Wind energy prediction: a forecasting model based on ANNs and meteorological data on mesoscale

Abstract

Energy is essential to society for ensuring good quality of life by modern standards. Nowadays, fossil fuels are still the most used energy source, but, due to their depletion and contribution to climate change, the pursuit of a sustainable development has promoted an ever-growing trend to use new and pollution-free energy sources. Such a trend is impacting the energy scenario with massive transformations on a world scale. From the Kyoto protocol in 1997 to the COP 21 Paris agreement in 2015, great challenges have been introduced in terms of both emissions’ reduction and development of new energy sources, which are cleaner than the fossil ones. As a result, renewable energy sources (RESs) have seen a great development, favoured by a strong interest from governments, private companies, universities and public and private research centres. In fact, estimates suggest a RES penetration of over 55% in the next few years. Obviously, such a process is not likely to occur in the same fashion in all countries. As a matter of fact, RESs are not uniformly distributed, and incentive policies differ very much according to the single countries. Among RESs, wind power is the most widespread in the world after hydropower: over the last few decades, the global wind installed capacity has grown rapidly, particularly in Europe, Asia, and North America. However, the unpredictable and intermittent nature of wind is the main obstacle to its integration on a large scale: grid operators have difficulties keeping the grid in a safe state when large volumes of this energy are injected into the power system. Hence, in order to manage wind capacity, accurate wind power forecasting is necessary. However, forecasting the wind power production is quite challenging as wind is extremely variable and depends on weather conditions, terrain factors, and height above ground level. Furthermore, wind power strongly depends on wind speed, thus for a successful integration of this type of energy into any power system, it is important to design a wind speed prediction model with a forecasting error which is as low as possible. Unfortunately, wind is the most difficult meteorological phenomenon to predict: wind forecasting thus represents a great challenge for researchers, meteorologist, and wind power producers. In the literature, several forecasting models have been proposed, traditionally based on physical and statistical methods. In addition to those, a number of more advanced methods based on artificial intelligence have been investigated in recent years, in the attempt to attain more reliable wind-power forecasts. ...[edited by Author]