Distributed Inverter Design

Utilizing String Inverters in Large Commercial Systems

For the past decade, designers and installers working in the North American PV industry have welcomed the challenges presented by larger, more complex systems. Inverter manufacturers also embraced the challenge and began to offer high-capacity string and central inverters to meet the design and power-conditioning demands of these larger solar projects. My company, Sunlight Electric, completed its first PV installation in 2002. The project utilized a single SMA America Sunny Boy 2,500 W inverter. As system sizes grew, we installed multiple string inverters coupled with an ac aggregation panel to meet a given system’s capacity requirements. In 2004 we installed our first project that utilized a central inverter. It was an important milestone for the company, and we had a sense of accomplishment in knowing that we had graduated from “beginner” to “expert” inverters. However, in recent years we have taken a step back and have been developing large commercial systems based on distributed designs that utilize multiple string inverters.

Today, for projects between 100 kW and 1 MW, designers can specify one or more central inverters, or they can create a distributed design that utilizes numerous wall-mountable string inverters. In the abstract, designing a 1 MW PV system consisting of 400 2.5-kW string inverters does not seem all that far-fetched. Intuitively we appreciate that there are major benefits to distributed designs in terms of fault tolerance and performance. After all, the Tesla Roadster battery pack consists of approximately 7,000 individual cells, and Google employs nearly 2 million servers across 12 data centers around the world.

Despite this intuitive understanding, if you asked PV system designers to consider using 400 2.5-kW inverters on a 1 MW PV system, it would probably be difficult to engage in a serious discussion of the relative merits of such a design. The foregone conclusion is that this approach would not be cost effective due to greater inverter, labor and BOS costs. But consider two 500 kW inverters versus a single 1 MW inverter. Savvy designers and engineers recognize the benefit of the increased fault tolerance. After all, the chances of both inverters failing at the same time would be considerably less than the chance of a single inverter failure. But what about specifying four 250-kW inverters, 10 100-kW inverters or even 40 25-kW inverters? This is where the discussion starts to get interesting.

Despite tendencies in the US toward utilizing central inverters for commercial- and small utility-scale projects, Henry Dziuba, the president and general manager of SMA America, points out that “distributed inverter designs have long been a best practice in Europe.” New string inverter products allow integrators to deliver robust and cost-effective distributed systems for their customers. In this article, I present the potential benefits of the distributed inverter design approach and discuss engineering and installation best practices for integrating these systems based on the evolution of our design approach at Sunlight Electric.

Benefits of a Distributed Inverter Design

For commercial-scale projects, we began using central inverters at the earliest possible opportunity. In 2004 we specified our first central inverter: a Xantrex PV225 (225 kW) model for a 179 kWdc project at Frog’s Leap Winery in Napa, California. We did not even consider utilizing multiple string inverters, and we ruled out using two 75 kW central inverters because the cost of the single 225 kW inverter was actually lower.

A few years later, when we were designing a 125 kWdc ground-mount project for ZD Wines in Napa, we added the distributed approach to our list of options. We considered specifying a Xantrex PV225 central inverter, but the cost penalty of purchasing an additional 100 kW of capacity would have made the economics prohibitive. By then, SMA America had introduced the Sunny Boy 6000-US, and we realized that the distributed approach allowed a better matching of the PV array and inverter capacities. In this case, the string inverter option was also the most economical solution. With a higher weighted efficiency and a lower cost per watt, the 6 kW string inverter also won out over the 30 kW and 40 kW central inverter models that were available at the time. After we ruled out using a single large central inverter and multiple smaller central inverters, the best option appeared to be specifying 18 string inverters. As part of our typical stakeholder sign-off process, I contacted SMA’s technical support staff to discuss integrating 18 SB 6000-US inverters and learned there were even more compelling benefits associated with a distributed inverter design.

Optimized inverter-to-array ratio. Central inverters have larger power-capacity increments than string inverters do, making it more difficult to optimize the ratio between the inverter and array capacities. The distributed approach gives designers more flexibility to optimize the ratio between the inverter’s dc input capacity and the size of the PV array, and can reduce a project’s total inverter cost and improve the owner’s ROI.

Decreased dc BOS cost. Using string inverters with integrated combiner boxes eliminates the need for separate dc source-circuit combiners. In comparison, the higher-capacity dc inputs on a central inverter are designed to accept PV output circuits that are aggregated in separate dc combiner boxes, which drives up conductor sizes and associated costs.

