Search documents

Browse topics

Document details

RoboPV: an Aerial Robots’ Embedded Code for Intelligent Monitoring and Fault Detection of Large-Scale PV Plants Using Deep Neural Network Models
A.M. Moradi Sizkouhi, S.M. Esmailifar, M. Karimkhani, M. Aghaei
Architecture, Photovoltaic (PV) Power Plant, Autonomous Aerial Monitoring, Aerial Robots, Encoder-Decoder
PV Systems Engineering, Integrated/Applied PV
Subtopic: Operation, Performance and Maintenance of PV Systems
Event: 8th World Conference on Photovoltaic Energy Conversion
Session: 4DO.2.6
1094 - 1098
ISBN: 3-936338-86-8
Paper DOI: 10.4229/WCPEC-82022-4DO.2.6
0,00 EUR
Document(s): paper


In this paper, a novel embedded software package, called RoboPV, is introduced for autonomous aerial monitoring of large-scale photovoltaic (PV) plants. RoboPV enables aerial robots to automatically perform aerial monitoring of PV plants, from optimal trajectory planning to real-time fault detection using deep neural network models. This software package is made of four systematically integrated parts: boundary detection, path planning, dynamic processing, and fault detection. The first requirement of efficient path planning is to identify the PV plant boundary closed curve. This requirement is prepared by an encoder-decoder deep learning architecture which is trained by labeled real flight aerial images of PV plants. Subsequently, a novel path planning algorithm is conducted by RoboPV to design an optimal flight path with full coverage of whole regions of the PV plant. Another deep neural network model has been also trained for real-time fault detection in PV plants through PV faults labeled dataset. Using this highly precise trained neural network, aerial images are analyzed in real-time during the flight and when a probable fault is detected, RoboPV commands the aerial robot to maneuver and concentrate more on the detected fault. In this study, several decision-making and maneuver algorithms are developed for various real-world flight conditions to improve the performance of RoboPV during an autonomous aerial inspection. RoboPV is a modular processing library that can be installed on any micro-computer processor with a low computational burden. Moreover, supporting the MAVLink communication protocol enables RoboPV to connect with an intelligent Pixhawk flight autopilot and navigate a wide range of multi-rotors. To demonstrate the performance of RoboPV, a six degrees of freedom dynamic model of a multi-rotor has been developed in a Simulink environment, and processor in the loop (PIL) tests have been performed for aerial monitoring of two different real megawatt-scale PV plants.