CN117318614A - Photovoltaic inverter fault prediction method - Google Patents
Photovoltaic inverter fault prediction method Download PDFInfo
- Publication number
- CN117318614A CN117318614A CN202311285987.1A CN202311285987A CN117318614A CN 117318614 A CN117318614 A CN 117318614A CN 202311285987 A CN202311285987 A CN 202311285987A CN 117318614 A CN117318614 A CN 117318614A
- Authority
- CN
- China
- Prior art keywords
- data
- fault
- inverter
- model
- photovoltaic inverter
- 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.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 61
- 238000012549 training Methods 0.000 claims abstract description 42
- 238000012545 processing Methods 0.000 claims abstract description 23
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 238000004140 cleaning Methods 0.000 claims abstract description 12
- 238000003745 diagnosis Methods 0.000 claims abstract description 11
- 238000002790 cross-validation Methods 0.000 claims abstract description 6
- 238000012423 maintenance Methods 0.000 claims description 29
- 239000011159 matrix material Substances 0.000 claims description 13
- 238000011282 treatment Methods 0.000 claims description 12
- 238000012795 verification Methods 0.000 claims description 12
- 238000013527 convolutional neural network Methods 0.000 claims description 11
- 238000012544 monitoring process Methods 0.000 claims description 11
- 238000004458 analytical method Methods 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 9
- 230000002159 abnormal effect Effects 0.000 claims description 8
- 238000001914 filtration Methods 0.000 claims description 8
- 230000009467 reduction Effects 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 7
- 238000011156 evaluation Methods 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 5
- 238000011418 maintenance treatment Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 238000009826 distribution Methods 0.000 claims description 4
- 230000003595 spectral effect Effects 0.000 claims description 4
- 239000013598 vector Substances 0.000 claims description 4
- 238000000546 chi-square test Methods 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 238000012847 principal component analysis method Methods 0.000 claims description 3
- 230000000717 retained effect Effects 0.000 claims description 3
- 238000001228 spectrum Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 238000000513 principal component analysis Methods 0.000 description 5
- 230000008439 repair process Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 230000004044 response Effects 0.000 description 3
- 238000009825 accumulation Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013523 data management Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02S—GENERATION OF ELECTRIC POWER BY CONVERSION OF INFRARED RADIATION, VISIBLE LIGHT OR ULTRAVIOLET LIGHT, e.g. USING PHOTOVOLTAIC [PV] MODULES
- H02S50/00—Monitoring or testing of PV systems, e.g. load balancing or fault identification
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- H02S40/00—Components or accessories in combination with PV modules, not provided for in groups H02S10/00 - H02S30/00
- H02S40/30—Electrical components
- H02S40/32—Electrical components comprising DC/AC inverter means associated with the PV module itself, e.g. AC modules
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Inverter Devices (AREA)
- Photovoltaic Devices (AREA)
Abstract
The invention discloses a photovoltaic inverter fault prediction method, which comprises the following steps: collecting operation data of each inverter in the inverter cluster, including voltage, current and temperature information, and cleaning and preprocessing; extracting representative features from the operation data of each inverter; establishing a historical fault database of the inverters in the inverter cluster, wherein the historical fault database comprises fault types, time and characteristic data information; model training using a fault database, evaluating performance of the model using cross-validation, and adjusting model parameters to improve accuracy; when new data enter the system, the new data are input into a fault diagnosis model to carry out fault prediction, whether the photovoltaic inverter cluster has faults or not is judged according to the output result of the model, and the type of the faults is determined; triggering an alarm or a processing measure according to the fault prediction result; the photovoltaic inverter fault prediction method can simultaneously perform unified fault prediction on a plurality of photovoltaic inverters.
Description
Technical Field
The invention relates to a photovoltaic inverter fault prediction method.
Background
The photovoltaic inverter (Photovoltaic Inverter) is a device that converts direct-current electric energy generated by a solar photovoltaic panel into alternating-current electric energy. The electrical energy generated by the photovoltaic panels is in the form of direct current, while most electrical grids and appliances use alternating current. Thus, photovoltaic inverters play a critical role in photovoltaic power generation systems.
At present, maintenance of the photovoltaic inverter usually adopts post-maintenance, and maintenance personnel have difficulty in grasping the health state of the photovoltaic inverter in real time. The fault prediction technology can help overhaulers to predict possible faults of the photovoltaic inverter in advance, however, most of the existing fault prediction methods at present depend on the whole life cycle operation data of equipment, the established fault prediction model is only suitable for single equipment, and a method for carrying out unified and accurate fault prediction on a plurality of photovoltaic inverters is lacking.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a photovoltaic inverter fault prediction method capable of simultaneously carrying out unified fault prediction on a plurality of photovoltaic inverters.
