China International Youth Conference on Electrical Engineeing (CIYCEE 2025)

Accurate forecasting of photovoltaic (PV) power is crucial for enhancing the reliability and economic viability of renewable energy systems. In this paper, we propose a novel hybrid architecture, GTCN, which synergistically integrates a Temporal Convolutional Network (TCN), a gated cross-attention mechanism, and a parallel Gated Recurrent Unit (GRU) decoder. The model efficiently captures both long-range temporal features and dynamic sequential dependencies in PV time series. Extensive experiments on real-world PV datasets demonstrate that our approach significantly outperforms traditional models such as LSTM, GRU, and standalone TCN in terms of forecasting accuracy and robustness.

Deep learning; Gated temporal convolutional network; Photovoltaic power prediction.