Forecasting power output of PV systems Market overview, technical solutions and implementation prop

The PV industry is growing worldwide, more grid-connected PV plants are injecting electricity to the grid and forecasting systems are required, wheter for system operators to control the stability of the grid than for market operators to cover the power demand. The present work is a comprehensive overview of PV- based forecasting systems, separated in six chapters, including the state-of-the-art, a market overview, two national case studies and a proposal for a new low cost solution. The development of this thesis was made within an italian company, called Building Energy, currently expanding its development pipeline all around the world. The rst chapter introduces the importance of forecasting the power generated from renewable sources and Building Energy as a company. Subsequently, the second chapter is a compilation of the state-of-the-art, with particular attention to Numerical Weather Prediction NWP models, used for larger forecast ranges and with more utility to the two case studies analyzed in the next chapters. The third chapter correlates the previous ndings to services available in the industry, a market overview including two italian, one dutch, one italian/swiss and one american weather providers, their speci c quotations and infomation that a manager type reader would appreciate. In Chapter 4 and 5, two national cases are studied: Italy and South Africa. The choice of reviewing these two countries was based on the Building Energy projects. The italian case, more complex in terms of the regulatory framework and electricity market, required a deeper economic analysis. The purpose was to clarify wheter it is convenient or not, to a single PV power producer, to change energy trader and assume its own forecasting, as a consequence of high unbalancing fees. It was found that in certain conditions, perform a better owned forecast, is economically more convenient. For the South African case, the analysis was straightforward, depending only on the authoroties regulations. Finally, the last chapter presents a proposal for a new low cost forecast system, based on free online meteorological data and the implementation of an Articial Neural Network ANN. It describes the system, indicating the main components and tools which have been tested. Its implementation is currently in evaluation process in the company.