The technical scheme adopted for solving the technical problems is as follows:
a photovoltaic inverter fault prediction method, comprising the steps of:
data acquisition and processing: collecting operation data of each inverter in the inverter cluster, including voltage, current and temperature information, and cleaning and preprocessing;
feature extraction and selection: extracting representative features from the operation data of each inverter, including spectral features, time domain features and energy features;
establishing a fault database: establishing a historical fault database of the inverters in the inverter cluster, wherein the historical fault database comprises fault types, time and characteristic data information;
model training and optimizing: model training using a fault database, evaluating performance of the model using cross-validation, and adjusting model parameters to improve accuracy;
and (3) predicting and judging faults: when new data enter the system, the new data are input into a fault diagnosis model to carry out fault prediction, whether the photovoltaic inverter cluster has faults or not is judged according to the output result of the model, and the type of the faults is determined;
alarm or treatment measures: and triggering an alarm or processing measure according to the fault prediction result.
Preferably, the data acquisition and processing method comprises the following steps:
transmitting data to the internet of things device using the I2C, SPI or analog output interface to measure voltage, current and temperature;
cleaning and preprocessing the collected original data, including abnormal value removal, noise filtering, calibration and standardization;
selecting a relational database to store the acquired data, and setting a data acquisition system to transmit the data to a back-end server in real time;
the last 24 hours of data was retained in a once-per-minute fashion.
Preferably, the feature extraction and selection method comprises the following steps:
converting the operation data of the inverter into frequency domain signals by using Fourier transformation, extracting frequency distribution conditions of the frequency domain signals, including frequency energy, frequency peak values and the like, and analyzing and processing the characteristics to obtain representative frequency spectrum characteristics;
dividing the operation data of the inverter into different time periods by using a sliding window method, and then carrying out statistics and analysis on each data section, wherein the statistics and analysis comprise mean value, variance and waveform morphology so as to extract representative time domain characteristics;
converting the operation data of the inverter into a power signal, and calculating the characteristics of total energy, average power, power factor and the like of the inverter through an accumulated square root algorithm to extract representative energy characteristics;
and selecting and reducing the dimension of the extracted features to reduce the feature dimension.
Preferably, the feature is subjected to dimension reduction processing by adopting a principal component analysis method, and the method comprises the following steps of:
carrying out standardization processing on the feature data so that the mean value of each feature is 0 and the variance is 1;
according to the normalized characteristic data, calculating a covariance matrix among the characteristics;
performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors, wherein the eigenvalues represent variances of the covariance matrix in the direction of the eigenvectors, and the eigenvectors represent the directions of principal components of the covariance matrix;
sorting the characteristic values according to the sizes, and selecting characteristic vectors corresponding to the first k larger characteristic values as main components;
projecting the original characteristic data onto the selected main component to obtain dimension-reduced data, wherein the projection is realized through dot multiplication operation;
and determining a final dimension reduction dimension k according to the reserved total variance ratio.
Preferably, the method for establishing the fault database comprises the following steps:
monitoring the running state of each inverter in the inverter cluster, and recording the fault type, time and related characteristic data when faults occur;
cleaning and preprocessing the collected fault data, including abnormal value removal, noise filtering, calibration and standardization, so as to ensure the accuracy and consistency of the data;
the MySQL system is selected to store the collected fault data, a database table is created to record fault type, time and characteristic data information, and a data structure and index are set.
Preferably, the method for model training and optimization is as follows:
acquiring fault type, time and characteristic data information from a fault database as a training data set;
according to the fault type and the feature data, using chi-square test to evaluate the correlation between the feature and the fault type, and selecting the feature with higher correlation for training the model;
dividing a training data set into a training set and a verification set, dividing the data set into a plurality of subsets by adopting a cross verification method, wherein one part of the data set is used as the verification set each time, and the other part of the data set is used as the training set so as to evaluate the performance of the model;
the convolutional neural network is selected for training, and in the training process, parameters of the model are adjusted according to the characteristic data of the training set and the corresponding fault type, so that the model can more accurately conduct fault prediction;
and evaluating the trained model by using the verification set, calculating indexes such as accuracy, recall rate, precision and the like of the model, and optimizing the model according to an evaluation result.
Preferably, the convolutional neural network architecture includes:
input layer: the state quantity of the photovoltaic inverter comprises voltage, current and power;
output layer: the output result is mapped to the [0,1] interval using a sigmoid activation function, indicating the probability of failure.
Preferably, the method for predicting and judging the fault comprises the following steps: and judging the fault rate of each photovoltaic inverter in the photovoltaic inverter cluster according to the output result of the model, and determining the fault type of the photovoltaic inverter with the fault rate higher than 50%.
