences under different working states, the infrared thermal imager detection method utilizes the infrared thermal imager to shoot thermal imaging photos to find out the hot spot faults, and fault detection is carried out in modes of manual holding, fixed-point setting, unmanned aerial vehicle carrying the infrared thermal imager and the like, so that the hot spot fault positions can be accurately positioned, but purchase and maintenance equipment cost is higher, the dependency on the performance of shooting equipment is strong, the shooting equipment is easily influenced by weather, certain manpower workload is realized, time and labor are wasted, and the efficiency is lower. The I-V curve scanning method requires a great deal of hardware and labor, and has low detection efficiency.
To solve the cost and efficiency problems, a method of detecting hot spots from current has emerged. For example: the Chinese patent application with publication number of CN110620551A discloses a hot spot detection method, which comprises the following steps: acquiring current values of all photovoltaic group strings in the photovoltaic array in real time; determining the average value of the current of each photovoltaic group string in a preset length time period according to the acquired current value of each photovoltaic group string; determining a current average value of each photovoltaic group string combination in the preset length time period according to the acquired current value of each photovoltaic group string; and determining whether each photovoltaic group string has a hot spot fault according to the current average value of each photovoltaic group string in the preset length time period and the current average value of the combination of the photovoltaic group strings in which each photovoltaic group string is positioned in the preset length time period. The Chinese patent application with publication number of CN108964606A discloses a hot spot fault detection method of a photovoltaic system, which comprises the following steps: s1: establishing a theoretical current model of each group of strings of each array of the photovoltaic system; s2: predicting theoretical current output values of each group of strings according to the theoretical current model;
s3: monitoring current output values of each group of strings in real time; s4: preprocessing the current output values of each group of strings, and screening out an array possibly with hot spots; s5: judging whether the screened array possibly having hot spots has hot spots or not to obtain a hot spot array; s6: further judging a hot spot group string in the hot spot array; s7: and determining the position of the fault component in the hot spot cluster.
The method for detecting the hot spots according to the current has the advantages of high feasibility, good economy and the like. However, the accuracy is to be improved by only relying on one parameter of the current. Meanwhile, the timeliness of data sharing is poor, and staff cannot timely obtain feedback.
In addition, a method for setting a temperature sampling device of the photovoltaic module to collect temperature data of each part of the surface of the photovoltaic module and detecting whether hot spots exist or not independently according to the temperature data or comparison with a thermal imaging graph is also provided. However, when the temperature data is solely used, the accuracy is low, and when the temperature data is solely used, the thermal imaging cost is high due to the comparison between the temperature data and the thermal imaging diagram.
Disclosure of Invention
Aiming at the defect that the existing photovoltaic system hot spot detection method only depends on a single parameter and is low in accuracy, the invention provides the photovoltaic array hot spot detection method which comprehensively analyzes multi-channel data of current, voltage, power and temperature and improves judgment accuracy.
In order to achieve the above purpose, the invention adopts the following technical scheme: a photovoltaic array hot spot detection method, the photovoltaic array hot spot detection method comprising:
s1, constructing a data set comprising a training set and a testing set by taking current, voltage, power and temperature as input characteristics and the failure degree of a photovoltaic panel as an output label;
step S2, building a neural network classifier, optimizing by adopting an optimizer, training by using a training set, and continuously adjusting based on a test result of a test set to obtain an optimal model;
step S3, collecting current, voltage, power and temperature data of the photovoltaic array during operation and uploading the current, voltage, power and temperature data to a management system, wherein the management system adopts a model obtained in the previous step to infer, and judges whether the photovoltaic array meets the early warning rule according to an inference result and a preset early warning rule;
and S4, when the early warning rule is met, early warning information is generated, and the management system pushes the early warning information to the relevant party.
According to the photovoltaic array hot spot detection method, four channel data including current, voltage, power, temperature and the like are comprehensively analyzed, and a multi-mode model is built based on a machine learning algorithm, so that the hot spot fault degree of a photovoltaic panel is accurately judged. The directly collected data are current, voltage and temperature, and the power is calculated by the current and the voltage.
In the step S2, a neural network classifier is built based on PyTorch, and Adam is adopted as an optimizer for optimization.
