The Greening of Google
Corporate rooftops are the latest frontier in solar energy generation
It's another brilliant day at the world headquarters of the hottest company on the planet. Some shirtless employees are playing a lunchtime game of volleyball while others stride across campus with laptops tucked under their arms. The place fairly crackles with energy, and in more ways than one.
Up here on a roof at Google's leafy and sprawling Mountain View, Calif., campus, with the shouts of the volleyball game just barely audible, sunlight glints off 9212 polysilicon solar panels stretching out toward the horizon. Amid the irregular jumble of angular roofs, a single south-facing wave stands out, a pitch and roll frozen in place against a backdrop of foothills.
Today, like most days, the panels will generate 9000 kilowatt-hours of electricity before the sun fades into a fat orange ball and disappears into the Pacific. All are connected to Mountain View's section of the electricity grid. The solar modules blanket virtually all the free roof space on the eight buildings at the center of the Googleplex [see " "]. Even part of the parking lot is covered: two rows of carports, shaped like miniature gas stations, support yet more panels. When the last building is fully connected, by the end of this year, the panels will produce 1.6 megawatts of electricity. It'll be enough to satisfy 30 percent of the buildings' peak demand or power a thousand California homes.
Google's project is the largest corporate installation of solar panels in North America. It has grabbed headlines since Google announced it a year ago. That said, it isn't even in the worldwide top 10 of roof-mounted solar projects. A handful of factories in Germany and Japan take that honor, as well as a couple of roofs in Spain and the Netherlands. At the very least, the search giant's solar play adds one more country to the list of star performers in the world of commoditized sunshine. And it seems clear that Google's array won't be tops in North America for long.
After languishing through much of the 1990s, the market for photovoltaic installations in the United States and several other countries took off about five years ago, and it's now increasing by 40 percent annually in the United States alone [see sidebar, "Photovoltaic Hot Spots"]. Spain's bullish market grew 100 percent in the past year. And percentages never tell the full story, as Noah Kaye, a spokesman for the Solar Energy Industries Association (SEIA), points out. ”The German market was relatively flat in the past year, but Germany still installed more [photovoltaics] than the U.S. did,” says Kaye, on behalf of the trade and lobbying group.
California has nonetheless become the second-fastest-growing solar market in the world, and that surge, especially in the United States, is being driven mainly by activity on corporate rooftops. Travis Bradford, president of the nonprofit Prometheus Institute for Sustainable Development, in Cambridge, Mass., calls corporate attention to solar power ”an exploding interest.” In 2006, the commercial sector accounted for 60 percent of newly installed capacity in the United States, up from 13.5 percent in 2001, according to data from the U.S. Department of Energy.
”We've stopped reporting the biggest systems,” Bradford adds. ”A new record is set every few months.”
In March, Applied Materials of Santa Clara, Calif., announced a plan to install 1.9 MW of solar power on the rooftops of its Sunnyvale, Calif., complex. And it's not just high-tech titans retooling their roofs: Tesco, the British-based supermarket chain, says it intends to put up a 2-MW solar installation at an office complex in northern California. Wal-Mart, the world's largest retailer, intends to outshine all these companies with multipart plans to put more than 5.6 MW's worth of solar panels on the roofs of 22 stores in California and Hawaii. Two other discount-retailing giants, Target and Kohl's, have also begun transforming their roofs into tiny, independent utilities.
”It's not an illusion,” says Craig Cornelius, program manager for the Department of Energy's solar division. ”Corporate solar is really happening.”
Amid the enthusiasm, it's important to keep this latest twist in the solar saga in perspective. Solar energy of all kinds fulfills less than 0.1 percent of electrical demand in the United States, and affordable, commercially available panels have hovered near 15 percent efficiency for years. Despite the recent burst of corporate enthusiasm, the prices of solar modules are expected to continue inching down at just 5 percent a year, and grid parity--the point at which solar panels can compete subsidy-free with utilities--isn't expected until 2015 at the earliest.
