Boulder City, Nevada, is no stranger to renewable energy projects. More than 70 years ago, many residents were employed in the construction of the iconic Hoover Dam. With a generating capacity of over two gigawatts (GW), the plant supplies 15 near-by cities including Boulder City itself and Los Angeles across the state border. So it was a return to form of sorts, when construction began in 2010 on the vast Copper Mountain Solar 1 plant. Worth 48 megawatts (MW) the plant, which is owned and operated by Sempra Generation, is a minnow in comparison with its hydropower neighbor, but is still of impressive scale. Also providing cross-border electricity to California, on completion it is the biggest PV plant in the U.S. Covering 380 acres (1.58 square-kilometers) and using 775,000 PV panels, the electricity it produces is sold to the Pacific Gas and Electric Company under a 20-year contract. It really is a vast piece of land, describes University of California, San Diego (UCSD) researcher Jan Kleissl.
Copper Mountain is also set to get bigger. A second stage, Copper Mountain 2, will expand the installation by 150 MW. Being built in two stages, Sempra expects the initial 92 MW to be completed by 2013 with the final 58 MW by 2015. With such a large installation, a passing storm, weather front or even high-level clouds can significantly affect the electrical output of the plant.
These fluctuations can cause utilities and grid operators, such as California ISO, major headaches and often underpins arguments as to PVs unreliability as mainstream source of electricity. James Blatchford from California ISO says uncertainties caused by clouds in the short-term are significant: These [
] can have a profound effect not only on the cost of energy replacements but also on the reliability of the system. Additionally, if electricity pricing systems which are designed to ensure greater efficiency within the renewable energy supply market are put in place, such variation could cost Sempra big money in the form of penalties for not delivering on forecasts. Under variable or dynamic pricing systems, the market and prices received for renewable energy fluctuates with accurate production forecasts incentivized and inaccuracies being penalized. The time frames for such fluctuations can be as little as five minutes. Penalties for inaccurate forecasts may initially seem onerous or punitive in regards to the PV generators, but on closer inspection they make a lot of sense, argues UCSDs Kleissl.
Expensive energy
The more uncertainties [in supply] the more reserve power they [utilities] have to procure in advance and that is expensive energy and if its not used then it was a waste. But if its not enough, then the lights may go out! Kleissl continues that a power purchasing system designed with these mechanisms has been put in place in PV markets in Spain, where in order to receive a premium rate for the electricity generated, operators must make accurate predictions on their output. In the Mid-west of the U.S., reports Kleissl, similar short-term supply markets for wind-generated electricity are being established.
Forecasts lead to cost reductions
The Department of Energys (DOE) SunShot Program and the National Renewable Energy Laboratory (NREL) are funding acquisition and deployment of the instrumentation for Kleissls research. The SunShot Program Manager Ramamoorthy Ramesh explains that short-term PV production forecasts can allow utilities to quickly compensate for fluctuations in production. If you are a utility company and you know that a cloud is going to pass over a big solar installation and you have five minutes notice [
] you have to fire up a gas turbine engine, which will boost up the electricity that you produce. If the time frame is reduced, then expensive reserve energy is required.
The SunShot programs goal is to drive down the cost of PV by 75 percent. Last month it announced that it has financed USD145 million in projects in pursuit of this goal. While the Copper Mountain cloud forecasting project has the possibility of driving up the prices PV electricity producers can charge utilities in the future, it could make these plants more profitable and attractive to investors and potentially drive up the size of installations.
The DOEs Sandia National Laboratories pursue a range of PV research in the name of energy security, however are not involved in this specific forecasting project. Sandia researcher Joshua Stein has researched forecasting methods and believes the accurate short-term forecasting may have additional benefits. If you can accurately predict in the short-term, it may make it easier to integrate other technologies such as energy storage systems to help mitigate some of the large ramp rates that are concerning some utilities.
Forecasting today
At present, forecasts rely primarily on satellite information. However, with large parks and for short-term forecasts, satellite data is deficient. For larger parks, SunShots Ramesh explains: Imagine you had a million solar panels, a huge installation, lets say 200 MW. Where the cloud is walking over the left-most solar panel, the right-most solar panel is still generating electricity. To be able to predict such a large plants output, spatiotemporal data for different parts of the installation needs to be collected and processed. Satellite images lack sufficient resolution to do this with many PV parks appearing as only as one pixel. UCSDs Kleissl says: Satellites give you [ ] one value of cloudiness for the entire power plant, so you cannot really map out where exactly the clouds are.