Preferably, the method for alarming or processing measures is as follows: and notifying operation and maintenance personnel to carry out maintenance treatment in a mode of short message, mail or APP.
The beneficial effects of the invention are as follows:
by continuously monitoring the operation data of the inverter cluster, the faults of the inverter can be detected in time, corresponding treatment measures are adopted in advance, and the fault expansion or larger loss is avoided; according to the method, the characteristic extraction and selection technology is utilized to extract representative characteristics from the inverter operation data, so that the fault prediction model is more efficient and accurate. Meanwhile, the accuracy and performance of prediction are further improved through model training and optimization.
Based on a large number of historical fault databases, a fault diagnosis model is constructed by utilizing a machine learning technology, so that a prediction result is more credible. According to the prediction result, corresponding measures can be timely taken to carry out maintenance, replacement and other operations, and the operation efficiency of the photovoltaic system is improved; the fault type of the photovoltaic inverter can be judged through the fault prediction model, and corresponding treatment measures can be adopted in a targeted manner for different types of faults, so that the maintenance efficiency and the success rate are improved; the method can realize automatic monitoring and fault prediction of the inverter cluster, reduce the workload of manual inspection and maintenance, and improve the operation and maintenance efficiency and reliability of the photovoltaic system.
Detailed Description
The principles and features of the present invention are described below with examples given for the purpose of illustration only and are not intended to limit the scope of the invention. The invention is more specifically described by way of example in the following paragraphs. Advantages and features of the invention will become more apparent from the following description and from the claims.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and/or" as used herein includes any and all combinations of one or more of the associated listed items.
Examples
A photovoltaic inverter fault prediction method, comprising the steps of:
data acquisition and processing: collecting operation data of each inverter in the inverter cluster, including voltage, current and temperature information, and cleaning and preprocessing;
feature extraction and selection: extracting representative features from the operation data of each inverter, including spectral features, time domain features and energy features;
establishing a fault database: establishing a historical fault database of the inverters in the inverter cluster, wherein the historical fault database comprises fault types, time and characteristic data information;
model training and optimizing: model training using a fault database, evaluating performance of the model using cross-validation, and adjusting model parameters to improve accuracy;
and (3) predicting and judging faults: when new data enter the system, the new data are input into a fault diagnosis model to carry out fault prediction, whether the photovoltaic inverter cluster has faults or not is judged according to the output result of the model, and the type of the faults is determined;
alarm or treatment measures: and triggering an alarm or processing measure according to the fault prediction result.
By installing sensors in the photovoltaic inverter clusters, operation data of each inverter including information such as voltage, current and temperature are collected in real time. Cleaning and preprocessing the acquired data to remove abnormal values and noise; representative features are extracted from the preprocessed data. Spectral features may be extracted using spectral analysis methods, temporal features may be extracted using statistical methods, energy features may be extracted using energy calculation methods, etc. The most discriminative feature is selected for fault prediction by feature selection techniques.
And establishing a fault database according to the known photovoltaic inverter fault data. Each fault sample should contain a fault type tag and corresponding characteristic data. The fault database should be large enough and representative to improve the accuracy and generalization ability of the model; model training is performed using the established fault database. Common machine learning algorithms include Support Vector Machines (SVMs), decision trees, random forests, neural networks, and the like. And evaluating the performance of the model by using a cross-validation method, and optimizing the model parameters according to the evaluation result so as to improve the accuracy and stability of fault prediction.
When new inverter data enters the system, the new inverter data is input into a trained fault diagnosis model to conduct fault prediction. The model outputs a prediction result, judges whether the photovoltaic inverter cluster has faults or not, and determines the type of the faults. A threshold may be set to determine the outcome of the prediction, e.g., exceeding a certain threshold indicates that a fault exists; and triggering corresponding alarming or processing measures according to the fault prediction result. For example, if the prediction indicates a fault, the system may send an alarm message to the service personnel and provide detailed fault information and suggested treatment. The operation and maintenance personnel can repair or replace related equipment in time.
The method for collecting and processing the data comprises the following steps:
transmitting data to the internet of things device using the I2C, SPI or analog output interface to measure voltage, current and temperature;
cleaning and preprocessing the collected original data, including abnormal value removal, noise filtering, calibration and standardization;
selecting a relational database to store the acquired data, and setting a data acquisition system to transmit the data to a back-end server in real time;
the last 24 hours of data was retained in a once-per-minute fashion.