As an improvement, the neural network classifier includes an input layer, a hidden layer and an output layer, 3-layer neurons, wherein the input layer contains 4 neurons for inputting data of 4 channels of voltage, current, power and temperature, respectively, the hidden layer contains 16 neurons for extracting features, and the output layer contains 4 neurons for outputting probabilities of 4 categories of normal, mild, moderate and severe.
As an improvement, the activation function of the hidden layer adopts a Relu function, and the class probability output is carried out by connecting with a Softmax function after the output layer.
As an improvement, during model training, all data are subjected to Z-score transformation to eliminate the influence of dimension; when the model is inferred, the data to be predicted is firstly subjected to Z-score transformation by using the mean value mu and the standard deviation sigma of the training sample, and then is input into the model for analysis, and the prediction type is output;
the formula for the Z-score transformation is:
as an improvement, in step S3, the early warning rule includes the voltage, current, power, temperature of the photovoltaic module and the warning threshold value of the hot spot fault degree output by the model.
In step S3, the largest one of the probabilities of the 4 types of output is taken as the criterion.
As an improvement, in step S4, early warning information is transmitted to the management system and each user terminal through a 5G communication technology, so as to realize remote transmission of on-site dynamic real-time monitoring data and remote interactive setting of working parameters; in step S4, the early warning information is sent to remind workers of maintenance in a mode of sending short messages through the software APP.
As an improvement, the method further comprises step S5, wherein the data is transmitted to the cloud platform by adopting a 5G intelligent gateway technology architecture, and the data is displayed in a chart form for a worker to check the data and the state of the photovoltaic panel.
As an improvement, the deployment mode of the 5G intelligent gateway is as follows: the intelligent sensor deployment position is on site, the access mode/network is Zigbee, and the intelligent sensor deployment position is used for acquiring data; the deployment position of the 5G intelligent gateway is on site, and the access mode/network is 5G, and is used for 5G networking, protocol transmission, data collection and intelligent control; the communication gateway deployment position is an external network machine room, and the access mode/network is the Internet and is used for authentication, protocol forwarding and security gateway; the communication service deployment position is an office machine room, and the access mode/network is a photovoltaic power station office network and is used for application communication; the application service deployment position is an office machine room, and the access mode/network is a photovoltaic power station office network and is used for monitoring data analysis and remote control.
The photovoltaic array hot spot detection method has the beneficial effects that: the four channel data of current, voltage, power, temperature and the like are comprehensively analyzed, a multi-mode model is established based on a machine learning algorithm, and compared with the existing method for analyzing and judging only according to the current or the temperature, the hot spot fault degree of the photovoltaic panel can be accurately judged. Furthermore, a neural network classifier is built based on PyTorch, adam is adopted as an optimizer for optimization, a Relu function is adopted as an activation function of a hidden layer, and a Softmax function is connected to an output layer for class probability output, so that a more accurate judgment result can be obtained. The current, voltage and temperature data required during detection are directly acquired by a sensor, so that the calculated amount of the model is small, and the calculation is rapid; only in the model creation stage, infrared image data is needed, and in the detection process after the model creation is completed, the infrared image is not needed to be shot any more, so that the detection is faster and more convenient.
Drawings
Fig. 1 is a flowchart of a method for detecting a hot spot of a photovoltaic array according to a first embodiment of the present invention.
Fig. 2 is a diagram of a neural network model structure of a photovoltaic array hot spot detection method according to the first embodiment of the present invention.
Detailed Description
The technical solutions of the inventive embodiments of the present invention will be explained and illustrated below with reference to the drawings of the inventive embodiments of the present invention, but the following embodiments are only preferred embodiments of the inventive embodiments of the present invention, not all. Based on the examples in the implementation manner, other examples obtained by a person skilled in the art without making any inventive effort fall within the scope of protection created by the present invention.
A photovoltaic array hot spot detection method, the photovoltaic array hot spot detection method comprising:
s1, constructing a data set comprising a training set and a testing set by taking current, voltage, power and temperature as input characteristics and the failure degree of a photovoltaic panel as an output label;
step S2, building a neural network classifier, optimizing by adopting an optimizer, training by using a training set, and continuously adjusting based on a test result of a test set to obtain an optimal model;
step S3, collecting current, voltage, power and temperature data of the photovoltaic array during operation and uploading the current, voltage, power and temperature data to a management system, wherein the management system adopts a model obtained in the previous step to infer, and judges whether the photovoltaic array meets the early warning rule according to an inference result and a preset early warning rule;
and S4, when the early warning rule is met, early warning information is generated, and the management system pushes the early warning information to the relevant party.
