An Advanced Forecasting Model for Global Irradiance in Solar Electric Vehicles Using a Fuzzy C-Means-CNN-Wavelet BiLSTM Network with an Attention Mechanism

Abstract

Precise prediction of solar radiation is crucial for improving energy efficiency and planning routes in electric vehicles powered by solar energy. Traditional forecasting methods frequently struggle to handle the unpredictability of weather conditions. Such limitations negatively impact the vehicle’s efficiency and range, underscoring the importance of developing more sophisticated prediction methods.Methods such as Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNNs) have shown promise. However, they struggle to capture intricate temporal dependencies and prioritize key features. Hybrid approaches like CNN-LSTM also have limitations. For instance, they lack advanced tools like wavelet transformations and attention mechanisms. Additionally, clustering methods remain underutilized, which constrains their accuracy.This study introduces FZ-CNN-WBiLSTM-AM, an innovative hybrid model. The model integrates fuzzy clustering (FZ), (CNNs), Wavelet-enhanced BiLSTM, and an attention mechanism (AM). Fuzzy clustering improves data preprocessing by addressing variability. CNN extracts temporal patterns, while WBiLSTM captures both historical and future dependencies. The attention mechanism enhances focus on critical temporal features.The proposed model demonstrates exceptional performance. It achieves coefficients of determination (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) close to 0.99. Furthermore, it significantly reduces error metrics such as MAPE and RMSE. The model consistently outperforms existing approaches, showing resilience under diverse weather conditions. This work provides a groundbreaking solution for precise solar radiation forecasting. It advances the efficiency and autonomy of solar-powered EVs.