Cost-effective granular monitoring. Using multiple string inverters in a distributed design increases the number of inverter-direct monitoring points. A centralized design requires zone- or string-level monitoring to provide a similar level of system performance visibility. In many systems, utilizing inverter-direct monitoring yields sufficient monitoring granularity at a lower cost.

Reduced space and infrastructure requirements. String inverters can be mounted on an existing wall, carport column or racking structure. Central inverters located in outdoor environments require an inverter pad and often an associated shade structure, consuming potentially valuable space and adding cost that does not necessarily add value for the customer.

Maximizing uptime. In the event of an inverter failure, a distributed design increases the PV system’s ability to continue operating near its intended production levels. If one inverter fails, the loss is equal to the capacity of the inverter. For example, in the ZD Wines project, a single inverter failure would result in a loss of just 5.5% of the total system capacity. Alternatively, if the system utilized a single central inverter, a failure would result in a 100% loss of the system’s production.

Easy replacement and improved uptime. Two technicians can remove and replace a wall-mounted string inverter in a matter of hours. The low cost per unit for a string inverter enables system owners to keep a spare in stock or on-site, so just one truck roll can restore operation. Alternatively, repairing or replacing a central inverter typically requires the inverter manufacturer to deploy a specialized technician and heavy equipment to remove and replace the unit if necessary. This may take days or even weeks to schedule and complete.

Design flexibility. Where necessary, installers can strategically deploy string inverters to maximize array production that varying array conditions may otherwise limit. The distributed approach provides increased design flexibility to address adverse site conditions, such as localized shading and variable array orientations and pitches, because each string inverter employs one or more MPPT channels.

After getting SMA’s input on the benefits of distributed inverter design for large commercial projects, it seemed too good to be true. The only downside was increased labor costs to install the individual inverters and the added ac aggregation panel or panels required to combine the inverter outputs before tying into the site’s main service. Having conducted our analysis on potential designs for the ZD Wines system, we determined that the savings and benefits of a distributed design more than offset the added costs. Since I had not seen anyone in California deploy such a system, I asked the technical support representative at SMA if anyone had done anything like this, and I can still recall his reply: “We do this in Europe all the time! Americans just love their big iron.”

Wall-Mountable String Inverters for 3-Phase Applications

Recent developments are reframing the distributed versus central inverter discussion. In 2007, SMA America introduced its Sunny Tower system. At the time, it integrated six 6-kW or 7-kW inverters into a pre-engineered unit designed for 3-phase applications. The product allowed integrators to work with familiar equipment, and it eased the transition to working with larger 3-phase systems. Since then, several manufacturers have introduced string inverters developed for 3-phase applications, ranging in rated capacity from 9 kW to 30 kW. The lightweight units mount directly on a wall, rack or carport column and, depending on the product, interconnect to 3-phase distribution systems at 208, 277 or 480 Vac. They offer inverter-direct monitoring, and some models, such as the Aurora Trio units from Power-One, are listed for 1,000 Vdc maximum input voltages. This higher dc voltage rating increases the possible number of modules per source circuit and offers further reductions in the number of dc components and installation labor hours.

Currently, eight manufacturers offer high-capacity wall-mountable string inverters developed specifically for 3-phase applications in North America: AE Solar Energy, Chint Power Systems, Fronius USA, Ideal Power Converters, KACO new energy, Power-One, SMA America and SolarEdge Technologies. Table 1 (p. 30) lists the relevant models that are available in the US or will be launched in the near future. Comprehensive specifications for these inverters, as well as string inverters developed for single-phase applications, are included in the “2013 String Inverters: Developments and Specifications” article (pp. 42–52). This dataset also identifies lower-capacity string inverters with 208 Vac outputs that can be effectively deployed in 3-phase applications. These new and improved string inverter offerings strengthen the case for considering the distributed design approach for commercial projects as large as 1 MW or even larger.

Whether new products are filling a market niche and enabling new designs, or the market is driving manufacturers to meet the needs of their customers, the use of distributed designs in commercial and small-scale utility projects is catching on. “We are seeing great response from the marketplace on our line of 12 kW–24 kW string inverters for larger commercial projects,” says Mike Dooley, the vice president of marketing for AE Solar Energy. “Our customers are starting to understand the benefits of distributed string inverter designs.” Dziuba of SMA agrees: “As new products have become available in the US market, string inverter designs are increasing in popularity. With many projects, this concept provides compelling advantages, including increased design flexibility, simpler O&M and a superior financial outcome.”

