Photovoltaic plant condition monitoring using thermal images analysis by convolutional neural network-based structure

Abstract

The size and the complexity of photovoltaic solar power plants are increasing, and it requires an advanced and robust condition monitoring systems for ensuring their reliability. This paper proposes a novel method for faults detection in photovoltaic panels employing a thermographic camera embedded in an unmanned aerial vehicle. The large amount of data generated by these systems must be processed and analyzed. This paper presents a novel approach to identify panels to detect hot spots, and to set their locations. Two novels region-based convolutional neural networks are unified to generate a robust detection structure. The main contribution is the combination of thermography and telemetry data to provide a response of the panel condition monitoring. The data are acquired and then automatically processed, allowing fault detection during the inspection. A detailed description of the methodology is presented, including the different stages to build the neural networks, i.e. the training process, the acquisition and processing of data and the outcomes generation. A thermographic inspection of a real photovoltaic solar plant is done to validate the proposed methodology. The accuracy, the efficiency and the performance of the approach under different real scenarios are evaluated statistically obtaining satisfactory results.

Low-Cost Electronics for Online I-V Tracing at Photovoltaic Module Level: Development of Two Strategies and Comparison between Them

“…Perhaps the most widely used method is infrared thermography (IRT), for hot spot detection [8,9]. Some authors use the thermographic images obtained by drones to subsequently use artificial intelligence systems for automatic fault detection [10]. Another important technique is based on electroluminescence (EL) images [11], where in some cases, low-cost systems are used [12], and in many cases, the use of drones is also possible [5].…”

“…The last one depends on the ADC resolution and the maximum dynamic range of the signals measured (adjusted with the resistive divisors in the case of voltages). For the module voltage, the resistor divider has been calculated for a full scale of 50 V, while the resolution of the 10 bit ADC is 2 10 levels, so the quantification noise has a maximum amplitude of 50/2 10 = 48.8 mV. Our direct measurements over the module with an oscilloscope show a composed noise (thermal + interference) with a maximum amplitude of 2 mV; thus, the main contribution to noise is the quantification one.…”

Autonomous Underwater Vehicles and Field of View in Underwater Operations

“…The mapping system is focused on the data acquisition employing optical, magnetic or acoustic systems, e.g., multi-beam novel echo sounders and sonars [20], together with individual sensors and cameras with data/image processing [21,22]. The combination of visual and sonar data leads to new improvements in comparison with traditional monitoring processes.…”

Condition Monitoring of Manufacturing Processes under Low Sampling Rate

“…Condition-based monitoring (CBM) is defined as the act of monitoring the condition of a machine or a process [1], while prognostics and health management (PHM) is defined as an algorithmic way of detecting, predicting, monitoring and assessing operation problems and health changes of systems [2] as well as taking decisions on them. Both are used to monitor machinery, such as pumps [3], bearings and gears [4,5], electronics [6], plane parts [6] and machine tools (Computer Numerical Control (CNC)) [2,7]. This field of research is backed by a strong interest of the industry towards Smart Manufacturing, also called Industry 4.0, which aims at the autonomous management of production through virtualization of the production chain [8].…”