The real-time monitoring of the photovoltaic inverter can be realized by collecting data once every minute and transmitting the data to the back-end server. Thus, potential faults can be found and predicted in time, operation and maintenance personnel are helped to take corresponding treatment measures, and downtime and loss are reduced; by using the I2C, SPI or analog output interface, data such as voltage, current, temperature and the like can be directly obtained from the photovoltaic inverter, and manual input and human errors are avoided. In addition, the data preprocessing comprises the steps of outlier removal, noise filtering, calibration, standardization and the like, so that the accuracy and the reliability of the data can be improved.
The relational database is selected to store the collected data, so that the data can be conveniently queried, analyzed and managed. Meanwhile, the data acquisition system is arranged to transmit data to the back-end server in real time, so that timeliness and integrity of the data can be guaranteed; according to the scheme description, the data of the last 24 hours are reserved, unlimited accumulation of the data can be avoided, and the storage space and the cost are saved. Meanwhile, the frequency of data transmission can be reduced by collecting the data once per minute, and the pressure on the communication network and the server resources is reduced.
The method for extracting and selecting the characteristics comprises the following steps:
converting the operation data of the inverter into frequency domain signals by using Fourier transformation, extracting frequency distribution conditions of the frequency domain signals, including frequency energy, frequency peak values and the like, and analyzing and processing the characteristics to obtain representative frequency spectrum characteristics;
dividing the operation data of the inverter into different time periods by using a sliding window method, and then carrying out statistics and analysis on each data section, wherein the statistics and analysis comprise mean value, variance and waveform morphology so as to extract representative time domain characteristics;
converting the operation data of the inverter into a power signal, and calculating the characteristics of total energy, average power, power factor and the like of the inverter through an accumulated square root algorithm to extract representative energy characteristics;
and selecting and reducing the dimension of the extracted features to reduce the feature dimension.
By converting the operational data of the inverter into a frequency domain signal, a time domain signature, and a power signal, the signature can be extracted from multiple angles. Thus, more comprehensive and diversified characteristic information can be obtained, and the running state and performance of the inverter can be better described; representative features can be extracted by analyzing and processing the extracted features, including frequency energy, frequency peak, mean, variance, waveform morphology, total energy, average power, power factor, etc. The characteristics can better reflect the working state and characteristics of the inverter, and are helpful for subsequent tasks such as fault diagnosis, performance evaluation and the like.
By selecting and dimension-reducing the extracted features, the dimension of the features can be reduced, the calculation efficiency is improved, redundant or irrelevant features are removed, and the generalization capability of the model is improved. The feature selection and dimension reduction can help us to better understand the operation rule of the inverter, reduce the correlation between features and improve the accuracy and effect of subsequent tasks; by extracting the frequency domain signal, the time domain feature and the power signal, we can obtain the feature directly related to the inverter operation. The characteristics have clear physical significance, help us to better understand the inherent mechanism and performance of the inverter, and provide reference basis for fault diagnosis, fault prediction, optimization adjustment and the like.
The feature is subjected to dimension reduction treatment by adopting a principal component analysis method, and the method comprises the following steps of:
carrying out standardization processing on the feature data so that the mean value of each feature is 0 and the variance is 1;
according to the normalized characteristic data, calculating a covariance matrix among the characteristics;
performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors, wherein the eigenvalues represent variances of the covariance matrix in the direction of the eigenvectors, and the eigenvectors represent the directions of principal components of the covariance matrix;
sorting the characteristic values according to the sizes, and selecting characteristic vectors corresponding to the first k larger characteristic values as main components;
projecting the original characteristic data onto the selected main component to obtain dimension-reduced data, wherein the projection is realized through dot multiplication operation;
and determining a final dimension reduction dimension k according to the reserved total variance ratio.
The principal component analysis realizes data dimension reduction by selecting principal components in the covariance matrix, and reduces the number of features. The method is beneficial to simplifying the data set, reducing the computational complexity, reducing the correlation among the features and improving the accuracy and effect of the subsequent tasks; principal component analysis can automatically select those principal components that can account for a larger portion of the variance of the data based on the variance distribution of the data. By selecting the feature vectors corresponding to the first k larger feature values, the main information of the data can be reserved, and the features of the original data can be kept as much as possible.
Principal component analysis can identify and eliminate redundant features with lower variances. These redundant features provide less information on the data representation and therefore can be ignored by the principal component analysis, thereby reducing computational and storage overhead; the result of principal component analysis is a set of orthogonal principal components, each of which is a linear combination of the original features. These principal components tend to have better interpretability, which can help us understand the structure and characteristics of the data, and thus better interpret the results of the model.
By calculating the corresponding eigenvalues of each principal component, we can know the contribution degree of each principal component to the total variance. From the remaining total variance ratio, the final dimension k can be determined. Therefore, the number of the features can be reduced and the calculation efficiency can be improved while enough information is kept.
