Fault detection method for solar photovoltaic cell panels based on deep learning

Abstract

The invention discloses a fault detection method for solar photovoltaic cell panels based on deep learning. The method includes the following steps: cleaning collected sequential current data of a photovoltaic cell panel; extracting the transverse and longitudinal features of the processed current data; generating a detection data set from the extracted features and the current value of the original cell panel; processing the data set through a convolution neural network to obtain deep features of current data; and learning the features of training data in time dimension through a long and short term memory model, thus completing fault detection of the photovoltaic module. The experimental verification of massive current data of power plants proves that the method has accuracy of above 90%in fault detection of photovoltaic cell panels and has practicability and convenience in deployment and application in power plants.

Description

technical field [0001] The invention relates to the field of fault detection of solar photovoltaic panels, in particular to a method for detecting faults of solar photovoltaic panels based on deep learning. Background technique [0002] In recent years, as the environmental pollution caused by traditional fossil energy has become more and more serious, the demand for clean energy such as solar energy has increased, and more and more photovoltaic power stations have been built on the ground. The problems encountered also increased. Photovoltaic power stations are usually built in the wilderness with inaccessible people and harsh environmental conditions. A power station usually covers an area of ​​thousands of acres and has tens of millions of photovoltaic modules. However, the daily operation and maintenance personnel of the power station usually do not exceed about ten people. There are many kinds of faults. Accurate positioning and identification of fault types of photovo...

AI Technical Summary

Problems solved by technology

Due to the different construction periods of these photovoltaic power stations, there are great differences in the types of purchased equipment specifications. Even in the same photovoltaic power station, the replacement of a large number of photovoltaic modules due to equipment wear and tear will also cause equipment in the same power station. The old and the new are mixed, and the current data in the same time period between the branches under the same combiner box will also have obvious differences due to different equipment losses

Therefore, the method based on mathematical modeling of photovoltaic arrays cannot accurately detect large-scale photovoltaic arrays.

At the same time, power stations are often built in the suburbs with harsh environmental conditions, which poses great difficulties for obtaining infrared images of photovoltaic modules in a timely, safe and accurate manner.

Moreover, based on economic considerations, it is impossible for photovoltaic companies to update and add a variety of sensors and new data acquisition equipment for photovoltaic power plants that have been built and put into use for many years, so the multi-sensor method cannot be applied to existing power plants

Images