LSTM neural network techniques-based analytical predictive models for wind energy and mechanical power

Abstract

Nowadays, the importance of renewable energy is rapidly increasing. It is considered as an alternative clean source of energy due to environmental reasons. Therefore, this research presents a data analysis model to predict the generated electrical power based on wind energy and the long short-term memory (LSTM) model. The work focused on the Spring and Autumn seasons where wind speed has high variation and the data was collected every 15 min in a wide, open space area located in southeast Palestine. To investigate and validate the correctness and robustness of the work, three different scenarios were performed for each season to predict wind speed and direction, and mechanical power. Also, different performance metrics were applied. The results were very promising with an average error of less than 3% and an R-Squared value of 0.95. Since the price of electricity in Palestine is relatively high, the results showed also the possibility to generate electricity with lowered price of about 40% and a reasonable payback period of 11 years. The work confirms that wind energy is cost-effective and a good alternative to reducing global warming.