Several inverter manufacturers currently offer string inverters developed for 3-phase applications as well as higher-capacity central inverters (see Table 1). A related development is Advanced Energy’s acquisition of REFUsol in May 2013. Steve Reed, the string inverter product manager at AE Solar Energy (Advanced Energy’s solar division), states, “The foremost factor in Advanced Energy’s decision to acquire REFUsol was the added flexibility and choice that we can offer our customers.” The REFUsol product line offers “a compelling price-to-performance ratio, ease of installation, improved uptime, and quick serviceability for commercial applications where flexibility and modular design are essential,” Reed says.

Industry Opinions on Distributed Inverter Design

Although a distributed design is not always the most cost-effective option, many manufacturers and installers point to benefits of the approach. Ryan Parsons, the national sales manager of utility markets at Power-One, a manufacturer of both string and central inverters, states, “Our analysis indicates that string inverters make sense in a lot more applications than many integrators assume due to more MPPT points, greater design flexibility, and lower installation and BOS costs.”

Dan Wishnick, a manager of business development at Siemens Industry, which provides PPA funding for the US municipal, educational and health care markets, states: “In our experience, the sweet spot for the distributed string inverter design is just about anything under 10 MW. We find the costs to be comparable or lower for the distributed approach.”

Reed from AE Solar Energy adds: “We see the string inverter product used predominantly for commercial and small utility-scale projects. It works for both roof- or ground-mount applications, but lends itself especially well to carports and multiple azimuth and angled applications.” Reed states that the direct benefits include “lower total cost, primarily driven by lower BOS costs” and “higher and more consistent energy production in systems with adverse conditions, such as shading, soiling, arrays with multiple orientations and space-constrained projects.” He points out, “For larger systems, this cost reduction is lessened, and other factors, such as LCOE and site labor, may become more important.”

Jim Curran Sr., the president of Shamrock Renewable Energy Services, a solar integrator located in Northern California, notes, “Our analysis of BOS and installation labor costs of distributed inverter designs indicates an average savings of $.05 per watt-dc, or approximately 3%.” So even if string inverters cost a few cents per watt more than central inverters, the decrease in other costs often offsets the additional cost of specifying string inverters. This allows integrators to assess and compare design approaches on other factors, such as the benefits of inverter-direct monitoring, design flexibility, and increased fault tolerance and system availability.

A Distributed Inverter Design Case Study

Distributed inverter systems are now ingrained in our design process at Sunlight Electric. They are our default solution for commercial systems with capacities of up to 250 kWdc. We have deployed systems as large as 855 kWdc using this approach and have found the benefits to be compelling in increasingly larger systems. One of our recent proposals is a 743 kWdc system that interconnects to a 480 Vac 3-phase commercial electrical service. Traditionally, we would consider the project an excellent fit for two AE Solar Energy 333NX (formerly Solaron 333) 3-phase 480 Vac central inverters. We analyzed using 26 AE Solar Energy AE 3TL REFUsol 024K-UL string inverters, each with 12 source circuits of 10 modules per string. In the past, our inverter selection process for a project like this was relatively simple. It included reviewing the inverter cost per watt, the possible source-circuit and dc-aggregation configurations, and a few other minor considerations. A distributed inverter design requires deeper analysis.

Inverter costs. With matching warranties applied, the equipment costs for the central and string inverter designs were comparable. The cost for 27 string inverters (26 to deploy and a spare to keep on hand) was within 1% of the cost of the central inverter options we analyzed.

Labor and BOS costs. There would be increased labor and equipment costs associated with installing 26 string inverters and the required inverter ac aggregation panels. Would we simply be moving cost from the dc to the ac side of the system? The string inverter’s integrated dc source-circuit combiners eliminated the need for separate combiner boxes and reduced the overall costs for the required dc BOS components. The distributed design also avoided material and labor costs for the concrete pad required to support 4,000 pounds of inverters and the associated shade structure. Working with Shamrock Renewable Energy Services, our installation partner, and its subcontractor, IE Systems, we determined that the price premium for labor and BOS costs for the distributed design would be approximately $27,000, or about 1% of the total project cost.

Inverter performance. In this case, there were only modest differences in the CEC-weighted efficiency of the two different options—97.5% for the central inverters compared to 98% for the string inverters. This application would have a uniform array tilt and orientation. As such, there was no significant production benefit to having 26 individual string inverter MPPT channels compared to two channels in the dual central inverter design.

Space utilization. The prospective client had plenty of space for two ground-mounted inverters and the associated pad, so this was not a major design driver.

