Predictive analysis of RNN, GBM and LSTM network for short-term wind power forecasting

Abstract

One In the 21st century, Power sector has a drastic influence in today’s country economy. Role of renewable energy is increasing day by day. The renewable energy majorly comprises of hydro power, solar energy, wind energy. At present, ratio of wind energy to the total renewable energy is highest, as compared to hydro and solar. Also, with increase in population across the world, the enhancement in demand of electrical energy is also increasing. At the same time, revolution in the field of electronics and telecommunication also escalates the demand of electrical energy. With increase in the number of private players supplying the electricity, the competition rises proportionally. Therefore, in order to survive in the competitive market, private companies want to predict the future demand, which they can fulfill at the right time. Thus, they require some software based computing model which can forecast the future demand precisely. The existing predicting methodology utilizes linear models like: ARIMA, MA, AR at the same time also non-linear algorithms like: Neural Networks, GARCH, ARCH. As, the wind energy plays a major role in the renewable sector. So, we have to predict or forecast the power generated from wind energy using wind velocity via wind turbine, by three neural network based model that is RNN, GBM and LSTM, finally we have to find out which is better one depending upon the performance parameters values.

Subject Classification:

• Primary LSTM
• Secondary RNN
• GBM

Keywords:

• Short term forecasting
• RNN
• GBM
• LSTM
• Hidden Layers
• epochs
• Mean square error
• Mean absolute percentage error