| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.65 MB | Adobe PDF |
Advisor(s)
Abstract(s)
A implementação de um parque eólico envolve uma série de desafios, incluindo a
avaliação da sua viabilidade económica e a obtenção de financiamento. Portanto, a análise do
recurso eólico e da estimativa de energia associada desempenha um papel crucial. No entanto,
a estimativa da produção anual de energia elétrica pré-construtiva enfrenta várias fontes de
incerteza, que podem ser atribuídas a medições realizadas ou a pressupostos dos modelos
utilizados.
O objetivo deste estudo consiste em identificar uma metodologia simples para estimar a
produção anual de energia elétrica de Longo Termo pós construção. Isso foi realizado através
da análise da correlação entre diversas fontes históricas de dados de vento e as produções
mensais reais observadas em parques eólicos em operação. Além disso, este estudo investigou
como o período de operação do parque eólico afeta a estimativa de produção de energia a Longo
Termo em projetos com maior tempo de operação. Também avaliou o impacto das diferentes
fontes históricas de dados de vento na estimativa de produção de energia de Longo Termo. Em
parques eólicos que fazem parte de um complexo eólico, o método foi aplicado ao complexo e
a estimativa de produção anual de energia foi comparada com a estimativa agregada da
produção individual dos parques eólicos. O estudo desenvolveu-se em três fases distintas: a
análise e validação da metodologia adotada, a análise da influência do tempo de correlação em
parques com 20 e 4 anos de operação e o estudo da influência da agregação espacial em
complexos eólicos.
A metodologia utilizada foi analisada através de um estudo da relação entre o coeficiente
de correlação (R²) e o desvio da produção de energia de Longo Termo em relação ao valor
médio de produção real do parque, indicando que o R² é um indicador confiável para a seleção
adequada da série de reanálise a ser utilizada no cálculo da estimativa de produção de energia.
Para além disso, realizou-se uma avaliação do cumprimento dos pressupostos da regressão
linear e os resultados revelaram que, de uma forma geral para todos os parques e complexos, o
modelo de regressão linear empregado é estatisticamente significativo. Além disso, os resíduos
gerados pelo modelo apresentam uma distribuição normal, e na maioria dos casos, não foi
identificada auto correlação nos resíduos. Ao analisar os parques com 20 anos de operação, observou-se uma tendência de redução
gradual da dispersão dos dados à medida que o número de anos aumenta. Isso sugere que, para
obter estimativas de Longo Termo mais precisas e próximas da realidade, é recomendável
utilizar o maior número possível de anos de dados operacionais. Em relação à mediana, não
foram identificadas diferenças significativas entre os grupos, pois as combinações resultaram
em valores medianos semelhantes. As métricas de erro, como o RRMBE e RRMSE, seguiram
o comportamento esperado, apresentando uma diminuição na dispersão à medida que o número
de anos de operação aumenta. Isso sugere que as previsões estão, em média, mais próximas dos
valores reais, com menos viés ao longo dos anos, indicando uma maior consistência nas
estimativas. Quanto aos parques com 4 anos de operação, os resultados apontam para diferenças
e nuances entre os grupos, principalmente em relação à dispersão dos resultados, no entanto,
essas diferenças não se mostraram estatisticamente significativas. No contexto dos complexos
eólicos, a análise revelou que não há uma diferença significativa entre a utilização da série de
reanálise mais adequada para o complexo e a melhor para cada parque individualmente. Para
além disso, parece não parece haver uma vantagem evidente na realização de estudos
individuais para cada parque relativamente ao estudo para o complexo, visto que a diferença
entre essas abordagens é insignificante em comparação com a incerteza típica do modelo e da
própria estimativa de produção.