Granular monitoring. In this system, inverter-direct monitoring was the lowest-cost option and provided a sufficient level of granularity. It also provided the buyer with greater confidence that a technician could quickly identify, locate and address any underperformance issues. In comparison, the central inverter approach required zone-level monitoring at a minimum, or smart combiners for string-level monitoring, to achieve an acceptable level of dc-side monitoring granularity.

Maximizing uptime. The benefits of a more fault-tolerant design and improved uptime were relatively easy to quantify. To demonstrate these benefits to the client, we presented a worst-case inverter failure scenario for each design approach. In the case of the central inverter design, should one of two inverters fail on a Saturday morning, in all likelihood we would not be able to dispatch a service technician until Tuesday. If we could not repair the unit ourselves, we would call the manufacturer that Tuesday to open a support case and schedule one of its field technicians to visit the site as soon as possible. With luck, the technician would arrive before the end of the first week. If we could not repair the inverter within the first week, we would be looking at early the following week for restored operation. In this admittedly worst-case scenario, half of the system could be down for as many as 10 days.

If this failure were to occur in June, the customer would have to make unexpected utility purchases exceeding 19,000 kWh. Under PG&E’s Small General Time-of-Use Service rate plan—which the client selected to capture premium energy rates during peak PV generation hours—the effective rate for 19,000 kWh would be $0.286/kWh, for a total loss of $5,434. Based on the California performance-based incentive rates during that time, the lost PV generation would add another $475 in losses. Thus, the net loss during a single central inverter failure could be as high as $5,909 over the 10-day period.

Our experience indicates that these worst-case failures do happen. Joshua Weiner, the president of SepiSolar, a specialized solar engineering firm, advises PV system designers to request mean time between failure (MTBF) rates for the inverters they are considering, in writing, from primary providers or independent sources like Black & Veatch Bankability Reports. Weiner says, “If you can get MTBF data, the calculus becomes MTBF rates multiplied by the number of units considered for both central and distributed options.” Although not always readily available, inverter MTBF data are quite valuable for estimating inverter failures over the life of the system. A designer can perform a detailed reliability analysis that accounts for the higher number of inverters deployed in a distributed design.

The loss would be considerably smaller should one of the 26 string inverters fail in the distributed design. If that failure were to happen on a Saturday morning in June, in all likelihood our technician would not be on-site until Tuesday. If we could not repair the unit, we would simply swap it out with the spare string inverter already on-site. The total value of lost production would be $600 for the customer, and we would complete the operation in a single truck roll. In the meantime, we could send the failed unit back to the manufacturer for a replacement, so a new spare inverter would be on-site.

The fault-tolerant distributed approach and its improved uptime minimize the potential financial losses due to an inverter failure. In either scenario, assuming that one inverter may fail over the life of the system, the financial losses for the customer could be $5,909 or $600—a difference of $5,309—depending on the design approach. This begs the question: Is it worth paying an additional $27,000 for the distributed design? At first glance, it does not seem so. However, since the system expense qualifies for the 30% federal investment tax credit and will depreciate at a combined state and federal marginal tax rate of 40%, the true after-tax premium for the distributed design is approximately $8,100.

To spend an additional $8,100 to avoid a loss of $5,309 due to a worst-case scenario comes closer to making sense. Since each of the proposed options includes a 20-year inverter warranty, the added cost would more than break even if this worst-case failure were to happen twice in 20 years, which is certainly within the realm of possibility.

In this case, the system owner and operator recognized the benefit of granular inverter-direct monitoring and the distributed design’s potential to minimize revenue losses in the event of an inverter failure. With this deeper understanding, the prospective customer found that the distributed approach provided sufficient peace of mind to justify the added up-front expense. Our comparison not only proved valuable for our potential client, but also validated our opinion that distributed designs are worth considering for large-scale commercial applications. The introduction of higher-capacity 3-phase string inverters listed for 1,000 Vdc applications will make the value proposition even more compelling.

CONTACT

Rob Erlichman / Sunlight Electric / Sonoma, CA / sunlightelectric.com

MANUFACTURERS

AE Solar Energy / 877.312.3832 / advanced-energy.com

Chint Power Systems / 855.584.7168 / chintpower.com/na

Fronius USA / 877.376.6487 / fronius.com

Ideal Power Converters / 512.264.1542 / idealpowerconverters.com

KACO new energy / 415.931.2046 / kaco-newenergy.com

Power-One / 805.987.8741 / power-one.com

SMA America / 916.625.0870 / sma-america.com

SolarEdge Technologies / 877.360.5292 / solaredge.com

Article Discussion