The method for establishing the fault database comprises the following steps:
monitoring the running state of each inverter in the inverter cluster, and recording the fault type, time and related characteristic data when faults occur;
cleaning and preprocessing the collected fault data, including abnormal value removal, noise filtering, calibration and standardization, so as to ensure the accuracy and consistency of the data;
the MySQL system is selected to store the collected fault data, a database table is created to record fault type, time and characteristic data information, and a data structure and index are set.
By monitoring the operating status of each inverter in the inverter cluster and recording the fault type, time and associated characteristic data, a complete fault database can be established. Therefore, the follow-up analysis and treatment of faults can be facilitated, and the rapid positioning and problem solving are facilitated; the collected fault data is cleaned and preprocessed, including abnormal value removal, noise filtering, calibration and standardization, so that the accuracy and consistency of the data can be ensured. The cleaning and preprocessing operations can improve the quality of the data, reducing erroneous decisions and false analyses due to data problems.
Selecting MySQL system as the database storing failure data may provide efficient data management and query functions. MySQL has good expandability and stability, supports concurrent read-write operation, and can accelerate query speed through indexes. By creating database tables and setting data structures and indexes, fault data can be better organized and managed; after the fault database is established, fault data can be conveniently shared and utilized. The system not only can be used for fault analysis and processing by engineers and technicians, but also can be used as a knowledge base for relevant researches and applications such as fault prediction, fault diagnosis and the like. The fault database can also perform data interaction and integration with other systems, so that wider data application is realized.
By establishing a fault database, fault data can be managed and analyzed, and a basis for supporting fault management and optimization is provided. Through statistics and analysis of fault data, a fault mode and a rule can be identified, corresponding improvement measures and optimization schemes are provided, and reliability and performance of the inverter cluster are improved.
The method for training and optimizing the model comprises the following steps:
acquiring fault type, time and characteristic data information from a fault database as a training data set;
according to the fault type and the feature data, using chi-square test to evaluate the correlation between the feature and the fault type, and selecting the feature with higher correlation for training the model;
dividing a training data set into a training set and a verification set, dividing the data set into a plurality of subsets by adopting a cross verification method, wherein one part of the data set is used as the verification set each time, and the other part of the data set is used as the training set so as to evaluate the performance of the model;
the convolutional neural network is selected for training, and in the training process, parameters of the model are adjusted according to the characteristic data of the training set and the corresponding fault type, so that the model can more accurately conduct fault prediction;
and evaluating the trained model by using the verification set, calculating indexes such as accuracy, recall rate, precision and the like of the model, and optimizing the model according to an evaluation result.
By analyzing the fault type, time and feature data in the fault database, the fault data can be fully utilized, useful information is extracted as a training data set, and features with higher correlation with the fault type are screened out for model training, so that the accuracy of a model is improved; the data set is divided into a plurality of subsets by adopting a cross-validation method, wherein one part of the data set is used as a validation set each time, and the other part of the data set is used as a training set, so that the performance of the model can be better evaluated, and meanwhile, the overfitting phenomenon is prevented. The structure of the convolutional neural network is selected for training, so that the spatial relationship and feature extraction of input data can be better processed.
By calculating and analyzing indexes such as accuracy, recall rate and precision of the model, the performance of the model can be evaluated, the problems of the model can be found, the model can be optimized and improved, the generalization capability and the prediction accuracy of the model can be improved, and the optimal fault prediction effect can be achieved; by establishing a fault prediction model, potential faults can be found in advance, repair can be performed in time, and fault diffusion and accumulation are avoided, so that the reliability and stability of the inverter cluster are improved.
The convolutional neural network architecture includes:
input layer: the state quantity of the photovoltaic inverter comprises voltage, current and power;
output layer: the output result is mapped to the [0,1] interval using a sigmoid activation function, indicating the probability of failure.
The convolutional neural network can effectively process the spatial relationship of input data, and the information such as state quantity and control command of the photovoltaic inverter is usually multidimensional and comprises parameters such as voltage, current and power. The convolution layer and the pooling layer can extract spatial features in the data, so that the model can learn and understand the complex multidimensional data better; in convolutional neural networks, the convolutional layer can extract features on input data through sliding convolutional kernels, and features with different sizes and shapes can be learned through different weights and offsets of a plurality of convolutional kernels. The characteristic extraction capability is strong, and the association mode between the state quantity of the photovoltaic inverter and the control command can be effectively captured.
