Development of Short-Term Wind Power Forecasting Methods
Bo Cao, Liuchen Chang
Abstract
This paper presents new approaches for short-term wind power forecasts developed by the authors. Day-ahead and hours-ahead wind power forecasts were derived from the wind speed forecast data generated by High Resolution Deterministic Prediction System (HRDPS) Model at the Environment and Climate Change Canada (ECCC). Following a statistical analysis to verify the accuracy of the ECCC wind speed forecasts themselves, a power curve transfer model was developed to offer day-ahead wind power forecasts by converting the ECCC wind speed forecasts to wind power forecasts. An hours-ahead wind speed forecasting method was developed using a fusion approach to predict wind power for look-ahead times ranging from 30 minutes to six and a half hours with 5-min time steps to meet forecast delivery requirements of utilities and system operators. Using operational data over several years from six wind farms in different locations in Canada, the forecasting methodologies were validated for their good performance on the basis of statistical metrics and error distribution analyses.