Heaven and earth
The approach being developed by Kleissl is seen as complementary to satellite techniques. It transitions seamlessly into our technology, satellites take it from above, we take it from below. Satellites can see further out, we can see more detail nearby. The other complimentary technology is point-sensors, which give accurate calculations of output, but only the one stationary reading, from where the sensor itself is located.
Project kit
In contract, Total Sky Imagers (TSIs) have been placed at different ends of the Copper Mountain plant, taking images of the sky every 30 seconds. Yankee Environmental Systems (YES) sell these imagers and co-founder Mark Beaubien describes the TSI: It is essentially an industrial quality environmentally hardened all-sky web imager. Its [
] focused on the cloud base, the atmospheric boundary layer, where the clouds start. Then the other part of the TSI is an image processing computer which takes each image and does a sophisticated parametric model of the sky. The TSI has two further components, a heated mirror onto which the image is reflected and a rotating shadow band to ensure glare from the sun doesnt detract from the image quality.
Kleissls research is not the first to use TSIs, but it is the first time they have featured on a PV installation of this size. NREL has been working with the TSI at its Solar Radiation Research Laboratory (SRRL) in Golden, Colorado. NREL will draw on this experience and plans to remain involved in the project through to the forecasting phase.
While Kleissl doesnt wax lyrical about the TSIs themselves thats old technology, no different to a camera he says that the next stage is more complicated. Processing the images [
] we locate the clouds very easy to do with your own eyes, for a computer it is a challenge to a computer dark clouds may actually look similar to blue sky. The UCSD team then tries to analyze the velocity and direction of the clouds. This is done by overlaying two different sky images, in various ways, determines whether the clouds have moved by a certain number of pixels, corresponding to a distance traveled.
In terms of funding the research, through SunShot, NREL put in USD126,00 for the acquisition and deployment of the TSIs. The decision to commence the project was made in July 2010, when California ISO observed Kleissls use of TSIs and the ability to improve forecasting. The ISO coordinated the efforts between UCSD, Sempra Energy and the Copper Mountain Facility for the installation and data exchange, says the ISOs James Blatchford. Installation of the TSIs was completed in July 2011. The Californian Energy Commission is providing USD534,000 to develop sky imager solar forecasts and integrate them with satellite and weather forecasts over the next three years.
Forecasting
After the collection and sorting of the data, to make actual PV output forecasts is then the next step. Using the most recent image as a reference, the speed and direction of the clouds is applied. Assuming that this doesnt change within the forecast period, the final step is to project the shadows of the cloud onto the solar park below. In this, the position of the sun is a crucial factor. Depending on the height of the cloud and time of the day, [the shadow] could be many kilometers off from where the clouds actually are over the site, says Kleissl.
There are clearly a lot of moving parts and techniques to be refined in the various stages of the project. pv magazine can report that the first images from the project have been sent to California ISO; however, the real goal of the research project is to take what Kleissl describes as, pretty pictures and provide some accurate forecasts for the power plant. That would probably take another two or three months to get it on an operational framework. And then the remaining two years [of the project] will mainly look at integrating different forecast products.
Challenges
While the UCSD researchers have taken into account many variables in their approach, clouds behavior remains a difficult thing to predict. Sandia National Laboratories Stein sums it up: Clouds are not static, they are forming as a function of convection currents in the atmosphere and so weve actually taken some imagery here in Sandia that shows that under certain circumstances clouds can pop into view and then disappear [
] while youre trying to track them, so thats one of the challenges. Stein also believes that, for forecasts beyond 30 minutes, TSIs may not be able to add significantly to forecast accuracy. Kleissl also admits that if plants get really big, more than 500 MW, the resolution of satellites become sufficient for forecasting across different parts of the installation.
The shape of plants themselves introduces another variable. They are installed in real-world environments, where there are obstacles on the ground that they must be built around. As such, larger parks can take non-standard shapes, further complicating the task of forecasting. Youve got a weird shaped plant and a weird shaped hole in the clouds, describes YES Beaubien.
Whether the algorithms developed by the team at the Copper Mountain plant research project can overcome these challenges is yet to be seen. Sandias Stein has some confidence in the research: I think the group thats leading [it] has demonstrated using TSIs [and] the concept works. If sophisticated short-term PV forecasting does become more widespread, another argument against PV may be silenced and utilities confidence in PV expanded. As, in the words of the SunShot Program Manager Ramesh: At the end of the day, this is why were doing it, because we want to know what impact [clouds] have on the utility grid.
###MARGINALIE_BEGIN###
###MARGINALIE_END###
This content is protected by copyright and may not be reused. If you want to cooperate with us and would like to reuse some of our content, please contact: editors@pv-magazine.com.