According to the method for detecting the hot spots of the photovoltaic array, disclosed by the invention, the four channel data of current, voltage, power, temperature and the like are comprehensively analyzed, a multi-mode model is established based on a machine learning algorithm, and compared with the existing method for carrying out analysis and judgment only according to the current, the hot spot fault degree of the photovoltaic panel can be more accurately judged.
Example 1
Referring to fig. 1 and fig. 2, a method for detecting a photovoltaic array hot spot according to a first embodiment of the present invention includes:
s1, constructing a data set comprising a training set and a testing set by taking current, voltage, power and temperature as input characteristics and the failure degree of a photovoltaic panel as an output label;
step S2, building a neural network classifier, optimizing by adopting an optimizer, training by using a training set, and continuously adjusting based on a test result of a test set to obtain an optimal model;
step S3, collecting current, voltage, power and temperature data of the photovoltaic array during operation and uploading the current, voltage, power and temperature data to a management system, wherein the management system adopts a model obtained in the previous step to infer, and judges whether the photovoltaic array meets the early warning rule according to an inference result and a preset early warning rule;
and S4, when the early warning rule is met, early warning information is generated, and the management system pushes the early warning information to the relevant party.
In this embodiment, in step S1, a data set including 1000 samples is constructed with the failure degree (0 indicates normal, 1 indicates mild, 2 indicates moderate, and 3 indicates severe) of the photovoltaic panel as a label, wherein 80% of the samples form a training set, and the remaining 20% of the samples form a test set. Specifically, voltage, current and temperature sensors are installed on a photovoltaic power generation panel array and data are collected, meanwhile, hot spot effect data are collected on site by using an unmanned plane, and a professional judges the hot spot effect level according to a judging standard (0 is used for indicating normal, 1 is used for indicating mild, 2 is used for indicating medium, and 3 is used for indicating serious). The data of the sensors on each photovoltaic panel and the evaluated hot spot effect level are combined together as one piece of data, wherein the voltage, current and temperature data are used as input values, and the hot spot effect level is used as a label (expected value). In the above method, 1000 pieces of data are collected repeatedly in total, and a data set is constructed.
In this embodiment, in step S2, a neural network classifier is built based on the pyresch, adam is used as an optimizer to perform optimization, a Softmax function is used to perform normalization, and the hot spot fault degree of the photovoltaic module is determined according to the maximum class probability output by the model (for example, the probability distribution of output is normal 10%, mild 80%, moderate 5% and severe 5%, and the mild is the predicted hot spot fault degree of the photovoltaic module). And (3) training by using the training set in the step (1), continuously adjusting the super parameters based on the test result of the test set, and finally judging the hot spot fault degree of the photovoltaic module by using the optimal model.
In this embodiment, in step S2, the structure of the neural network classifier used is as follows: comprising 3 layers of neurons in total of an input layer, a hidden layer and an output layer. Wherein the input layer contains 4 neurons for inputting data of 4 channels of voltage, current, power and temperature, respectively. The hidden layer contains 16 neurons for feature extraction. The output layer contains 4 neurons for outputting probabilities of the 4 normal, mild, moderate, and severe classes. The activation function of the hidden layer adopts a Relu function, and the output layer is connected with a Softmax function to output class probability. During model training, all data are subjected to Z-score transformation to eliminate the influence of dimension. During model reasoning, the data to be predicted is subjected to Z-score transformation by using the mean value mu and the standard deviation sigma of the training samples, and then is input into model analysis, and the prediction type is output.
The Relu function formula is expressed as:
the formula for the Z-score transformation is:
in this embodiment, in step S2, the model training method (process) is as follows: let the input data be vector X, output be vector Y, the weight matrix between the input layer and the hidden layer be W, the weight matrix between the hidden layer and the output layer be V.
