Short term wind power forecasting using machine learning techniques

Wind power forecasting is essential for proper planning and scheduling the load for the grid. This provides us with a great advantage by ensuring efficient management of grid thereby reducing loss and thus, the cost of power production. As we dig deeper into the wind power forecasting, the analysis of huge data set becomes more complex and become very difficult to accurately forecast the output power. Thus, it is necessary to implement appropriate methods to forecast wind power accurately. This paper proposes a novel model of wind power forecasting using random forest technique. For investigating the effectiveness of this model, the comparison was made with two methods namely support vector regression and Decision tree. For training and testing of this model Kolkata region data of wind power and meteorological data is used. The result shows that Random Forest and Decision Tree techniques provide better forecasting accuracy with lesser MAPE. For Random Forest and Decision tree, the MAPE values obtained are 1.8999 and 1.2169 respectively compared to SVM with a MAPE of 20.8346.