Outperforming Self-Attention Mechanisms in Solar Irradiance Forecasting via Physics-Guided Neural Networks

TL;DR

This paper introduces a lightweight, physics-guided CNN-BiLSTM model for solar irradiance forecasting that outperforms complex Transformer-based models by leveraging domain knowledge and hyperparameter tuning.

Contribution

It presents a novel hybrid neural network architecture explicitly guided by physical features, challenging the complexity-first paradigm in meteorological time series forecasting.

Findings

The proposed model achieves an RMSE of 19.53 W/m^2, outperforming attention-based baselines with RMSE 30.64 W/m^2.

Explicit physical feature integration improves forecast accuracy over raw data-driven models.

Hyperparameter tuning via Bayesian Optimization enhances model performance.

Abstract

Accurate Global Horizontal Irradiance (GHI) forecasting is critical for grid stability, particularly in arid regions characterized by rapid aerosol fluctuations. While recent trends favor computationally expensive Transformer-based architectures, this paper challenges the prevailing "complexity-first" paradigm. We propose a lightweight, Physics-Informed Hybrid CNN-BiLSTM framework that prioritizes domain knowledge over architectural depth. The model integrates a Convolutional Neural Network (CNN) for spatial feature extraction with a Bi-Directional LSTM for capturing temporal dependencies. Unlike standard data-driven approaches, our model is explicitly guided by a vector of 15 engineered features including Clear-Sky indices and Solar Zenith Angle - rather than relying solely on raw historical data. Hyperparameters are rigorously tuned using Bayesian Optimization to ensure global optimality. Experimental validation using NASA POWER data in Sudan demonstrates that our physics-guided approach achieves a Root Mean Square Error (RMSE) of 19.53 W/m^2, significantly outperforming complex attention-based baselines (RMSE 30.64 W/m^2). These results confirm a "Complexity Paradox": in high-noise meteorological tasks, explicit physical constraints offer a more efficient and accurate alternative to self-attention mechanisms. The findings advocate for a shift towards hybrid, physics-aware AI for real-time renewable energy management.