When the input layer feeds forward to the hidden layer, the input X and W are calculated as follows to obtain the output value L of the hidden layer 1 ：
L 1 ＝relu(W·X+B w )
Wherein relu is a laserLiving function, B W Is the bias value of the hidden layer.
Then L is 1 Then the output value L of the output layer is obtained by the following operation with V 2 。
L 2 ＝relu(V·L 1 +B v )
Wherein B is v Is the bias value of the output layer.
Output value L 2 And (5) carrying out Softmax function normalization treatment to obtain a final result Y.
Y＝softmax(L 2 )
The obtained result Y and the expected value T calculate the cross entropy as a loss function, and a gradient descent method is adopted to optimize the loss function, and the loss function loss is calculated on the hidden layer output L first 1 And the weight matrix V is adjusted according to the product of the gradient value and the learning rate as the amplitude to obtain a new value V ', and then the weight matrix W is sequentially adjusted according to the error back propagation algorithm according to the product of the gradient value and the learning rate as the amplitude to obtain a new value W'.
The cycle is then repeated according to the above steps until the loss function is minimized or the number of iterations reaches 2000 stops. At this time, the model training is completed.
In this embodiment, in step S3, an early warning rule is set. The early warning rule can set the voltage, current, power and temperature of the photovoltaic module and the warning threshold value of the hot spot fault degree output by the model, and the management system decides whether to generate early warning information according to the preset value and the actual value.
In this embodiment, in step S3, the model reasoning method (process) is as follows: during model reasoning, unknown data X new Is input as an input value into the input layer of the model, and forward operation is performed according to the steps instead of X until a calculated value Y is obtained new . Y is then new I.e. the inferred value for the unknown data. Let Y be new ＝[0.05 0.80 0.10 0.05]And judging the hot spot effect grade of the photovoltaic panel as a second grade (light) according to the principle of maximum probability. And the weight is not adjusted by loss function calculation and gradient descent method in the reasoning process.
In the embodiment, in step 4, early warning information is transmitted to the management system and each user terminal through a 5G communication technology, so as to realize remote transmission of on-site dynamic real-time monitoring data and remote interactive setting of working parameters, and the early warning information can also remind workers to maintain in a mode of sending short messages through the software APP.
In the embodiment, in step 4, the 5G-DNN-based photovoltaic panel monitoring may access the 5G network through a data network name (Data Network Name, DNN), and the monitoring data flow accesses the communication network in a 5G-DNN manner. And configuring a gateway platform containing a private SIM card, taking a gateway general packet radio service (General Packet Radio Service, GPRS) support node network as a network element of a core network, and determining gateway configuration and a specific interface of the gateway GPRS support node according to communication and connection requirements of photovoltaic panel data monitoring. And the data passes through the 5G base station and realizes the functions through the operator core network.
In this embodiment, the method further includes step 5: the 5G intelligent gateway technical architecture is adopted to transmit data to the cloud platform, and the data is displayed in a chart form, so that workers can view the data and the state of the photovoltaic panel at each moment every day. The 5G intelligent gateway integrates AI early warning, edge calculation, data summarization and data compression functions, a data center does not need to be uploaded through a 5G network, an instruction is returned, and the processing process is completed in a local edge calculation layer. The data is sent to the cloud platform via 5G encrypted transmission and displayed on a personal computer side (Personal Computer, PC) management platform and a handheld side application platform.
The deployment mode of the 5G intelligent gateway is as follows:
(1) the intelligent sensor deployment position is on site, the access mode/network is Zigbee, and the intelligent sensor deployment position is used for acquiring data;
(2) the deployment position of the 5G intelligent gateway is on site, and the access mode/network is 5G, which is used for 5G networking, protocol transmission, data collection, intelligent control and the like;
(3) the communication gateway deployment position is an external network machine room, and the access mode/network is the Internet and is used for authentication, protocol forwarding and security gateway;
(4) the communication service deployment position is an office machine room, the access mode/network is a photovoltaic power station office network, and the communication service deployment position is used for application communication and the like;
(5) the application service deployment position is an office machine room, and the access mode/network is a photovoltaic power station office network and is used for monitoring data analysis, remote control and the like.
The following description will be given of a specific application example.
