Forecasting of wind power generation using wind speed and temperature for thambapawani wind farm in Sri Lanka

Abstract

Wind power generation is a rapidly growing renewable energy resource in the world, both on a small and large scale. By integrating wind power generation systems, it helps to maintain grid stability, meet renewable energy targets, reduce greenhouse gas emissions, and promote economic growth while enhancing energy security by diversifying energy sources. Due to the intermittent nature of the wind and the influence of several weather parameters such as wind direction, ambient temperature, humidity, atmospheric pressure, the utilization of energy produced by the wind is challenging while maintaining the grid stability. Addressing this challenge involves the development of accurate forecasting models. Hence, in this study, accurate wind forecast models are built using two main weather parameters: wind speed and temperature for the newly implemented largest on-shore wind farm, "Thambapawani", A univariate model is built for the active power variable using the Seasonal Autoregressive Integrated Moving Average (SARIMA) method. Two different Vector Autoregressive (VAR) models were built with average wind speed and average temperature. However, all these models fail to grasp the intermittent nature of wind power alone. Therefore, hybrid models were generated using the above-mentioned models as mean models and Generalized Autoregressive Conditional Heteroskedasticity models as conditional variance models. All hybrid models were validated using the same test data set and evaluated with one of the goodness of fit tests called the root mean squared test. In this research, the forecasting horizon is 48 hours and the data resolution is 1 hour. The hybrid model of SARIMA (1,1,1) (1,1,1)24 with GARCH (1,1) is selected as the best-fit model that has the lowest RMSE value compared to the other two hybrid models in order to forecast wind power generation at “Thambapawani” Wind Farm in Sri Lanka. Keywords: wind power, grid stability, SARIMA, VAR, GARCH