They're big but can they get bigger? We uncover research that will help the builders of photovoltaic plants construct tomorrow's solar giants.
In late January this year, Greek Prime Minister George Papandreou unveiled plans to build the world's largest photovoltaic plant, some 200MW in capacity, by 2013. A bold move and, if successful, the plant will be one of several worldwide serving up solar power on the hundred megawatt scale.
But as global plans to build bigger solar plants gather speed, the industry must also find better ways to reliably predict power output. Weather, be it wind speed, cloud cover or local temperature, massively affects the power generated by photovoltaic (PV) panels and research on how large-scale installations function under different conditions is sparse.
One man hoping to change this is Josh Stein, a researcher from the US-based Sandia National Laboratories. He and colleagues are developing systems to provide data on the potential power output and variability of a power plant at any location. As he explains, the team's goal is to 'get to the point where they can predict what's going to happen at larger scale plants as they go towards hundreds of megawatts'.
This would be gold-dust for any utility. According to Stein, a utility wishing to build a large solar PV plant has myriad questions about a proposed site. For example, what happens to power output when cloud covers only part of a large installation, leaving the rest in sunlight? And if wind sweeps soil across an installation, partially obscuring some panels, by how much will power output drop?
Accurate answers to such questions are crucial to predict the power output of future large-scale PV plants and must be resolved if the industry is to build these installations affordably. As Stein highlights, faced with uncertainties, system integrators exercise caution when estimating the possible energy output of future plants and so financiers charge a premium to lend money.
This, though, could soon change. Stein and his colleagues have made an exciting discovery following studies at the ten-acre La Ola Solar Farm on the Hawaiian island of Lana'i.
The 1.2MW plant is the state's largest solar power system and can produce enough power to supply up to 30 per cent of the island's peak demand, making it one of the highest rates of solar PV penetration in the world. This, twinned with the sun and cloud mix at the site, says Stein, has provided an optimum environment for prediction and modelling research.
Researchers first installed a network of 24 wireless, solar radiation sensors. Each sensor measured the solar irradiance – incident solar radiation per unit area – at its location, every second so the researchers could see how atmosphere effects, such as aerosols, as well as cloud cover and wind speed affected the solar energy available.
Stein and colleagues then used these data to determine the average irradiance across the plant, and compared this to the plant's power output. They found a very strong correlation between power plant output and the average irradiance, even at the single second intervals. Stein is excited by the results.
'This is probably one of the first irradiance networks that has been set up with an operational PV plant,' he says. 'The results are somewhat intuitive but when we set out trying to create a hypothesis, there was a lot of difference of opinion on what we would find.'
The next step is to find out whether the relationship between average irradiance and power output holds at larger power plants. With this in mind, the team is currently installing a network of around 50 sensors at an 18MW solar power plant in Alamosa, Colorado.
'If we find the same result, that is, this relationship scales up, I believe we will then have a method for going out and evaluating candidate sites for 100MW, 200MW PV farms,' says Stein. 'We could go out with our mobile sensor network, collect solar irradiation data and estimate output for those plants.'
But while the researchers deliver a much-needed tool for predicting potential plant output, utilities and system integrators still need methods to address other questions. For example, once online, how might the quality of power from a plant interact with the connecting network and how much storage does a plant require, if any at all?
The first step is to sift through existing tools already used by PV suppliers and utilities worldwide, and establish a code of practice for modelling a PV system. 'Each has a different parameter set so there isn't a standard way of modelling a PV system,' explains Stein.
What's more, established suppliers such as US-based First Solar and SunPower have developed proprietary methods but are reluctant to share data. 'Data validation is crucial, but today's data is highly guarded – we can't even access it from operational plants,' continues Stein. 'So financiers label future projects as risky, contributing to the high costs of future plants.'
Clearly the team's latest result in predicting power plant output is a step towards providing standard data and tools industry needs to economically develop and build large solar plants, but what next?
Sandia researchers are now busy collecting data on how PV modules operate under different irradiance and temperature levels, potential wiring losses, how dirty a particular system will be if an operator just allows rain to wash the modules; the list goes on.
'We're trying to get more characteristic data from partners to understand how systems operate,' says Stein, '[The industry] has to have more transparency in terms of how real systems behave and how much power they put out, so financiers lower the risk-rating on these things.'
Sounds like a tall order? Stein agrees but adds: 'We're using the information learned at La Ola and I've just finished another study predicting output for [future] plants as big as 300MW in Nevada. We believe there are no technical barriers to going, even, 100 per cent solar. The barriers are purely financial.'