Posters

Abstract

Improving the air density time series estimation is one of the key aspects of reducing uncertainty on the expected annual energy production on wind farms. In this manner, this work brings a comparison between temporal downscaling techniques of a Reanalysis hourly temperature data in order to fill the gaps of a 10-minute basis in-situ temperature measurement in the Brazilian northeast region. This work compares temporal downscaling methods, constant, linear and Markov-based interpolation, to find the best approach to cope with this problem. Our Markov-based temporal downscaling technique can be described as follows: 1. Calculate the Markov state transition matrix utilizing the sazo-normalization on the wind met mast temperature data 2. Calculate the the sazo-normalized Reanalysis temperature data, ensuring the same percentiles levels of the previous calculated Markov state transition matrix 3. Apply the calculated Markov state transition matrix on the sazo-normalized Reanalysis temperature data 4. Temporally downscale the Reanalysis data (create a 10-minute temperature Reanalysis data) 5. Find the linear relation with the met mast data and the 10-minute temperature Reanalysis data 6. Fill the gaps in the met mast Given the seasonality and the stochastic nature of temperature data, the Markov-based temporal downscaling technique provides a good estimation of the gaps in the measurements, reducing the uncertainty on the expected air density time series, and thus, in the annual energy production of a specific Casa dos Ventos future wind farm.