The rapid deployment of photovoltaic generation increases uncertainty in power-system operation and strengthens the need for ultra-short-term forecasts with reliable uncertainty estimates. Point-forecasting approaches alone are often insufficient for dispatch and reserve decisions because they do not quantify risk. This study investigates probabilistic forecasting of short-horizon solar generation using quantile regression on a public dataset of solar output and meteorological variables. This study proposes a hybrid attention–convolution model that combines an attention-based encoder to capture long-range temporal dependencies with a causal temporal convolution module that extracts fast local fluctuations using only past information, preventing information leakage. The two representations are fused and decoded jointly across multiple future horizons to produce consistent quantile trajectories. Experiments against representative machine-learning and deep-learning baselines show improved probabilistic accuracy and competitive central forecasts, while illustrating an important sharpness–calibration trade-off relevant to risk-aware grid operation. Key novelties include a multi-horizon quantile formulation at 15 min resolution for one-hour-ahead PV increments, a HybridCNN–STTransformer that fuses causal temporal convolutions with Transformer attention, and a horizon-token decoder that models inter-horizon dependencies to produce consistent multi-step quantile trajectories; reliability/sharpness diagnostics and post hoc calibration are discussed for operational risk-aware use.

This study presents PAGER-Life, an adaptive hybrid framework for the early prediction of lithium-ion battery service life under heterogeneous operating conditions. The system integrates physically interpretable degradation features and standardized charge–discharge curves using linear regression, decision tree-based models, and neural networks. An error-aware routing mechanism combines expert predictions based on out-of-fold errors, while conformal prediction enables uncertainty quantification. The framework was evaluated using the MATR and HUST datasets with observation horizons of 20, 50, and 100 cycles, following a protocol that accounts for the grouped data structure and prevents information leakage. A comparative analysis involving 114 physical cells and 14 alternative methods demonstrated the following performance metrics for PAGER-Life: Mean Absolute Percentage Error (MAPE) of 12.827%, Mean Absolute Error (MAE) of 166.766 cycles, Root Mean Square Error (RMSE) of 210.715 cycles, and a macro-averaged coefficient of determination (R²) of 0.495. Ablation studies indicated that removing the physically grounded degradation features substantially degraded predictive performance. The results indicate that the adaptive integration of complementary prognostic models offers a promising approach for the early estimation of battery service life under heterogeneous operating conditions.