It's too soon to say whether these costly corporate installations will go down in history as the first of a limited series of impulsive, feel-good publicity moves by tech start-up billionaires, or as the beginning of a longer-term movement that will help sustain the market for solar photovoltaics during the next decade and enable solar to finally become cost-competitive. One thing is certain: the movement will flourish only to the extent it is nurtured by a complex patchwork of economic and bureaucratic conditions. Of the 9509 new grid-tied solar installations in the United States in 2006, which totaled 101 MW, 70 percent of them were in California. And that's not just because it's sunny. As it turns out, California subsidizes solar in a particularly generous way.
Even without subsidies , solar panels may have found their logical home, at last, in the commercial world. The nice, flat roof design on most commercial buildings, unlike the pleasingly angular but less workable residential roof, is one obvious advantage. But some of the most compelling reasons are intangible.
”It's a key part of attracting and retaining employees,” says Doreen Reid, a senior associate at The Climate Group, a London-based nonprofit that helps companies reduce their carbon emissions. ”Students coming out of college are more conscious of a company's environmental image.”
Robyn Beavers, Google's corporate environmental programs manager, confirms the transformative power of solar cells. ”I've had so many people e-mail me and say, ’This is why I love working at Google' and, ’How can I install solar at home?' ” says Beavers, who presided over the installation project. Google's solar enterprise is part of a larger mission to promote the growth of solar energy, she says. Google founders Sergey Brin and Larry Page have invested heavily in Nanosolar, a start-up that specializes in thin-film solar cells. (Both declined to be interviewed for this article.)
For its rooftops, Google chose Sharp modules capable of generating 208 watts each. The polycrystalline silicon cells average 12.8 percent module conversion efficiency. Because solar panels produce dc current, each system requires inverters to change the current into usable ac, and Google used a set of utility-grade inverters with an average of 96 percent conversion efficiency made by SatCon Technology Corp. of Boston. It partnered with EI Solutions, a solar project developer with headquarters in San Rafael, Calif., to do the electrical design work.
Google won't say how much the whole project costs, other than to indicate that it expects to recoup its investment in five to seven years. Nonetheless, experts estimate that a solar installation costs between US $3 and $5 per watt in California, and between $6 and $10 per watt in the rest of the United States after factoring in local and federal rebates for the cost of the system. (According to the Northern California Solar Energy Association, the average cost of installing large systems in the Bay Area in 2006 was $8.58 per watt before rebates, on par with national figures.) Using data from California's Solar Initiative program and based on a $2.80-per-watt incentive rate, Google likely retrieved about $4.5 million from California on a project that in total probably cost more than $13 million. Federal tax breaks through the Energy Policy Act of 2005 also help to burnish the appeal of what is still, for many, a prohibitively expensive system.
For other companies, an important piece has been added to the picture for solar. Where research and development have so far failed to slash the price of solar, clever financing schemes have filled the breach. Google's solar project, for all its trendy impact, was financed the old-fashioned way--with cash. But customers without billions of dollars in liquid reserves tend to shy away from such a move.
Rather than requiring that customers buy all the equipment for an installation, which can run into the millions of dollars, solar service providers are persuading customers to sign agreements that in effect turn those providers into miniature utilities.
The office-supply company Staples was among the first to pursue such a scheme--in 2004 with SunEdison, a prominent Beltsville, Md., solar electricity service provider, for a 280-kW installation on two of its California warehouses. The solar installation covered about 10 percent of the facilities' electric loads.
In this arrangement, SunEdison installs the solar modules on a customer's property and is responsible for maintaining them. But, crucially, it does not charge the customer for them. Instead, the customer signs a long-term agreement, usually lasting about 15 years, that locks the customer into buying back the electricity generated by those panels at a fixed rate. Typically, that rate is lower than retail utility prices. Prometheus's Bradford estimates that 40 percent of recent commercial installations have gone this route, and he says that it's likely to grow more popular as additional companies move into solar.
Through such long-term contracts, corporations are cushioning themselves from fluctuations in electricity prices, over both the long and short terms. In the short term, on-site generation lightens the customer's demand from the local utility precisely when the utility needs a break: during peak periods of demand, which are also, not coincidentally, the hours of the typical business day--yet another reason that corporate roofs make sense as hubs of solar activity. During these peaks, electricity prices can double, triple, or even quadruple. ”Solar energy is generated when companies need it most, which is typically 10 a.m. to 5 p.m. on hot, sunny days,” says Kaye, the SEIA spokesman.
