TECHNOLOGY I SAFETY
well as have a minimal chance of false detection. The cost of nuisance trips can be quite high. From the UL1699B detection requirements, an arc detect system must be able to quickly, but reliably, detect the presence of an arc without misinterpreting other interference to be an arcing event.
Due to the complexity of these detection requirements, a microcontroller (MCU) capable of digital signal processing is typically used. But the use of an MCU adds complexity to a critical application such as arc detection. A subtle bug in the programming could result in the MCU locking up or missing an arcing event. Even rigorous coding practices can’t protect against a fault in an MCU or random events, such as random bit flipping due to radiation. With a system out in the field for more than 20 years, such rare events are not so unlikely. UL1699B requires all MCU-based arc detection systems to follow a modified version of UL1998, which addresses software implementation for fail-safe and fault tolerant systems.
The firmware in the RD-195 utilizes SafeTI, a modular safety library developed for TI’s C2000 MCU family. This library includes functionality to ensure that the MCU instructions are not corrupted, the MCU is properly performing operations, MCU sub-systems (peripherals) are operating correctly, and if a fault is detected, the system enters a safe mode. This comprehensive library is useful for applications beyond that of arc detection. This library has been UL certified as recognized components based on UL1998:2008 Class 1 standard. As an additional level of protection, Arc Detection DSP routines are compartmentalized for ease of inclusion into a customized application. The Arc Detection Library includes self-testing to catch any errors in processing. This library is also UL certified as a recognized off- the-shelf (OTS) software component.
Development of the arc detection DSP used in the RD-195 took some time. Several different signal processing methods were evaluated during development. After extensive evaluation, time domain techniques, which attempted to extract the arc signal from any periodic interferer signals, were deemed ineffective or too complex. A learning algorithm was not adopted since an arcing event could occur while learning, which would compromise effective detection. Techniques which monitored several frequencies for absolute level had potential issues with cross-talk from non-arcing strings. Other methods were determined to require too much processing to fit within our system constraints. Consequently, a new method was developed for this application using heuristics controlled by a set of adjustable parameters to dynamically scan for interfering
Figure 7: Monitored string power over time
signals and mathematically filter an interfering signal. A set of parameters were determined to provide effective detection for the majority of inverters on the market. The collection of all inverters on the market presented conflicting constraints on the detection parameters. For some inverters, 50 kHz was the optimum detection band, while for others the interference was at its worst extreme at 50 kHz. As a result, for some of the uncommon inverters (such as the one whose response is shown in Figure 6) the solution needed to use a different set of detection parameters.
During development, a technique used for optimizing DSP detection parameters was to automatically collect raw ADC data when a false detection occurred. In this way, the detection parameters could be adjusted to avoid false detection from these types of events. The RD-195 was routinely evaluated on a variety of PV systems for extended periods to ensure that the final parameters used are robust against false detection events. Some of these evaluations ran for more than a month, with occasional changes to system and periodic checks of the arc detection characteristics. During these extended tests the status of the tested RD-195 was monitored and the string current and voltage was recorded. Figure 7 illustrates the power during one extended test.
It is easy to overlook the need for a self-test circuit, but it provides a comprehensive verification of the entire signal path. Without the self-test circuit, a fault in the transformer, the input protection, filtering, or the ADC might not be detectable. As a result, effective detection would not be present. The self-test circuit in the RD-195 generates a noise pattern which emulates an arcing event. This noise is coupled into the transformer via a dedicated secondary. It goes through the signal path and is processed by the Arc Detect Library. The only fault this self-test implementation will miss is if the primary path of the transformer is open – this fault would be quickly apparent as no power is being produced.
Summary
Detection of an arc is not easy to implement as there are many critical aspects to focus on – from interference to dynamic range to self-testing. A fuse does a great job for the protection it provides, but it can’t provide arc detection. TI has worked to provide an extensible solution which can fit nearly any PV system needs. To assist developers, TI offers the RD-195 evaluation board with a C2000 MCU running UL1699B recognized firmware.
Figure 6: Some inverter interference is extremely difficult to differentiate from an arcing event
©2013 Permission required. Angel Business Communications Ltd.
Issue III 2013 I
www.solar-uk.net 35
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48