The sigmoid activation function is used to map the output result to the [0,1] interval and represent the probability of failure, and such output result has interpretability. The probability of occurrence of different fault types can be known by analyzing the output result, and the fault detection and maintenance work guidance is facilitated; the convolutional neural network has fewer parameters, and the design of sharing weights and offsets reduces the complexity of the model, so that the training process of the model is relatively simple and efficient. Moreover, the structure and the feature extraction capability of the convolutional neural network enable the model to obtain a good prediction effect under the condition of a small sample size.
The method for predicting and judging the faults comprises the following steps: and judging the fault rate of each photovoltaic inverter in the photovoltaic inverter cluster according to the output result of the model, and determining the fault type of the photovoltaic inverter with the fault rate higher than 50%.
And through the output result of the model, the fault rate of each photovoltaic inverter in the photovoltaic inverter cluster can be predicted and judged. Thus, potential faults can be found in time, and corresponding maintenance and repair work can be carried out, so that larger loss caused by further expansion of the faults is avoided; the fault type of the photovoltaic inverter with the fault rate higher than 50% is determined, and standardized judgment can be carried out according to a set threshold value. This can simplify the failure judgment process and improve the efficiency. Moreover, photovoltaic inverters with failure rates higher than 50% often represent serious failures, requiring priority.
By determining the fault type of the photovoltaic inverter with a fault rate higher than 50%, repair and maintenance work can be performed in a targeted manner. The photovoltaic inverter with serious faults can be preferentially processed, the repair efficiency is improved, and the influence of the faults on the whole photovoltaic inverter cluster is reduced; the method of fault prediction may help discover potential signs of faults early. By monitoring and analyzing the fault rate change in the output result of the model, measures can be taken in advance to carry out maintenance work, so that faults are avoided, and the influence of the faults on the photovoltaic inverter cluster is reduced.
The method for alarming or processing measures comprises the following steps: and notifying operation and maintenance personnel to carry out maintenance treatment in a mode of short message, mail or APP.
The maintenance personnel is informed to carry out maintenance treatment in a mode of short messages, mails or APP, and the quick response of the problems can be facilitated. If a fault occurs, operation and maintenance personnel can timely receive the notification and timely take measures to process, so that larger loss caused by further expansion of the fault is avoided; the operation and maintenance personnel are informed to carry out maintenance treatment in a mode of short messages, mails or APP, and standardization and rapid transmission of information can be achieved. This allows the service personnel to quickly learn the severity of the problem and the action to be taken, thereby improving response speed and work efficiency.
By adopting a plurality of notification modes, the operation and maintenance personnel can be ensured to receive the notification in time, and the most convenient mode is selected for response. For example, if the operation and maintenance personnel are out of the business, a short message notification may be most convenient and timely; and the operation and maintenance personnel are informed of maintenance processing in a mode of short messages, mails or APP, so that problems can be found in real-time monitoring and timely processed. Therefore, operation and maintenance personnel can know the working state of the photovoltaic inverter cluster, quickly respond to faults, and ensure the operation efficiency of the photovoltaic inverter cluster to the greatest extent through diagnosis and maintenance.
The beneficial effects of the invention are as follows:
by continuously monitoring the operation data of the inverter cluster, the faults of the inverter can be detected in time, corresponding treatment measures are adopted in advance, and the fault expansion or larger loss is avoided; according to the method, the characteristic extraction and selection technology is utilized to extract representative characteristics from the inverter operation data, so that the fault prediction model is more efficient and accurate. Meanwhile, the accuracy and performance of prediction are further improved through model training and optimization.
Based on a large number of historical fault databases, a fault diagnosis model is constructed by utilizing a machine learning technology, so that a prediction result is more credible. According to the prediction result, corresponding measures can be timely taken to carry out maintenance, replacement and other operations, and the operation efficiency of the photovoltaic system is improved; the fault type of the photovoltaic inverter can be judged through the fault prediction model, and corresponding treatment measures can be adopted in a targeted manner for different types of faults, so that the maintenance efficiency and the success rate are improved; the method can realize automatic monitoring and fault prediction of the inverter cluster, reduce the workload of manual inspection and maintenance, and improve the operation and maintenance efficiency and reliability of the photovoltaic system.
The above-mentioned embodiments of the present invention are not intended to limit the scope of the present invention, and the embodiments of the present invention are not limited thereto, and all kinds of modifications, substitutions or alterations made to the above-mentioned structures of the present invention according to the above-mentioned general knowledge and conventional means of the art without departing from the basic technical ideas of the present invention shall fall within the scope of the present invention.
