Why do we physically inspect PV sites?
Before diving into the details of aerial inspection tools in PV systems, it is important to consider why we perform physical inspections of PV sites in the first place. With the rise of advanced data analytics, improved sensor networks, and seemingly endless Big Trends (Big Data! Machine Learning! Blockchain?), why are physical site inspections required, and will they be required in the future?
Historically, the majority of operational plans for PV assets include physical site inspections, and broadly speaking the motivation for these inspections is driven by two main factors:
1. Underlying uncertainty of system analytics
2. Detection of non-energy risk factors
Academic literature has established that the accuracy of PV system modelling is between 5%-10%.** This uncertainty can be decreased by looking at peer to peer inter-comparisons such as a ranking of normalized combiner outputs. In this case, the uncertainty can be decreased to around 3-4% which translates to a probability of false negative (i.e. missing a single string out) of approximately 60% for a site with 28 strings per combiner. This measurement is also susceptible to shifts in global mean production, or in other words, faults which occur in the majority of combiners would not be detected.
In addition to underlying uncertainty, production data analytics alone also cannot identify non-energy risk factors. For example, groupings of hot-spots or sub-string failures on a site will not cause a near-term appreciable energy loss, but can be signals of potentially serious serial defects on a site.
Because of these issues, physical site inspections are a standard part of the operating plans for PV assets. Traditionally, these inspections have been performed by an I-V curve trace of all or a subset of strings on a site, and this data is used to answer two questions:
1. Is a string or module broken?
2. Is the system degrading over time?
There is a growing recognition, however, that this tool is not an effective tool for achieving either of these goals. I-V trace is generally too expensive and labor intensive to be applied to an entire site, and exposes technicians to high voltage DC components during the testing procedure, posing an arc-flash hazard.
Because of these deficiencies there has been a significant shift in the industry over the past few years to move towards aerial inspections in lieu of traditional manual physical inspections..
What are aerial Thermal inspections?
PV aerial thermal inspection refers to the collection of high resolution infrared and visible imagery of a site from an aerial platform. The fundamental concept being used is simple: All modules are receiving the same amount of energy for the sun, and those that are not converting that energy into electricity will turn it into heat. Therefore, energy losses will show up as module heating.
Aerial thermal inspections can be used to detect a wide variety of site defects, from common fault modes such as hot-spots, module breakage, sub-module faults (i.e. diode engagement) and string outages, to more subtle defects such as Potential Induced Degradation (PID), MPPT issues and junction box resistance.
Typically, these inspections are performed annually as part of site preventative maintenance in lieu of traditional string testing. Ideally, the scans are performed in advance of regular preventative maintenance work, especially for unmanned sites, as this ensures that technicians can arrive with the knowledge, tools, and equipment needed to quickly remediate issues.
Aerial inspection process
The aerial inspection process can be considered in three stages: Acquisition, Analytics and Remediation.
Acquisition is performed from an aerial platform, either UAS (drone) or manned aircraft. In the case of a drone, the drone is transported to the site, and a flight team will perform initial site safety checks, set up defined landing zones and proceed with the inspection. In the case of aircraft inspection, the aircraft will begin from a nearby airfield and fly the site as part of a regular flight pattern.
Both platforms will fly a designated pattern of successive scans over the site, building up an imagery database covering all modules in the system. For drones, the IR cameras which can be used have a longer integration time, meaning flight speed is limited to reduce the possibility of motion blur. In addition, the system must land to change batteries every 15-20 minutes of operation. This results in a capture rate for standard resolution inspections of around 20MW/day, though operations at lower resolution are able to cover more than 40MW/ day.
Aircraft inspections are able to fly faster and with fewer passes due higher resolution and integration times in their camera systems, and have endurance for full site inspections. Therefore, aircraft based systems able to achieve a scanning rate of up to 150MW/hr.
It is important that proper irradiance is maintained during the period of inspection in order to ensure all possible fault modes are detectable. IEC 62446-3* specifies a minimum irradiance threshold of 600W/m2 in the plane of array of the PV modules, and this threshold is mirrored in the NREL O&M best practices guide. There is an additional advantage to performing these inspections over a short time interval, as it allows for intercomparing of the thermal properties of different portions of an array while the system is at a relative thermal equilibrium.
Once data is collected for a site, it must be analyzed to produce an actionable report. This is the most critical stage of inspection, as a poor analysis can lead to missed faults and poor localization leading to inefficiency during remediation. A typical site inspection can yield anywhere from from multiple gigabytes to multiple terabytes of raw sensor data, and advanced systems must be put in place to reduce this data into a usable output for site operations.
For most aerial applications, commercial tools are available using a process called photogrammetry to generate stitched composite imagery of a site by finding unique “keypoints” in successive frames. Unfortunately, this tool cannot be used effectively when applied to IR aerial inspections of PV systems, due to the regularity of the imagery collected. When keypoint matching is attempted, multiple matches will occur causing image distortion and missed modules. Because of this, standard aerial post-processing tools cannot be utilized, and software utilizing a fundamentally different methodology for analysis must be utilized.
The methodology employed by Heliolytics utilizes Artificial Intelligence (AI) processes to analyze these complex datasets. AI tools provide a powerful way to classify a defect according to not only the distribution of thermal deviations within the module but also to correlation to collected visible imagery, and information from the faults around it.
These inspections can find a wide variety of issues, including:
Once the data has been analyzed, it must be passed to stakeholders in the correct format to allow for actionable remediation. For field personnel, this means accurate site maps which clearly identify the location and type of defect to be remediated. For site operators, it is important that the results from inspections can be integrated into existing digital workflows, to make sure that faults are tracked through their remediation phase, and records can be kept of the process.
Aerial inspections represent a new paradigm in the inspection of PV systems. These tools offer increased accuracy and safety as compared to traditional inspection tools, and provide the ability to effectively inspect 100% of a PV system on a regular basis. Through offering increased site output, reduced labor costs and increased asset visibility, these tools allow for new levels of asset optimization.
Why (didn't) we use drones for inspections?
Why do we use piloted aircraft for our inspections?
As Heliolytics has grown over the past six...
String to inverter: BoS component failures
Electrical Balance Of System (BOS) components can be key failure points in PV installations, and...
(Webinar) Data Insights on PV reliability
In this webinar presented at Solar Plaza Asset Managment North America, we discuss DC health...