Wind power forecasting over India: value-addition to a coupled model seasonal forecasts

Abstract Accurate predictions of wind power generation several months in advance are crucial for the effective operation and maintenance of wind farms and for facilitating efficient power purchase planning. This study evaluates the performance of the seasonal prediction system of the National Centre for Medium-Range Weather Forecasting in forecasting near-surface winds. An analysis of 23 years of hindcast data, from 1993 to 2015, indicates that the seasonal prediction system effectively captures the inter-annual variability of near-surface winds. Specifically, predictions initialized in May demonstrate notable accuracy, with a skill score of 0.78 in predicting the sign of wind speed anomalies aggregated across various wind farms during the high wind season (June to August). Additionally, we critically examine the peculiarity of a case study from 2020, when the Indian wind industry experienced low power generation. To enhance forecasting accuracy, we employ statistical techniques to produce bias-corrected forecasts on a seasonal scale. This approach improves the accuracy of wind speed predictions at turbine hub height. Our assessment, based on root mean square error, reveals that bias-corrected wind speed forecasts show a significant improvement, ranging from 54% to 93%.