At a certain photovoltaic power station, the photovoltaic panel array contains 6 strings of 20 modules (panels) each, totaling 120 photovoltaic panels. Each panel is provided with independent voltage, current and temperature sensors, and the detection ranges are respectively 0-110V, 0-10A and minus 30-500 ℃. The sensor communicates with the edge computing device via a serial port. The edge computing device takes raspberry group 4B as a main board, and the computing resources are configured into 8GB memory, 1.5GHz4 core CPU and 16GB flash memory. The edge computing device side software is developed by Java. An algorithm model constructed by a PyTorch deep learning framework is operated on the edge computing device, the model is a 3-layer error back propagation neural network, and the hot spot fault degree of the panel is calculated according to voltage, current, power (calculated by the voltage and the current) and temperature data sent by the sensor in real time.
For example, a certain photovoltaic panel assembly sensor measures 36V of output voltage, 0.5A of current and 105 ℃ of surface temperature at a certain time, and the probability of category 1-4 (normal-serious) is calculated by the input neural network to be 0.05, 0.10, 0.50 and 0.35 respectively, so that the maximum probability category is the moderate hot spot effect. Other data are exemplified in table 1.
Table 1 photovoltaic panel sensor data and hot spot effect identification example data
The photovoltaic array hot spot detection method of the first embodiment of the invention has the beneficial effects that: the hot spot fault degree of the photovoltaic panel is estimated based on current, voltage, power and temperature detection, and four channel data of the current, the voltage, the power, the temperature and the like are comprehensively analyzed, so that compared with the existing method for analyzing and judging only according to the current, the hot spot fault degree of the photovoltaic panel can be more accurately judged; constructing a multi-mode hot spot recognition model based on deep learning, analyzing multi-channel state data acquired by a sensor on a photovoltaic panel array by using the model, and evaluating the hot spot fault degree of the photovoltaic panel; constructing an early warning rule, pushing early warning information and fault degree to a management system and each user terminal in a grading manner, dynamically monitoring data in real time, and setting working parameters in a remote interaction manner; deploying the method at the edge end by means of a 5G network and a cloud platform, and realizing deployment, application and intelligent early warning of the method; the 5G transmission module is used for realizing real-time sharing of data, and a worker can timely receive related early warning information and find out the problem according to the related data; the system has the functions of on-site dynamic real-time monitoring data, remote interaction setting of working parameters, early warning, classified pushing, historical data cloud platform recording and the like.
While the invention has been described in terms of specific embodiments, it will be apparent to those skilled in the art that the invention is not limited to the specific embodiments described. Any modifications which do not depart from the functional and structural principles of the present invention are intended to be included within the scope of the appended claims.
Claims (10)
1. A photovoltaic array hot spot detection method is characterized in that: the method for detecting the hot spots of the photovoltaic array comprises the following steps:
s1, constructing a data set comprising a training set and a testing set by taking current, voltage, power and temperature as input characteristics and the failure degree of a photovoltaic panel as an output label;
step S2, building a neural network classifier, optimizing by adopting an optimizer, training by using a training set, and continuously adjusting based on a test result of a test set to obtain an optimal model;
step S3, collecting current, voltage, power and temperature data of the photovoltaic array during operation and uploading the current, voltage, power and temperature data to a management system, wherein the management system adopts a model obtained in the previous step to infer, and judges whether the photovoltaic array meets the early warning rule according to an inference result and a preset early warning rule;
and S4, when the early warning rule is met, early warning information is generated, and the management system pushes the early warning information to the relevant party.
2. The method for detecting the hot spots of the photovoltaic array according to claim 1, wherein the method comprises the following steps: in step S1, voltage, current and temperature sensors are installed on a photovoltaic power generation panel array and data are collected, meanwhile, an unmanned plane is used for collecting hot spot effect data on site, a professional judges hot spot effect grades according to judging standards, the data of the sensors on each photovoltaic panel and the judged hot spot effect grades are combined together to be used as one piece of data, the voltage, the current and the temperature data are used as input values, and the hot spot effect grades are used as labels.
