Photovoltaic array fault diagnosis method based on fine-tuning densely connected convolutional neural network

Abstract

The invention relates to a photovoltaic array fault diagnosis method based on a fine-tuning densely connected convolutional neural network. First, collect electrical characteristic data and environmental data under actual working conditions, and then use Simulink to build a model array to simulate actual working conditions; obtain simulated For the electrical characteristic data, secondly, through the mutation point detection algorithm, the abnormal data in the actual and simulation are eliminated, and the complete electrical waveform data is obtained, sampled, compressed, and spliced ​​into a two-dimensional feature matrix. Then, design a densely connected convolutional neural network, use the simulation training set and the Adam optimization algorithm to pre-train the network, and then use a small amount of actual working condition training set to fine-tune the network. Finally, the FT-DenseNet fault diagnosis network is used to detect and classify the photovoltaic power generation arrays under the test set of working conditions. The method of the invention obtains a classification network with high precision, strong robustness and good generalization ability under the condition of small samples, and can effectively improve the accuracy of photovoltaic array fault detection and classification.

Description

technical field [0001] The invention relates to the technical field of photovoltaic power generation string fault detection and classification, in particular to a photovoltaic array fault diagnosis method based on fine-tuning densely connected convolutional neural networks. Background technique [0002] In recent years, solar energy has been widely developed as a promising renewable energy. Photovoltaic energy is a form of solar energy that plays an integral role in curbing global warming, reducing the use and emissions of fossil fuels, and according to the latest announcement from the World Energy Organization, global PV installations and power generation are growing day by day. With the rapid development of the photovoltaic industry and the rapid growth of photovoltaic installed capacity, more and more attention has been paid to the service life and safety of photovoltaic arrays. Photovoltaic power plants are mostly built in sparsely populated and large areas. Frequent m...

AI Technical Summary

Problems solved by technology

Such as artificial neural network (ANN), decision tree (DT), random forest (RF), probabilistic neural network (PNN), wavelet neural network (WNN), support vector machine (SVM), and other traditional machine learning algorithms, in photovoltaic Significant achievements and breakthroughs have been made in the field of fault diagnosis and detection, but they all require a large number of training samples, and it is difficult to obtain fault samples of actual photovoltaic arrays, which cannot be truly applied to large photovoltaic arrays.

Some unsupervised learning algorithms, such as density peak clustering algorithm (DPCA), fuzzy C-means algorithm (FCM) and other algorithms have achieved good results in fault diagnosis, but their accuracy is still limited

With the increasing popularity of deep learning, convolutional neural networks and cyclic neural networks have entered the sight of photovoltaic fault diagnosis. Their powerful feature extraction capabilities can obtain high-quality fault features with strong representation capabilities, which further contribute to the accuracy of classification. improved, but still requires the support of a large number of samples

Images