Capacitor choice is key to solar PV economics
While efforts continue in many laboratories on thin-film and amorphous-silicon cells, it is the mono-crystalline cell that continues to lead in efficiency, with researchers seeking every possible percentage point beyond the low-20% region. That hard-won conversion efficiency can easily be wasted and the very feasibility of solar PV as a reliable energy source challenged, without an effective design in the other – and in many respects more critical – major component of the system: the inverter. PV cells produce DC, but very few applications employ that DC output directly. Most, perhaps 95%, provide AC power to conventional electrical installations, and feed that power into the AC grid. Within the renewable-energy sector, it is widely recognized that the critical component in the power chain is the inverter.
Inverters – the weakest link?
The inverter must convert the variable DC devices from the PV cell array into AC that is tightly regulated to grid voltage levels; that is of sufficiently good (“clean”) waveform quality to be acceptable to the grid operator; and that is closely synchronized to grid frequency, with the precise phase relationship to the grid waveform to produce the correct flow of power. It must do so in demanding – even extreme – environmental conditions, maintenance-free, ideally for an expected minimum lifetime of 20 or 25 years.
Any energy-production system must be demonstrably cost-effective; arguments continue about the lifetime cost equation of many renewable-energy technologies, but it is clear that capital costs cannot be too high; and that any significant maintenance costs are, effectively, unaffordable. If the system is to recover its own energy-cost-of-manufacture, and then go on to provide years of return-on-investment, it must be a fit-and-forget installation.
The most common PV system architecture in today’s installations connects a string of modules together in series, yielding a high DC voltage that is fed to a central inverter that generates the AC to feed to the grid. One inverter for multiple modules enables some degree of control of capital cost of that element of the installation, but has a number of drawbacks. High DC voltages must be routed around the installation: with accessible cabling exposed to the elements potentially represents a safety hazard. Resistive losses in that series-connected DC chain can be significant, but of greater concern is the impact on overall system performance of any impairment of the PV array.
To greatly over-simplify, a PV cell may be modeled as a voltage source with an internal resistance that varies inversely with level of illumination: less incident light equals higher resistance and less output. If one module in a “string” is shaded, or dirty, or failing in some way, the performance of the entire array is degraded as a consequence of the series connection.
Micro inverter configurations
The industry has recognized for many years that a more desirable configuration is to have one inverter per module; the so-called micro inverter architecture. Each module, via its micro inverter, is separately and directly connected to the grid (subject to energy-flow metering for fiscal purposes). This presents many challenges which have only recently been successfully resolved with an innovative design, one that automatically operates at the maximum power-transfer point for every module, at all times.
It is immediately apparent that such a design must be manufacturable within strict costs limits, which implies everything that conventionally accompanies designs intended for volume production – a manageable bill-of-materials, and operation out-of-the-box with no manual set-up. But the greater challenge is reliability.
Today’s electronic circuitry is conventionally thought of as very reliable, but the design challenge involved in a micro inverter design is of a different order to almost any other discipline. The micro inverter will be mounted behind its accompanying module – in the open air, on a rooftop or similar location – and will therefore be subjected to extreme environmental conditions. This can range from severe cold on a winter’s night, to perhaps 85ºC in full sunlight (and therefore, at full power output); with both driving rain on one hand, yet poor airflow for cooling on the other.
25 years is well over 200,000 hours, a figure that simply dwarfs the rated lifetime of the great majority of electronic components. Many items of electronic equipment – even mass-produced items – function well on such timescales but they generally operate in a relatively benign environment. Some of the most demanding specifications are issued by the automotive industry, but the real operating life of most vehicles is only a few thousand hours. It is apparent that despite its restricted cost targets, the micro inverter’s components list and build quality will have to be of a very high order.
Components to avoid
Within the roll-call of components available to the electronic system designer, however, there are two stand-out devices that cannot meet the demands of a micro inverter architecture; electrolytic capacitors, and opto-couplers.
A premium-quality industrial electrolytic capacitor has a rated (operating) lifetime of just 7000 hours , clearly far below that required for the micro inverter design. It is well-known that electrolytics have failure mechanisms that are likely to come into play in a design such as a solar-module inverter; the electrolyte tends to “dry out” in high-temperature operation; and faced with extreme temperature cycling, seals can crack and degrade, allowing that loss-of-electrolyte process to accelerate.
To avoid this problem, Enecsys created an innovative architecture for its micro inverter that eliminates the need for inherently low-lifetime components. In an inverter, and with a design constraint that the unit must operate at the point of maximum power-transfer (that is, maximum available power extracted from the module and fed to the grid) at all times, some form of energy storage is unavoidable. The module is producing DC while the inverter outputs AC: at the zero-voltage-points of the AC waveform, the inverter must continue to draw power from the PV module, and store that power, albeit on a tens-of-milliseconds timescale.
