Predicting wind turbine energy production with deep learning methods in GIS: A study on HAWTs and VAWTs

Abstract

The increasing global demand for renewable energy necessitates accurate forecasting methods to optimize wind energy production, particularly in regions with varying climatic conditions. This study addresses this need by utilizing advanced deep learning techniques and Geographical Information Systems (GIS) to estimate the energy output of wind turbines. Specifically, it focuses on predicting the energy production of both horizontal axis wind turbines (HAWTs) and vertical axis wind turbines (VAWTs) using a combination of Markov and Cellular Automata-Markov (CA-Markov) models, alongside deep learning methods such as long short-term memory (LSTM), LSTM-Wavelet, and Support Vector Regression (SVR). Additionally, the study evaluates the energy output of each turbine type, factoring in their construction costs within the study area. The analysis reveals significant variations in energy output over time, with maximum values increasing from 85,017 Wh in 2000 to 166,050 Wh in 2020 in the northern region, while minimum outputs also rose significantly. Projections for 2030 suggest that approximately 17% of the northern region experience a substantial increase in wind power potential. Among the forecasting methods, the LSTM-Wavelet hybrid model demonstrated superior accuracy, surpassing the 90% threshold, primarily due to its effective handling of data instability and noise reduction. This study underscores the potential of using sophisticated modeling techniques to enhance wind energy forecasting, contributing to more efficient energy management in regions with high energy demand and limited resources.