Analysis of Solar Irradiance Data Sets

The most accurate measurements of solar radiation are obtained by a pyranometer placed at a location for a number of years, usually on the order of a decade or more, measuring the direct radiation every few minutes. However, the volume of data generated by this technique makes it impractical (and unnecessary) to provide the full data set for each location for PV system design. Instead, the data can be presented in several other formats.

The most conceptually straight forward method of reducing the data set is to average the data over the measuring period. This form of data is called average daily, monthly or yearly radiation data. Although this data is useful for basic system design, the day-to-day variation in the solar radiation is lost. The loss of the day-to-day variation is critical since the design and performance of a system with, for example, 5 kWh/day nearly every day is quite different than one with 8 kWh/day on some days followed by several cloudy days with 2 kWh/day.

Comparison of TMY and average solar radiation data.

The most common format for solar radiation data is TMY data (or TMY2 or TMY3 data used by the National Renewable Energy Laboratories in the USA) which includes daily variability in the data. TMY data sets are described in the following page. However, average solar radiation data, particularly for each month of the year is also extensively used in rough estimates on the amount of PV panels required.

An additional useful, although less common data which can be determined from the full radiation data sets, is the probability of having a certain number of cloudy days which occur in a row, whereby the definition of a cloudy day is usually a day where less than 50% of the theoretically expected radiation is received. For example, at a certain location, 4 cloudy days in a row may occur once a year and 5 cloudy days in a row may occur once every 5 years. This information is particularly useful in estimating storage sufficient requirements. However, this information is less commonly tabulated and, if used, must be determined from the original data sets.