A conventional design would employ a substantial electrolytic capacitor adjacent to the PV module itself, as an energy reservoir. Generation of the AC waveform through its sine-wave cycle, at the inverter output, is reflected at the inverter input (PV module terminals) as a varying current: without the energy reservoir, the current waveform peaks would cause the PV module output voltage to dip, shifting operation away from the ideal maximum power transfer point. This is the case for conventional inverter architectures of both single- and dual-stage layout, shown in figure 1.
Figure 1: Conventional inverter designs need large electrolytic capacitors to act as energy reservoirs.
The inverter design created by Enecsys uses a two-stage configuration but with the key innovation that the bulk energy storage is re-located to a point between first and second stages. Directly connected to the PV module is a voltage-amplification stage – in effect, a high-operating-frequency DC/DC converter, that steps up the module output to a much higher DC level – an average level of about 405V, as shown in figure 2.
Figure 2: Creating a high DC-link voltage at 2X the grid frequency reduces the capacitance needed for energy storage.
Then, this DC link feeds a second stage that Enecsys calls a Buck-CSI. This provides voltage step-down (because the DC link voltage is always above the mains AC peak level) and current shaping, driving a power MOSFET (from Infineon’s CoolMOS range) to feed power out to the grid.
The design massively reduces the value of storage capacitor it needs, in two ways.
Firstly, it exploits the fundamental relationship that stored energy is proportional to the square of voltage. (E = CV2/2) The higher voltage of the DC link means a much lower value for C, for the same available energy. In a further innovative step, Enecsys’ engineers designed the Buck-CSI stage to be tolerant of a very high level of ripple on its input. In a conventional design, the bulk storage capacitor would be required to smooth the ripple to negligible levels to ensure correct operation of the output stage; this design can tolerate as much as 120V of ripple on the DC link. The topology of the inverter is shown in figure 3.
Figure 3: The Enecsys micro inverter topology moves energy storage to the DC-link, enabling high reliability, plastic film capacitors to be used.
The outcome of this is a design in which the largest, bulk-capacitor needed, for an inverter rated at 240W output, is around 30 µF. This brings it within the scope of robust and reliable capacitor technologies that avoid all of the problems associated with electrolytic types. Enecsys has used polyester film capacitors manufactured by EPCOS. These are the blue blocks visible in figure 4. A custom capacitor was needed to achieve a low-profile for the assembled inverter of just 30mm – available space in the mounting area behind PV modules is very limited. As well as providing the necessary density in a small space, the polyester film type provides the required ripple-current rating.
Figure 4: Custom capacitors enable low profile packaging so that inverters can fit behind solar modules.
Even film capacitors have specification-sheet lifetime rating of around 30,000 hours, which is still some way short of 25 years (even allowing for half of the 25 years being in darkness and therefore non-operating. But whereas an electrolytic offers thousands of hours rated life, with known limitations that make it unlikely to greatly exceed that rating, the film capacitor offers 30,000 hours with characteristics (for example, self-healing dielectrics) that make it very likely that it will continue to operate well beyond that figure.
Inherent stability eliminates opto-couplers
The other problematic component for extended life that often appears in inverter designs is the opto-coupler: ususally, these provide feedback for stability across a galvanic isolation barrier (i.e., a transformer). Opto-couplers suffer known degradation with age – for example, the parameter known as current transfer ratio declines. They typically comprise an LED and a photodiode embedded in polymer matrix, the polymer providing the isolation. The type of extreme environmental cycling experienced by a PV micro inverter is liable to cause a loss of clarity, and therefore optical coupling, in the polymer over time, potentially affecting the performance of the whole circuit.
In Enecsys’ design, the voltage amplification stage provides galvanic isolation, but operates with a stable control loop that only needs the input voltage as a reference: it does not call for a feedback signal to ensure overall stable operation. The Buck-CSI stage is, in any case, tolerant of considerable input voltage variations (to handle the ripple on the DC link), and therefore it also does not need to generate a feedback signal to ensure stability. The opto-coupler is simply not required.
By eliminating these two component types with their problematic long-term performance characteristics – and by implementing a number of other design innovations – Enecsys has been able to achieve something that has defeated many design teams in the industry; a micro inverter design that will give high efficiency, will be manufacturable in volume at a viable price point and that will run with minimal maintenance for the life of the PV module.
Lesley Chisenga is founder and chief architect of Enecsys – www.enecsys.com