Avoiding the Flaw of Averages in Solar Energy Investments

“Plans based on average assumptions are wrong on average.” This is how Stanford scientist Sam L. Savage defines the flaw of averages. Our natural inclination is to make decisions solely based on the long-term average values, without taking into account the variation around this average. In solar energy, this is precisely what many developers are doing: using a single-point estimate for the solar irradiance of a project site, and assuming that the level of incoming solar radiation will not change dramatically over time.

Much like the story of the statistician, who is perfectly happy in general, with his head in the fridge and his feet in the oven, this way of thinking may lead to some very bad decisions.

Solar variation
There’s a widespread perception in the solar energy industry that solar irradiance doesn’t vary too much from year to year at a given project site. Research suggests that, even on an annual basis, solar irradiance can vary significantly. Data for Ontario, Canada, obtained from both ground measurements and satellite observations, indicate that the variation can reach up to 10% from the long-term average for any given year.

This level of variation may not seem significant at first, but it’s the likelihood of back-to-back lower than average irradiance years that may cause a real problem. Such an event can dramatically hinder the debt servicing capability of a solar energy investor. So, how do we avoid the flaw of averages in the solar industry? The key is to understand the statistical behavior of solar irradiance by using historical time series data. This is precisely why large-scale solar energy investments require painstaking resource assessment studies. Although these studies are costly, developers and creditors are interested in them, as they decrease the level of uncertainty of an investment.

In recent years, a new generation of solar resource assessment tools has been developed, significantly cutting the cost of these studies. Today, obtaining a detailed analysis of solar irradiance is within reach of individual households, empowering investors to make the best possible decisions. When one considers the potential of a project site, a useful metric is the hourly average global horizontal irradiance (GHI). GHI is the total amount of short-wave radiation reaching a flat surface on the ground, generally measured in watts per square meter. As the standard yardstick of the industry, it’s a useful metric for comparing different project sites.

Any data collection method is prone to have missing data points. That’s why computing the hourly average of GHI can give a more accurate comparison of year-to-year variation (as opposed to the sum of all observations for a given year). One can always convert this value to the annual sum by multiplying it with the number of hours in a year. For example, the hourly average GHI in Toronto, Ontario is 158 watts/m2 and the annual average GHI is 158 x 8760 = 1,384,080 or 1.38 MWh/m2. However, these metrics are not sufficient for us to assess the natural variability of the solar resource.

Meaningful statistics
One way to move beyond the long-term averages is to derive meaningful statistics, such as the standard deviation of solar irradiance, from historical data sets. There are three main alternatives for finding historical data: onsite measurements, existing databases from ground and/or satellite measurements, and reanalysis data.

The most accurate data come from onsite measurements taken at the project site, ideally for a period of more than five years. Using instruments such as pyranometers, solar irradiance can be measured with great accuracy. Although this is by far the most reliable method, it’s not very practical for two main reasons: high cost and the long duration required for taking sufficient measurements. One’s own “weather station” composed of a good quality pyranometer, a data logger, and a transmitter will cost thousands of dollars. Perhaps more importantly, this type of data is not available on demand. At least one year of data has to be collected before any significant analysis can be performed. Even then, multi-year variations cannot be captured with one year of data.

The second alternative is to use existing databases, which contain solar irradiance data. These databases are maintained by the government and the private sector, and their spatial and temporal data coverage varies. In recent years, solar radiation estimates from meteorological satellite data have become more common. Complementing existing ground measurements, satellite data help increase the spatial coverage of the databases. Although these data sources may not be as accurate as onsite measurements, they are available on demand, and can be obtained at a reasonable cost.

The third data source is reanalysis data. Reanalysis is a scientific method that combines historical data from many different measurement systems using a consistent numerical model. These datasets cover multiple decades, and they enable us to conduct many different types of analysis. One disadvantage of reanalysis data is the relatively poor spatial resolution, so finding a data point close to a project site may not always be possible.

Of course, combining these different options according to one’s specific needs is probably the best way to avoid the flaw of averages. Moving beyond a single-point estimate of a solar resource may seem daunting at first. But, by setting realistic expectations of an investment right from day one, one can avoid a lot of headache down the road.   


Dr Ozgur Gurtuna is the president of Turquoise Technology Solutions Inc.

Turquoise Technology Solutions Inc.
www.turquoisetech.com