Signal processing and climate understanding

Clean energy, as generated by wind and sun, is a must nowadays. It has, however, to be competitive with other sources, offering acceptable costs and production risks. This work faces the usual lack of a long and reliable historic of climato-logical data and consequent uncertainty about expected plant production. Customized machine learning and signal processing techniques are used to map regional satellite measurements into local information, creating a virtual historical series of energy production. Two illustrative cases are presented for eolic and photovoltaic energy, where neural network inputs and operations are described.