Wind Power Prediction Model Using Machine Learning

Abstract

Before installing a wind turbine, it's essential to conduct wind power forecasting to gauge the effectiveness of the wind power initiative. Conventionally, wind speed measurements have been conducted instantaneously between various points. These measurement points solely indicate the locations where wind turbines will be positioned. However, these locations might exhibit reduced wind speeds, potentially making them less suitable for the optimal placement of the wind turbine. To address location challenges, we suggest conducting wind power predictions in areas where wind measuring instruments are yet to be installed. The study relies on the instantaneous measurements already performed at the site set up at the Dedan Kimathi University of Technology. To this end, a wind power forecasting model has been created. Real-time data from the site was gathered via a wireless sensor node utilising the Internet of Things (IoT). Additionally, a machine learning prediction model based on time series analysis was developed. Our forecasts were moderately aligned with the testing values, showing seasonality throughout the year. Therefore, the developed machine learning model captured the underlying patterns, trends, and seasonality in the wind data, making its forecasts reliable.