A two-step forecasting solution and upscaling technique for small size wind farms located in hilly areas of Romania

Abstract

Taking into account the well-known benefits of using renewable energy as an important energy source, some practical aspects regarding operation must be underlined. In the specific case of wind power plants (WPP) forecasting the generated wind power is the most important issue for assuring stability of Romania's National Power System (NPS) and economic efficiency of wind power plant operation. The wind power plant operator should report daily to National Dispatching Centre, the hourly energy production for the next day. In case of inaccurate energy production prediction, the National System Operator must take additional measures in order to maintain the stability of NPS. The purpose of our research is to develop a solution, based on Artificial Neural Networks, for forecasting the wind energy production of small output power wind farms located in hilly areas of Romania, thus improving the accuracy of the hourly prediction. The case study is based on two-year historical data for a wind power plant comprising two power production groups situated in Tulcea County, in southeastern Romania. By using this approach, we have also explored the possibility of algorithm generalization, starting from the detailed model of the first production group and generalize it to the second one.