Implementing a wind farm involves several challenges, including assessing its economic viability and obtaining financing. Therefore, the analysis of the wind resource and associated energy plays a crucial role. However, the estimation of pre-construction annual electricity production faces several sources of uncertainty, which can be attributed to measurements taken or to assumptions in the models used. The aim of this study is to identify a simple methodology for estimating annual long term electricity production. This was done by analyzing the correlation between various historical sources of wind data and the actual monthly productions observed in operating wind farms. In addition, this study investigated how the period of operation of the wind farm affects the estimate of Long-Term energy production in projects with a longer period of operation. The impact of different historical sources of wind data on the estimate of Long Term Energy Production will also be assessed. In wind farms that are part of a wind complex, the method will be applied to the complex and the estimate of annual energy production will be compared with the aggregate estimate of individual wind farm production. The study was carried out in three distinct phases: validation of the methodology adopted, analysis of the influence of correlation time on wind farms with 20 and 4 years of operation, and study of the influence of spatial aggregation on wind complexes. The methodology used was analyzed through a study of the relationship between the correlation coefficient (R²) and the deviation of long-term energy production from the park's actual average production value, indicating that the R² is a reliable indicator for the appropriate selection of the reanalysis series to be used in calculating the energy production estimate. In addition, an assessment of compliance with the linear regression assumptions was carried out and the results revealed that, for all parks and complexes, the linear regression model employed is statistically significant. In addition, the residuals generated by the model have a normal distribution, and in most cases, no autocorrelation was identified in the residuals. When analyzing parks with 20 years of operation, there was a tendency for data dispersion to gradually decrease as the number of years increased. This suggests that to obtain long-term estimates that are more accurate and closer to reality, it is advisable to use as many years of operational data as possible. Regarding the median, no significant differences were identified between the groups, as the combinations resulted in similar median values. The error metrics, such as RRMBE and RRMSE, followed the expected behavior, showing a decrease in dispersion as the number of years of operation increased. This suggests that the forecasts are, on average, closer to the real values, with less bias over the years, indicating greater consistency in the estimates. As for the parks with 4 years of operation, the results point to differences and nuances between the groups, mainly in relation to the dispersion of the results. However, these differences were not statistically significant. In the context of wind complexes, the analysis revealed that there is no significant difference between using the most appropriate reanalysis series for the complex and the best one for each individual farm. Therefore, there does not seem to be a clear advantage in carrying out individual studies for each park, since the difference between these approaches is insignificant compared to the typical uncertainty of the model and the production estimate.
Implementing a wind farm involves several challenges, including assessing its economic viability and obtaining financing. Therefore, the analysis of the wind resource and associated energy plays a crucial role. However, the estimation of pre-construction annual electricity production faces several sources of uncertainty, which can be attributed to measurements taken or to assumptions in the models used. The aim of this study is to identify a simple methodology for estimating annual long term electricity production. This was done by analyzing the correlation between various historical sources of wind data and the actual monthly productions observed in operating wind farms. In addition, this study investigated how the period of operation of the wind farm affects the estimate of Long-Term energy production in projects with a longer period of operation. The impact of different historical sources of wind data on the estimate of Long Term Energy Production will also be assessed. In wind farms that are part of a wind complex, the method will be applied to the complex and the estimate of annual energy production will be compared with the aggregate estimate of individual wind farm production. The study was carried out in three distinct phases: validation of the methodology adopted, analysis of the influence of correlation time on wind farms with 20 and 4 years of operation, and study of the influence of spatial aggregation on wind complexes. The methodology used was analyzed through a study of the relationship between the correlation coefficient (R²) and the deviation of long-term energy production from the park's actual average production value, indicating that the R² is a reliable indicator for the appropriate selection of the reanalysis series to be used in calculating the energy production estimate. In addition, an assessment of compliance with the linear regression assumptions was carried out and the results revealed that, for all parks and complexes, the linear regression model employed is statistically significant. In addition, the residuals generated by the model have a normal distribution, and in most cases, no autocorrelation was identified in the residuals. When analyzing parks with 20 years of operation, there was a tendency for data dispersion to gradually decrease as the number of years increased. This suggests that to obtain long-term estimates that are more accurate and closer to reality, it is advisable to use as many years of operational data as possible. Regarding the median, no significant differences were identified between the groups, as the combinations resulted in similar median values. The error metrics, such as RRMBE and RRMSE, followed the expected behavior, showing a decrease in dispersion as the number of years of operation increased. This suggests that the forecasts are, on average, closer to the real values, with less bias over the years, indicating greater consistency in the estimates. As for the parks with 4 years of operation, the results point to differences and nuances between the groups, mainly in relation to the dispersion of the results. However, these differences were not statistically significant. In the context of wind complexes, the analysis revealed that there is no significant difference between using the most appropriate reanalysis series for the complex and the best one for each individual farm. Therefore, there does not seem to be a clear advantage in carrying out individual studies for each park, since the difference between these approaches is insignificant compared to the typical uncertainty of the model and the production estimate.
Description
Keywords
Wind Energy Long-Term energy production estimation Data analysis Statistical tests Refinancing of wind projects