Add to that executive-level nervousness about the prospect of restrictions on carbon emissions and the volatility of electricity prices, and some businesses are finding the case for solar compelling. ”Companies can do all of the above,” says Rick Whisman, the director of west region system sales at PowerLight, in Berkeley, Calif., a subsidiary of SunPower, one of the largest solar cell producers and installers in the United States. ”They can make a wise decision on energy over the long term. They can reduce their footprint and also be prepared for regulations that may come into play in the future.”
Whisman and others are quick to point out that solar makes sense only as a component of a larger plan. ”What has made Google and Wal-Mart so noteworthy is the degree of thought that went into their planning,” says Bradford. Indeed, pursuing electricity generation on-site is of limited value unless accompanied by a suite of energy-efficiency measures to reduce a company's overall demand [see ”The Zero-Zero Hero,” IEEE Spectrum, September]. Wal-Mart is the undisputed leader in driving the adoption of compact fluorescent lightbulbs, and the megaretailer hasn't stopped there. It is also modernizing its truck fleet to be more aerodynamic and fuel-efficient, and at a store in Texas, the company is testing sustainable design with an experiment that includes rooftop wind turbines and on-site recycling. Google, meanwhile, is pushing ahead with a broad package of ambitious environmental programs [see sidebar, ”Google Goes Green”]. ”This is just a first step for us,” Beavers says.
Will the next steps take Google beyond California? Beavers won't say. Google has 15 U.S. offices and several power-hungry data centers outside of California, as well as offices and facilities in 23 other countries. But few, if any, of those places offer the incentives California does.
Driven by those incentives, in the past five years Californians have edged above the threshold of 30-kW demonstration projects to larger systems, such as Google's, spawning a cottage industry of experienced local solar-installation service companies. The California Solar Initiative credits companies based on performance metrics that can amount to one-fourth the cost of the system, which--when combined with a federal tax credit on some solar equipment, and depending on the cost of the panels and the installation--can cover more than 50 percent of a system's total cost. In 2007, rebates in California evolved from being per-watt, based on system size, to a formula that takes into account details of the physical placement of the panels, so that systems that are expected to perform better will be reimbursed more generously. The new calculations factor in the panels' tilt and shading, as well as altitude and azimuth, which are the two coordinates commonly used to describe the sun's apparent position in the sky. The rebates are still part of a tiered system designed to reduce the incentives over time, and in 2008 energy-efficiency requirements will be tied to those rebate dollars. Only New Jersey, among the other 49 states, has shown a similar level of leadership in photovoltaics.
The question now is whether the movement can expand beyond a few isolated states and countries. Data from the Interstate Renewable Energy Council, a nonprofit that disseminates information on rules and incentives relating to renewables, suggest that Arizona, Colorado, Massachusetts, New York, and Texas are also promising markets. New Jersey, with its second-only-to-California inducements, has the second-highest installed solar capacity, with 18 MW in 2006. By comparison, Florida, the Sunshine State, installed a meager 170 kW of solar energy in 2006, the year that its solar incentives program was launched. Its limited generation capacity speaks to the paltry nature of those incentives.
That disconnect between sunshine and solar output is even more pronounced outside the United States. The global leaders in solar energy, by virtually all metrics, are Germany and Japan. Both countries have sky-high electricity prices: on average 20 cents per kilowatt-hour in both Germany and Japan, double the average price of electricity in the United States, according to 2006 data from the International Energy Agency, in Paris. Starting in the mid-1990s, both governments began pouring money into renewable energy programs. As a result, today, in cloudy Germany, the renewable energy industry has become the country's second largest source of new jobs after the automotive sector. It employs some 200 000 people, according to Paul Runci, a senior scientist at Pacific Northwest National Laboratory, in Richland, Wash., who studies energy research and development trends.
