Estimativa da produção de energia solar fotovoltaica com base em modelos meteorológicos

Abstract

Photovoltaic energy had an exponential growth in the last few years in Brazil and soon it should become an important source of energy in the Brazilian electrical system. Unlike other sources, it is not possible to control the amount of energy generated by a photovoltaic system, since the irradiance has intermittent characteristics and seasonalities. So it requires good planning of the electrical system with estimations of production in various horizons, ranging from hours to years. Irradiance predictions are very important in this planning, and one of the main tools for forecasting it are the mesoescale numerical weather prediction models. The main model of this type, Weather Research and Forecasting Model (WRF), has been the subject of studies and optimizations aiming irradiance predictions. By the predictions of irradiânce and temperature it is possible to estimate the energy production by a photovoltaic system. This work involved the creation of several tools for a possible operationalization of an irradiance forecasting system. The tools automate several operations like retrieving data from ground stations and GSF and automatic runs of the WRF model. In addition, tests were carried out to verify the influence of proper parameterizations for irradiance predictions and different aerosol configurations in the WRF. The simulations were performed for the state of Rio Grande do Sul in the period of 20 days between March 12 and March 31 of 2018. The validation of the irradiance predictions used as reference sites of INMET. The WRF runs used as a boundary condition data from the Global Forecast System (GFS). Simulations were carried out with five sets of parameterizations, one with typical parameters and four with recommended parameters for irradiance predictions. Among the simulations with parametrizations specific to irradiance predictions were simulations disconsidering aerosols, using climatological aerosols, using ECMWF-CAMS aerosols and ECMWF-CAMS aerosols plus stochastic disturbances. Generation models were also created based on the WRF using the SAPM model for the distribution facilities of the domain of the irradiance forecast. The results showed that the specific parameterizations for irradiance predictions gave better results than typical parameterizations and the use of external aerosols and perturbations led to a small decrease of the error. WRF runs with irradiance prediction parameters were more accurate than GFS on days with little cloud coverage but performed worse on days with higher sky coverage. The power generation forecsts showed that the output power of the combined photovoltaic installations of the domain formed smooth curves without presenting significant oscillations in the energy production in the intervals of 30 min of the simulations.