Claims (9)
1. The photovoltaic inverter fault prediction method is characterized by comprising the following steps of:
data acquisition and processing: collecting operation data of each inverter in the inverter cluster, including voltage, current and temperature information, and cleaning and preprocessing;
feature extraction and selection: extracting representative features from the operation data of each inverter, including spectral features, time domain features and energy features;
establishing a fault database: establishing a historical fault database of the inverters in the inverter cluster, wherein the historical fault database comprises fault types, time and characteristic data information;
model training and optimizing: model training using a fault database, evaluating performance of the model using cross-validation, and adjusting model parameters to improve accuracy;
and (3) predicting and judging faults: when new data enter the system, the new data are input into a fault diagnosis model to carry out fault prediction, whether the photovoltaic inverter cluster has faults or not is judged according to the output result of the model, and the type of the faults is determined;
alarm or treatment measures: and triggering an alarm or processing measure according to the fault prediction result.
2. The photovoltaic inverter failure prediction method according to claim 1, characterized in that: the method for collecting and processing the data comprises the following steps:
transmitting data to the internet of things device using the I2C, SPI or analog output interface to measure voltage, current and temperature;
cleaning and preprocessing the collected original data, including abnormal value removal, noise filtering, calibration and standardization;
selecting a relational database to store the acquired data, and setting a data acquisition system to transmit the data to a back-end server in real time;
the last 24 hours of data was retained in a once-per-minute fashion.
3. The photovoltaic inverter failure prediction method according to claim 1, characterized in that: the method for extracting and selecting the characteristics comprises the following steps:
converting the operation data of the inverter into frequency domain signals by using Fourier transformation, extracting frequency distribution conditions of the frequency domain signals, including frequency energy, frequency peak values and the like, and analyzing and processing the characteristics to obtain representative frequency spectrum characteristics;
dividing the operation data of the inverter into different time periods by using a sliding window method, and then carrying out statistics and analysis on each data section, wherein the statistics and analysis comprise mean value, variance and waveform morphology so as to extract representative time domain characteristics;
converting the operation data of the inverter into a power signal, and calculating the characteristics of total energy, average power, power factor and the like of the inverter through an accumulated square root algorithm to extract representative energy characteristics;
and selecting and reducing the dimension of the extracted features to reduce the feature dimension.
4. A photovoltaic inverter failure prediction method according to claim 3, characterized in that: the feature is subjected to dimension reduction treatment by adopting a principal component analysis method, and the method comprises the following steps of:
carrying out standardization processing on the feature data so that the mean value of each feature is 0 and the variance is 1;
according to the normalized characteristic data, calculating a covariance matrix among the characteristics;
performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors, wherein the eigenvalues represent variances of the covariance matrix in the direction of the eigenvectors, and the eigenvectors represent the directions of principal components of the covariance matrix;
sorting the characteristic values according to the sizes, and selecting characteristic vectors corresponding to the first k larger characteristic values as main components;
projecting the original characteristic data onto the selected main component to obtain dimension-reduced data, wherein the projection is realized through dot multiplication operation;
and determining a final dimension reduction dimension k according to the reserved total variance ratio.
5. The photovoltaic inverter failure prediction method according to claim 1, characterized in that: the method for establishing the fault database comprises the following steps:
monitoring the running state of each inverter in the inverter cluster, and recording the fault type, time and related characteristic data when faults occur;
cleaning and preprocessing the collected fault data, including abnormal value removal, noise filtering, calibration and standardization, so as to ensure the accuracy and consistency of the data;
the MySQL system is selected to store the collected fault data, a database table is created to record fault type, time and characteristic data information, and a data structure and index are set.
6. The photovoltaic inverter failure prediction method according to any one of claims 1 to 5, characterized in that: the method for training and optimizing the model comprises the following steps:
acquiring fault type, time and characteristic data information from a fault database as a training data set;
according to the fault type and the feature data, using chi-square test to evaluate the correlation between the feature and the fault type, and selecting the feature with higher correlation for training the model;
dividing a training data set into a training set and a verification set, dividing the data set into a plurality of subsets by adopting a cross verification method, wherein one part of the data set is used as the verification set each time, and the other part of the data set is used as the training set so as to evaluate the performance of the model;
the convolutional neural network is selected for training, and in the training process, parameters of the model are adjusted according to the characteristic data of the training set and the corresponding fault type, so that the model can more accurately conduct fault prediction;
and evaluating the trained model by using the verification set, calculating indexes such as accuracy, recall rate, precision and the like of the model, and optimizing the model according to an evaluation result.
7. The photovoltaic inverter failure prediction method according to claim 6, characterized in that: the convolutional neural network architecture includes:
input layer: the state quantity of the photovoltaic inverter comprises voltage, current and power;
output layer: the output result is mapped to the [0,1] interval using a sigmoid activation function, indicating the probability of failure.
