Photovoltaic Power Forecasting Based on CNN-BiLSTM Optimized by an Enhanced Sparrow Search Algorithm with Sine-Cosine Mechanism and Cauchy Mutation

With the continuous growth of electricity consumption in society, conventional power generation methods are increasingly constrained by limited fossil fuel reserves and environmental pollution. As an abundant and sustainable resource, solar energy offers significant advantages-photovoltaic (PV) power generation is cost-effective, easy to deploy, and efficient. Consequently, PV power is expected to become a dominant electricity source in the future, making its development especially crucial. Nonetheless, the inherent variability of solar energy leads to instability in PV output, potentially causing damage to electrical systems. Accurate forecasting of PV output is thus essential. In this study, an enhanced sparrow search algorithm (ESSA) is employed for data preprocessing, followed by the use of a CNN-BiLSTM model for training and prediction. Experimental results demonstrate that the ESSA-CNN-BiLSTM framework achieves superior performance in EMA, MSE, and $\mathbf{R}^{\mathbf{2}}$ metrics, confirming its effectiveness in short-term photovoltaic power forecasting.