Explainable artificial intelligence for wind power forecasting model based on long short-term memory
Open Access
- 14 May 2025
- journal article
- research article
- Published by Springer Nature in Neural Computing and Applications
- Vol. 37 (19) , 14589-14611
Abstract
Expanding the world’s economy leads to higher requirements for energy storage. Traditional energy resources decline at the same time that environmental contamination levels increase in the world. Wind power is the most potential energy resource supported by its status as a significant renewable energy system. Wind power generation has become a popular and exciting method among nations worldwide for generating renewable energy. High wind power generation unpredictability results in unavoidable errors throughout the wind power prediction process, creating substantial difficulties in the optimal management of power systems. Wind power prediction errors remain unavoidable, but appropriate wind power uncertainty models help power system operators reduce their adverse impact on operational decision-making performance. In this paper, developed an appropriate machine learning model that efficiently forecasted wind power data through time series analysis. The long Short-Term Memory (LSTM), Gated Reference Unit (GRU), and Autoregressive Integrated Moving Average (ARIMA) are the machine algorithms used in this investigation. This paper proposed an X-double LSTM, which integrates explainable artificial intelligence (XAI) and long short-term memory (LSTM). The XAI-Shapley Additive Explanations (SHAP) model is modified to pinpoint the crucial elements affecting the power generation forecasting model’s accuracy in cutting-edge solar systems. Nine metrics are used to assess the efficacy of the proposed X-double LSTM model: root mean square error, mean bias error, correlation coefficient, relative root mean square error, Nash–Sutcliffe efficiency, mean square error, mean absolute deviation, coefficient of multiple determination, and Willmott index of agreement. For MSE, RMSE, MAE, MBE, r, R2, RRMSE, NSE, and WI, the suggested model improves by 0.000 11, 0.011, 0.008, 0.008, 0.99, 0.98, 2.5, 0.98, and 0.98, respectively. Other machine learning methods, including the Transformer model, single-layer LSTM networks, Autoregressive Moving Average (ARMA) models, Gated Recurrent Unit (GRU) networks, and Bidirectional (Bi-LSTM) networks, are compared to the performance of double LSTM. The twin LSTM model was demonstrated to perform better. The simulations and experimental findings show that the suggested model can precisely estimate wind power. Within the Google Collab environment, the suggested model uses TensorFlow and Keras.Keywords
Funding Information
- Mansoura University
This publication has 39 references indexed in Scilit:
- Wind power forecasting considering data privacy protection: A federated deep reinforcement learning approachApplied Energy, 2022
- A combined model for short-term wind power forecasting based on the analysis of numerical weather prediction dataEnergy Reports, 2022
- A hybrid attention-based deep learning approach for wind power predictionApplied Energy, 2022
- Short-term wind speed forecasting based on spatial-temporal graph transformer networksEnergy, 2022
- Wind power prediction based on outlier correction, ensemble reinforcement learning, and residual correctionEnergy, 2022
- Short-term wind speed forecasting system based on multivariate time series and multi-objective optimizationEnergy, 2022
- A novel genetic LSTM model for wind power forecastEnergy, 2021
- Wind power prediction using a three stage genetic ensemble and auxiliary predictorApplied Soft Computing, 2020
- Scene learning: Deep convolutional networks for wind power prediction by embedding turbines into grid spaceApplied Energy, 2019
- A hybrid wind power forecasting model based on data mining and wavelets analysisEnergy Conversion and Management, 2016