Assessment of the Prediction Capacity of a Wind-Electric Generation Model

Abstract

Harnessing the wind resource requires technical development and theoretical understanding to implement reliable predicting toolkits to assess wind potential and its economic viability. Comprehensive toolkits, like RETScreen®, have become very popular, allowing users to perform rapid and comprehensive pre-feasibility and feasibility studies of wind projects among other clean energy resources. In the assessment of wind potential, RETScreen® uses the annual average wind speed and a statistical distribution to account for the month-to-month variations, as opposed to hourly- or monthly-average data sets used in other toolkits. This paper compares the predictive capability of the wind model in RETScreen® with the effective production of wind farms located in the region of Ontario, Canada, to determine which factors affect the quality of these predictions and therefore, to understand what aspects of the model may be improved. The influence of distance between wind farms and measuring stations, statistical distribution shape factor, wind shear exponent, annual average wind speed, temperature and atmospheric pressure on the predictions is assessed. Based on the results, a maximum distance radius of 30km from the measuring station to the project site is recommended when using RETScreen toolkit. The wind shear exponent came off as the second most influential parameter after the annual average wind speed in terms of the prediction of accurate Power Generation. It was found that the former must be carefully chosen to match the type of terrain in order to achieve up to a 4-9% improvement in the current prediction capability of the software.