8. The photovoltaic inverter failure prediction method according to claim 7, characterized in that: the method for predicting and judging the faults comprises the following steps: and judging the fault rate of each photovoltaic inverter in the photovoltaic inverter cluster according to the output result of the model, and determining the fault type of the photovoltaic inverter with the fault rate higher than 50%.
9. The photovoltaic inverter failure prediction method according to claim 8, characterized in that: the method for alarming or processing measures comprises the following steps: and notifying operation and maintenance personnel to carry out maintenance treatment in a mode of short message, mail or APP.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311285987.1A CN117318614A (en) | 2023-10-07 | 2023-10-07 | Photovoltaic inverter fault prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311285987.1A CN117318614A (en) | 2023-10-07 | 2023-10-07 | Photovoltaic inverter fault prediction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117318614A true CN117318614A (en) | 2023-12-29 |
Family
ID=89261694
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311285987.1A Withdrawn CN117318614A (en) | 2023-10-07 | 2023-10-07 | Photovoltaic inverter fault prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117318614A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117929904A (en) * | 2024-03-20 | 2024-04-26 | 深圳市广晟德科技发展有限公司 | Inverter aging test method, device and storage medium |
| CN117951633A (en) * | 2024-03-27 | 2024-04-30 | 中节能甘肃武威太阳能发电有限公司 | Photovoltaic power generation equipment fault diagnosis method and system |
| CN117929904B (en) * | 2024-03-20 | 2024-06-21 | 深圳市广晟德科技发展有限公司 | Inverter aging test method, device and storage medium |
-
2023
- 2023-10-07 CN CN202311285987.1A patent/CN117318614A/en not_active Withdrawn
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117929904A (en) * | 2024-03-20 | 2024-04-26 | 深圳市广晟德科技发展有限公司 | Inverter aging test method, device and storage medium |
| CN117929904B (en) * | 2024-03-20 | 2024-06-21 | 深圳市广晟德科技发展有限公司 | Inverter aging test method, device and storage medium |
| CN117951633A (en) * | 2024-03-27 | 2024-04-30 | 中节能甘肃武威太阳能发电有限公司 | Photovoltaic power generation equipment fault diagnosis method and system |
| CN117951633B (en) * | 2024-03-27 | 2024-06-11 | 中节能甘肃武威太阳能发电有限公司 | Photovoltaic power generation equipment fault diagnosis method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108415789B (en) | Node fault prediction system and method for large-scale hybrid heterogeneous storage system | |
| CN110287552B (en) | Motor bearing fault diagnosis method and system based on improved random forest algorithm | |
| CN111507376B (en) | Single-index anomaly detection method based on fusion of multiple non-supervision methods | |
| CN115034248A (en) | Automatic diagnostic method, system and storage medium for equipment | |
| CN111353482A (en) | LSTM-based fatigue factor recessive anomaly detection and fault diagnosis method | |
| CN111259947A (en) | Power system fault early warning method and system based on multi-mode learning | |
| CN112414694B (en) | Equipment multistage abnormal state identification method and device based on multivariate state estimation technology | |
| CN117318614A (en) | Photovoltaic inverter fault prediction method | |
| CN116976707B (en) | User electricity consumption data anomaly analysis method and system based on electricity consumption data acquisition | |
| CN115441456A (en) | Fault diagnosis method and device for power grid dispatching support system | |
| CN117078227A (en) | Environment monitoring operation and maintenance platform based on identification analysis | |
| CN117194171A (en) | Heterogeneous cloud resource panoramic anomaly detection system | |
| CN117113135A (en) | Carbon emission anomaly monitoring and analyzing system capable of sorting and classifying anomaly data | |
| CN116714469A (en) | Charging pile health monitoring method, device, terminal and storage medium | |
| CN116720983A (en) | Power supply equipment abnormality detection method and system based on big data analysis | |
| CN116664098A (en) | Abnormality detection method and system for photovoltaic power station | |
| CN114154567A (en) | Wind power plant station operation data anomaly identification method based on machine learning | |
| CN115378000A (en) | Power distribution network operation state evaluation method based on interval two-type fuzzy clustering analysis | |
| CN114597886A (en) | Power distribution network operation state evaluation method based on interval type two fuzzy clustering analysis | |
| CN115598459A (en) | Power failure prediction method for 10kV feeder line fault of power distribution network | |
| CN111652313A (en) | Multi-source heterogeneous data mining method based on cluster analysis | |
| CN117391458B (en) | Safety production risk detection and early warning method and system based on data analysis | |
| CN115711206B (en) | Wind driven generator blade icing state monitoring system based on clustering weight | |
| CN117951633B (en) | Photovoltaic power generation equipment fault diagnosis method and system | |
| CN115511106B (en) | Method, device and readable storage medium for generating training data based on time sequence data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20231229 |