3. The method for detecting the hot spots of the photovoltaic array according to claim 1, wherein the method comprises the following steps: in step S2, a neural network classifier is built based on pyrerch, adam is adopted as an optimizer for optimization, the neural network classifier comprises an input layer, a hidden layer and an output layer, and 3 layers of neurons, wherein the input layer comprises 4 neurons and is used for inputting data of 4 channels of voltage, current, power and temperature respectively, the hidden layer comprises 16 neurons and is used for extracting features, and the output layer comprises 4 neurons and is used for outputting probabilities of 4 categories of normal, mild, medium and serious.
4. A method for detecting hot spots in a photovoltaic array according to claim 3, wherein: in step S2, the activation function of the hidden layer adopts a Relu function, and the output layer is connected with a Softmax function to output a class probability, where the Relu function is expressed as:
5. the method for detecting the hot spots of the photovoltaic array according to claim 4, wherein the method comprises the following steps: in step S2, during model training, all data are subjected to Z-score transformation so as to eliminate the influence of dimension; when the model is inferred, the data to be predicted is firstly subjected to Z-score transformation by using the mean value mu and the standard deviation sigma of the training sample, and then is input into the model for analysis, and the prediction type is output;
the formula for the Z-score transformation is:
6. the method for detecting the hot spots of the photovoltaic array according to claim 5, wherein the method comprises the following steps: in step S2, let input data be vector X, output data be vector Y, weight matrix between input layer and hidden layer be W, weight matrix between hidden layer and output layer be V, expressed as:
when the input layer feeds forward to the hidden layer, the input X and W are calculated as follows to obtain the output value L of the hidden layer 1 ：
L 1 ＝relu(W·X+B w )
Wherein relu is an activation function, B W A bias value for the hidden layer;
then L is 1 Then the output value L of the output layer is obtained by the following operation with V 2 ：
L 2 ＝relu(V·L 1 +B v )
Wherein B is v To output the offset value of the layer, the output value L 2 And (3) carrying out Softmax function normalization treatment to obtain a final result Y:
Y＝softmax(L 2 )
the obtained result Y and the expected value T calculate the cross entropy as a loss function, and a gradient descent method is adopted to optimize the loss function, and the loss function loss is calculated on the hidden layer output L first 1 The weight matrix V is adjusted according to the product of the gradient value and the learning rate as the amplitude to obtain a new value V ', and then the weight matrix W is sequentially adjusted according to the error back propagation algorithm according to the product of the gradient value and the learning rate as the amplitude to obtain a new value W':
and (5) repeating the steps according to the steps until the loss function is reduced to the minimum or the iteration times reach the preset times, and stopping.
7. The method for detecting the hot spots of the photovoltaic array according to claim 6, wherein the method comprises the following steps: in step S3, unknown data X is processed during model reasoning new Is input as an input value into the input layer of the model, and forward operation is performed according to the steps instead of X until a calculated value Y is obtained new Y is then new The method is to reason the unknown data, and the weight is not adjusted by loss function calculation and gradient descent method in the reasoning process.
8. A method for detecting hot spots in a photovoltaic array according to claim 3, wherein: in step S3, the early warning rule comprises warning thresholds of the voltage, current, power, temperature and hot spot fault degree output by the model of the photovoltaic module; in step S3, the largest one of the probabilities of the 4 types of output is taken as the judgment basis.
9. The method for detecting the hot spots of the photovoltaic array according to claim 1, wherein the method comprises the following steps: in step S4, early warning information is transmitted to a management system and each user terminal through a 5G communication technology, so that remote transmission of on-site dynamic real-time monitoring data and remote interactive setting of working parameters are realized; in step S4, the early warning information is sent to remind workers of maintenance in a mode of sending short messages through the software APP.
