SCADA system dataset exploration and machine learning based forecast for wind turbines

Abstract

Effective short-term wind power forecast is essential for adequate power system stability, dispatching and cost control. There are various significant renewable energy sources available, including wind power, which is one of the most favorable power source. As a result, wind power is an important green form of electricity generation. Wind power prediction is a crucial topic in reducing the energy system's unpredictability. The act of balancing energy supply and demand is critical. The main aim of this research is to offer a time-series dataset-based prediction tool for estimating wind power. In addition, for visualizing the dataset obtained from the SCADA (Supervisory Control and Data Acquisition) system, dataset exploration approach is also presented. The dataset analysis approach are shown in polar coordinate systems and pair plot diagram for better understanding of the relation between wind and power production relationship. With the help of given input features i.e. theoretical power, produced active power, wind direction, wind speed, month and hour, generated turbine power is forecasted using various machine learning algorithms. The performance of the developed model has been assessed using several statistical criteria. The experimental findings reveal that the XGBoost regression approach has greater prediction accuracy than the other methods with R2 of around 0.969, MSE of 0.003, RMSE of 0.064, MAPE of 0.282, and MAE of 0.026.