The Reference Year

As already mentioned, the most widely available information related to the solar radiation resource at a given location is the set of 12 monthly mean values of global horizontal daily irradiation, Gdm(0). The methods presented above allow estimation of all the radiation components incident on any surface of arbitrary orientation and at any moment of the average year, and even at any moment of a long sequence of years. This can be applied to all the problems related to the design of photovoltaic systems: sizing, prediction of energy yields, impact of shadowing, optimisation of tilt angles and so on.

Nevertheless, the solar radiation is still the object of systematic recording, and more and more irradiance and irradiation data are being accumulated and put at public disposal. Such data, whether in the form of crude recorded data or in the form of elaborate mathematical tools, attempt to properly represent the climate of the concerned location. The most widely used is the so-called reference year, also called the typical meteorological year TMY, or the standard year [36]. The TMY for a location is a hypothetical year in which months are real months, but are chosen from different years from the whole period for which data are available. In practice [37], the months are chosen such that the monthly mean of the daily global irradiation on the horizontal represents an average value for all values contained in the database. For example, January of 1986 was chosen for the TMY of Madrid, because it had a value of Gdm(0) = 1.98kWh/m2, the closest to the average value of Gdm(0) = 1.99kWh/m2 for all the months of January on record [14].


The most widely used TMY for photovoltaic applications is set in a one-hour time scale. Hence, it contains 4380 values[121] of global horizontal irradiation. Ambient temperature values are also specified for each hour. This huge number of initial data can lead to the impression that the corresponding results should be much more accurate than those obtained when simply using the 12 Gdm(0) values as input. However, this impression is largely wrong. On the one hand, because the representativeness of any data – it should be again remembered – is limited by the random nature of the solar radiation, small differences in the results are scarcely meaningful. On the other hand, because the results obtained from the 12 Gdm(0) values and from the TMY are very similar, provided the initial data are coherent (i. e. the monthly means in the TMY coincides with the 12 Gdm(0) values) and that the selected correlations and diffuse radiation models to transpose from horizontal to inclined surfaces are the same. The physical reasons for this lie in the, already mentioned, quasi-linear power-irradiance relationship in most PV devices, and in the fact, initially shown by Liu and Jordan [15], that the solar climate of a particular location can be well characterised by only the monthly mean daily clearness index. As already mentioned, they have demonstrated that, irrespective of latitude, the fractional time during which daily global radiation is equal to or less than a certain value depends only on this parameter. Surely, to go into this question in-depth would increase the reader’s boredom which is probably already large enough; hence, we will restrict ourselves to describing a representative case from our own experience:

In 1992, the Solar Energy Institute in the University of Madrid, IES-UPM, was involved in the design of the 1 MW PV plant in Toledo, Spain. It was the biggest European PV project at that time, so very careful studies were required at the initial project stage. Fortunately, a large historical database, containing 20 years of hourly irradiation data, was available from a nearby meteorological station, and was directly used to calculate the expected energy yields. Both static and sun-tracked photovoltaic arrays were analysed, while taking into account detailed features such as shadowing from adjacent rows, back-tracking features and so on. Moreover, the same calculation was also performed using as input the TMY, previously derived from the historical radiation sequence, and also using as the only input the 12 Gdm(0) values and computing for just the mean day of each month. The results from the three calculation procedures never differed by more than 2%! As a matter of fact, the results were much more sensitive to the considerations of the solar angle of incidence effects [38] described below.

A clever friend, not involved in this project, but aware of this anecdote, posed the questions: Then why go into such exhaustive detail when they give similar results? Why not just be simple? The proper answers are best found in human psychology. Many people simply desire not to believe in some ideas. Hence, the ones daring to declare them are automatically impelled to provide strong arguments in favour of such ideas. To a large extent, this is usually the case when defending the argument that modern complex software tools do not necessarily yield better results than simple (but judicious) traditional methods. That was the position of the IES-UPM in 1992, and today for the author of this chapter the position remains the same.

Updated: August 23, 2015 — 5:30 am