Stima della radiazione solare ultravioletta per mezzo di un modello neurale

In the last years the increase of ultraviolet (UV) radiation, that reaches the earth surface, has induced many authors to bring studies on the biological effects that this phenomenon can have on ecosystems and on human beings. In particular, for safeguard health of outdoor workers, an estimate of UV radiation has basic importance to establish maximum time of exposure. Although this problem is important, UV radiation is almost never measured in standard weather stations. The aim of present research has been to plan and to develop a neural model able to estimate UV radiation using meteorological data available in standard weather stations without apposite UV sensors. An objective of this work is, also, to estimate the UV with a minimum number of sensors. This allows to obtain a measurement system with markedly inferior cost with respect to a sensor for a direct measure of the UV. Our analysis brought us to consider, for UV estimation, two measured quantities (global solar radiation and air temperature) and two calculated quantities (theoretical solar radiation and solar elevation from horizon). The low cost of the only two used sensors, normally on the whole inferior to a UV sensor, makes the proposed system useful to carry out various measure points with reasonable costs. The obtained results in the experimental sites have shown that the neural model provides good capacity to estimate UV radiation. In the period of higher interest (spring-summer), the statistical precision of the difference between calculated values of UV (by the model) and measured values (by the instruments), normalized to the maximum measured value, is better than 0.2 %. For this reason the proposed neural model can be effectively utilised as an alternative for the models, alike “low cost”, where these do not give satisfactory results.