학술논문

Simultaneous Detection and Classification of Dust and Soil on Solar PhotoVoltaic Arrays Connected to A Large-Scale Industry: A Case Study
Document Type
Conference
Source
2022 18th International Conference on the European Energy Market (EEM) European Energy Market (EEM), 2022 18th International Conference on the. :1-6 Sep, 2022
Subject
Engineering Profession
Power, Energy and Industry Applications
Industries
Renewable energy sources
Training data
Carbon dioxide
Soil
Data models
Cleaning
Solar panel
soiling
detection
image processing
energy efficiency
Language
ISSN
2165-4093
Abstract
Solar PV technology has advanced significantly in recent years as a result of the widespread adoption of clean energy resources, and it is now the most preferred renewable energy resource. Large-scale industries use PVs in conjunction with co-generation plants to reduce carbon emissions while increasing revenue. However, due to continuous carbon emissions from co-generation plants, dust and carbon particles accumulate on the PV panels, necessitating extensive cleaning at the expense of increased labor expenses and decreased net revenue for operation. Furthermore, the efficiency of co-generation plants, which rely heavily on PV plant operation, is being harmed, limiting large-scale PV resource implementation in the industrial sector. To that end, this paper presents a novel image classification method for increasing the efficiency of large-scale industry co-generation plants by simultaneously detecting and classifying dust and soil on PV arrays. The proposed method takes into account PV array images captured by a high-resolution camera, industry meteorological data for different seasons, and the operating characteristics of a co-generation power plant as training data. The training data is then fed into the deep learning framework, which consists of coupled Convolutional neural network (CNN)-based models for feature extraction and Long short-term memory (LSTM) models for learning sequential meteorological data to enable joint classification and detection tasks. The experimental results show that our proposed method achieve an accuracy of 50.91% for CNN (Image data only), 94. 09% for LSTM (Meteorological data only), and 96.54% for CNN-LSTM (with combined image and meteorological data). The proposed system can help engineers of large-scale industry with large PV plants and co-generation for optimal cleaning regimes of PV panels and maximize co-generation plant efficiency.