Caracterização e previsão de potencial solar: estudo de caso para Parnaíba (PI), Maracanaú (CE) e Petrolina (PE)

• Ano de defesa: 2016
• Autor(a) principal: Melo, Francisco Evandro de
• Orientador(a): Carvalho, Paulo Cesar Marques de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Engenharia elétrica; Radiação solar; Análise de séries temporais
• Link de acesso: http://www.repositorio.ufc.br/handle/riufc/15925

The use of solar resource, as a supplementary source in the country’s energy matrix, requires a good forecast strategy that enables decision-making and strategic actions to keep the electrical potential generated in the national energy grid stable. Based on such premise, this dissertation presents the characterization and forecast of data series on solar irradiation registered at three locations in the Northeastern region of Brazil: from August 2012 to July 2013 in Parnaíba (PI), from May 2012 to April 2013 in Maracanaú (CE), and from May 2012 to March 2013 in Petrolina (PE). Those surveys consist of time series and, therefore, they require specific statistical methods for their own treatment and forecast. As seasonality is a characteristic found in time series data of solar irradiation, the characterization and forecast are made by using the low seasonality component of solar radiation: the atmospheric transparency index, known as Kt. The use of this component is justified for providing more accurate and reliable forecast results, with low interference from trend components present in time series data within the forecast process. The forecasts made in this study have used the ARIMA method, the Simple Exponential Smoothing method (SES), and the Moving Average (MA) method. Forecasts are obtained from a 30-day trial period for each location, adapted graphically between the dry season, and validated within periods of 30, 150 and 180 days in the rainy season. Then, the evaluation of the forecast methods used is done, being the ARIMA method a 30-day forecast and in need of validation within 30 days, which shows the lowest error rates, with Root Mean Squared Error for Prediction values (RMSEP) of 0.008 for Parnaíba (PI), 0.015 for Maracanaú (CE), and 0.010 for Petrolina (PE). Also, through the regression equations, by transforming the Kt values obtained with the ARIMA method forecast, one obtains a 30-day forecast of daily solar irradiation with averages of 6.4 kW/m² for Parnaíba, 5.69 kW/m² for Maracanaú, and 6.54 kW/m² for Petrolina