Multiscale Quantum-Enhanced Forecasting Framework for Photovoltaic Power Output and Electric Load

Accurate day-ahead forecasting of photovoltaic (PV) power and electric load is essential for reliable power system operation under growing renewable penetration and demand complexity. Both series, however, are strongly nonstationary, with multiscale dynamics driven by external factors. This study develops a multiscale quantum-enhanced forecasting framework integrating rolling variational mode decomposition (VMD), feature attention (FA), and bidirectional quantum temporal modeling. Rolling VMD separates slower trends from rapid fluctuations, FA weights weather and calendar inputs, and the bidirectional quantum module encodes nonlinear dependencies within the observed history to represent irradiance-driven PV variability and temperature- and calendar-related load dynamics. Historical context is combined with the corresponding previous-day target for direct day-ahead forecasting. Using the Desert Knowledge Australia Solar Centre (DKASC) photovoltaic dataset and the GEFCom2014 load dataset, the seasonal reference comparisons cover four seven-day intervals per task. Across the 672 hourly evaluation points per task, the five-run mean RMSE and R2 are 2.0299 kW and 0.9618 for PV and 7.5385 MW and 0.9633 for load. PV evaluation includes nighttime zeroing based on observed irradiance. Ablation results confirm the complementary contributions of the three modules, supporting a unified approach to supply- and demand-side day-ahead forecasting.