Portal de Periódicos da CAPES

Big Data Analytics Framework for Smart Forecasting of Wind Power Generation

2022; RELX Group (Netherlands); Linguagem: Inglês

10.2139/ssrn.4312770

1556-5068

Dr.R.Saravanan Ramaiah, M Murugesan, Senthil Kumaran Selvaraj, Fahmi Elsayed, Mostafa Rashdan, Marc Azab,

Big Data Technologies and Applications

To minimize the environmental pollution, encourage the usage of alternate fuels and limit the usage of fossil fuels, which strengthens the range of integration of renewable energy sources, namely wind and solar power, has taken significance in many nations. The interest in the economic and technological challenges of integrating renewable energy sources like wind power into power grid systems has grown in response to the rising demand. Forecasting is essential and critical for securing integration with the power grid systems. To balance the power supply and demand, the grid operators has to schedule an optimal power generation from fossil fuel-based power generation and give priority to wind power generation.In this work, an efficient System on Chip (SoC) based IoT sensors are used to collect the real time data. The field sample of wind speed and direction is taken in customized intervals (once in 5 seconds).For wind power generation forecasting, a powerful real time architecture with the combination of Apache Kafka, a messaging queue system together with Apache Storm a processing engine, Mongo DB a heterogeneous data base and with an intelligent machine learning algorithms for Clustering and Classification is used. The wind power generation mainly depends on three prime factors i.e. Wind speed, Wind direction and Wind density. Also, the environment temperature, and humidity also considered for this application. With the use of Density Based Spacial Clustering of Application with Noise (DBSCAN) algorithm, cluster the similar pattern of Wind velocity and direction which leads to maximum power generation through wind source. Based on the similar relationship of the pattern, it is possible to forecast the wind power generation in advance together with the standard static (historical) data already stored in the cloud server. Random forest classification algorithm is also used together with DBSCAN to improve the classification accuracy.