But for corporations, the solar story depends on more factors than rates and rebates. In the United States, state-by-state rules on how to attach solar plants to the grid and how to compensate producers for electricity they export to the grid vary tremendously. The portion of the Google system that has the solar panels, for example, connects to a secondary grid, which does not accommodate excess power fed back into it, according to Johann Niehaus, the lead engineer on the project for EI Solutions. To account for that, the system was scaled both to fit the available roof space and to generate less than 50 percent of the buildings' minimum demand, so that the solar modules never come close to producing more electricity than the campus consumes. What they've installed approaches that 50 percent limit.
That is just one way that interconnection to the grid can be complicated. Other problems stem from what numerous experts have described as electric power companies' lack of familiarity with distributed generation. Some utilities have been reluctant to open up their grids to ever-larger quantities of electricity that they cannot manage. At times, a utility may declare that a generation system warrants an engineering study, which can cost up to $50 000, to analyze the impact of adding the system's electricity to the grid. That may not be prohibitive for Google, but for some prospective buyers it is. ”They want to know, are you creating frequency disturbances? Voltage disturbances? How big are you in relation to the peak load on that circuit?” says Christopher Cook, a senior vice president for regulatory affairs and new markets at SunEdison. It's not unusual for a solar project to be killed because of the expense of commissioning an impact study. That's why, in one recent case described by Cook, a school in Virginia abandoned its plan to put solar panels on its roofs.
Some solar watchers have argued that those studies are sometimes unnecessary and redundant. The Interstate Renewable Energy Council and the Department of Energy, among others, are calling for guidelines to specify when they are needed.
”There are legitimate grid reliability concerns that can be addressed through technical means,” Cornelius, of the Department of Energy, says. ”And then there's the other reasons.” Wal-Mart initially had hinted at a solar plan for its stores that would add up to 100 MW, but SEIA's Kaye suggested that the company had found it unfeasible in many states because of slow response rates from utilities. Such bureaucratic bunglings related to connecting to the grid led Wal-Mart to scale down the project to stores in just a few states, at least for now.
Cornelius points to Connecticut as an example of how things can go wrong for solar. Electricity prices there are among the highest in the country--15 cents per kilowatt-hour for the commercial sector in 2007, compared with a national average of 9 cents--and the state is full of congested distribution systems. Nevertheless, interconnection problems have not yet been formally ironed out. These have ranged from the painfully mundane, such as utilities not processing applications quickly, to unresolved concerns for the safety of a utility's distribution engineers. As a result, project developers are not prepared to invest in installations until they are confident the utility will agree to connect them hassle-free. ”From the industry's perspective, we look at a state, and if it doesn't have interconnection rules, we say we can't do this project,” Cook says. ”It's not: let's go forward and see what happens. It's just simply: this state's not open for business.”
All these bureaucratic problems are surmountable, analysts say, largely by means of new industry standards. Approved in June 2003, IEEE Standard 1547 represents one step toward integrating the technical side of interconnection practices for distributed power sources. ”We had something like 3000 utilities, each with their own interconnection requirements,” says Richard DeBlasio, the technology manager for the National Renewable Energy Laboratory's Distributed Energy and Electric Reliability program, in Golden, Colo., who chaired the committee that drafted the standard. Since then, at least half of the 50 states have adopted the IEEE standard as well as another, UL 1741, as their minimum technical guidelines for equipment and safety requirements. If the political will is there, the environment should improve for photovoltaic installations. In fact, it's already happening: in June, Oregon's legislature approved a set of policies that can amount to a 50 percent tax credit for solar installations and manufacturing. And the historic materials crunch that has held the prices of panels aloft is likely to abate as new manufacturing capacity comes online starting in 2008.
Although California may be alone in the United States in swaddling itself in polysilicon panels, the leadership of the Googles and Wal-Marts of the world could cause corporate solar installations to pop up on rooftops in the rest of the country in almost no time at all.
Nvidia's per-vector scaling quantization (VSQ) scheme better represents the numbers needed in machine learning than standard formats such as INT4. Nvidia
Posits differ from floating-point numbers by the addition of an extra variable-length regime section. This gives posits a better accuracy around zero, where the majority of numbers used in neural networks reside.
Complutense University of Madrid/IEEE
Three 16-bit Brain Float numbers can carry the same precision as 32-bit floats while saving chip area during computations. John Osorio/Barcelona Supercomputing Center