10. The method for detecting the hot spots of the photovoltaic array according to claim 1, wherein the method comprises the following steps: step S5, transmitting the data to a cloud platform by adopting a 5G intelligent gateway technology architecture, and displaying the data in a chart form for a worker to check the data and the state of the photovoltaic panel;
the deployment mode of the 5G intelligent gateway is as follows: the intelligent sensor deployment position is on site, the access mode/network is Zigbee, and the intelligent sensor deployment position is used for acquiring data; the deployment position of the 5G intelligent gateway is on site, and the access mode/network is 5G, and is used for 5G networking, protocol transmission, data collection and intelligent control; the communication gateway deployment position is an external network machine room, and the access mode/network is the Internet and is used for authentication, protocol forwarding and security gateway; the communication service deployment position is an office machine room, and the access mode/network is a photovoltaic power station office network and is used for application communication; the application service deployment position is an office machine room, and the access mode/network is a photovoltaic power station office network and is used for monitoring data analysis and remote control.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310227297.4A CN116317943A (en) | 2023-03-10 | 2023-03-10 | Photovoltaic array hot spot detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310227297.4A CN116317943A (en) | 2023-03-10 | 2023-03-10 | Photovoltaic array hot spot detection method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116317943A true CN116317943A (en) | 2023-06-23 |
Family
ID=86818082
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310227297.4A Pending CN116317943A (en) | 2023-03-10 | 2023-03-10 | Photovoltaic array hot spot detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116317943A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116660317A (en) * | 2023-07-25 | 2023-08-29 | 北京智芯微电子科技有限公司 | Hot spot detection method, system, processor and storage medium of photovoltaic array |
| CN117237590A (en) * | 2023-11-10 | 2023-12-15 | 华能新能源股份有限公司山西分公司 | Photovoltaic module hot spot identification method and system based on image identification |
-
2023
- 2023-03-10 CN CN202310227297.4A patent/CN116317943A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116660317A (en) * | 2023-07-25 | 2023-08-29 | 北京智芯微电子科技有限公司 | Hot spot detection method, system, processor and storage medium of photovoltaic array |
| CN116660317B (en) * | 2023-07-25 | 2023-12-22 | 北京智芯微电子科技有限公司 | Hot spot detection method, system, processor and storage medium of photovoltaic array |
| CN117237590A (en) * | 2023-11-10 | 2023-12-15 | 华能新能源股份有限公司山西分公司 | Photovoltaic module hot spot identification method and system based on image identification |
| CN117237590B (en) * | 2023-11-10 | 2024-04-02 | 华能新能源股份有限公司山西分公司 | Photovoltaic module hot spot identification method and system based on image identification |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116317943A (en) | Photovoltaic array hot spot detection method | |
| CN108964606B (en) | Hot spot fault detection method for photovoltaic system | |
| CN103246265B (en) | Electromechanical equipment detection maintaining method | |
| CN109872003B (en) | Object state prediction method, object state prediction system, computer device, and storage medium | |
| CN109347668A (en) | A kind of training method and device of service quality assessment model | |
| CN107862052A (en) | A kind of fault case storehouse, fault tree and fault spectrum construction method | |
| CN109255440B (en) | Method for predictive maintenance of power production equipment based on Recurrent Neural Networks (RNN) | |
| KR101958474B1 (en) | System and method of balancing control for solar photovoltaic generation | |
| CN110335168B (en) | Method and system for optimizing power utilization information acquisition terminal fault prediction model based on GRU | |
| CN113408087B (en) | Substation inspection method based on cloud side system and video intelligent analysis | |
| CN105184886A (en) | Cloud data center intelligence inspection system and cloud data center intelligence inspection method | |
| CN115423009A (en) | Cloud edge coordination-oriented power equipment fault identification method and system | |
| CN113177652A (en) | Photovoltaic output prediction method | |
| CN116800199A (en) | Distributed photovoltaic operation electric energy quality monitoring and analyzing method | |
| CN116308304A (en) | New energy intelligent operation and maintenance method and system based on meta learning concept drift detection | |
| CN117411703A (en) | Modbus protocol-oriented industrial control network abnormal flow detection method | |
| CN115102287A (en) | Intelligent management and control system for new energy power station centralized area | |
| CN117574176B (en) | BIPV photovoltaic glass production process optimization method | |
| Lin et al. | On-line monitoring and fault diagnosis of PV array based on BP neural network optimized by genetic algorithm | |
| CN115905348A (en) | Industrial electricity utilization abnormity early warning method and system of multi-source heterogeneous data | |
| CN116578436A (en) | Real-time online detection method based on asynchronous multielement time sequence data | |
| CN115545962A (en) | Crop growth period control method and system based on multi-sensor system | |
| CN110874506A (en) | Low-temperature equipment fault prediction method | |
| CN113958463A (en) | Online monitoring method, system and device for fan blade | |
| CN112651454B (en) | Infrared data acquisition system and spiral data processing method for power equipment |
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 |