Pangu-Weather Driven Dual-Path Network with Dynamic Gated Attention for Ultra-Short-Term Wind Power Forecasting

A Physics-Regularized Gated Model with Patch-Based Temporal Encoding for Short-Term Probabilistic Wind Farm Power Forecasting

: Accurate probabilistic wind farm power forecasting is essential for reserve scheduling, dispatch decision-making, and risk-aware operation under high levels of wind power penetration. However, short-term wind power sequences exhibit strong nonstationarity, heterogeneous environmental variables exhibit time-varying predictive relevance under different meteorological regimes and operating states, and historical operating states and future numerical weather prediction (NWP) variables contribute differently over the forecasting horizon. In addition, direct quantile forecasts may suffer from quantile crossing or physically inconsistent wind-speed-power responses. To address these issues, this paper proposes a physics-regularized gated model for short-term probabilistic wind farm power forecasting, using hourly Local Peak Power (LPP) as the operational evaluation target. The temporal encoder follows the PatchTST patching strategy to capture both short-term ramping behavior and longer-range temporal dependence. A group-gated environmental variable selection and fusion module is designed to adaptively emphasize physically relevant meteorological variable groups and conditionally integrate historical representations with future NWP information. Moreover, a monotone multi-quantile prediction structure with a physics-based regularization term is introduced to improve probabilistic coherence and wind-speed-power consistency. Experiments are conducted against ten representative baselines covering empirical, tree-based ensemble, recurrent, convolutional, and Transformer-based probabilistic forecasting methods. The proposed model achieves the lowest mean absolute error (MAE), root mean squared error (RMSE), and continuous ranked probability score (CRPS) and obtains a 90% prediction interval coverage probability (PICP90) of 0.9001 and a 90% prediction interval normalized average width (PINAW90) of 0.3123, indicating near-nominal interval coverage with a moderate interval width. Relative to the strongest baseline for each metric, the proposed model reduces MAE by 4.92% compared with LightGBM-QR and reduces RMSE and CRPS by 2.78% and 1.94% compared with XGBoost-QR, respectively. Statistical significance analysis then confirms the overall performance gains, while ablation, interpretability, and sensitivity analyses verify the effectiveness of group-gated fusion, the future NWP branch, and physics-based regularization. These results demonstrate that the proposed model provides accurate, reliable, and physically consistent probabilistic forecasts for short-term wind farm operation.

Deep Multi-Modal Fusion and Kalman Online Adaptation for Electric Load Forecasting Against Extreme Rainfall Events

Hybrid Short-Term Wind Power Forecasting Using EEMD and Feature- Temporal Attention-Based TCN-BiLSTM Network

To improve short-term wind power forecasting accuracy and interpretability, this paper proposes an EEMD-FTA-TCN-BiLSTM model, where FTA denotes feature-temporal attention. The method first uses Ensemble Empirical Mode Decomposition (EEMD) to adaptively decompose the non-stationary power sequence into high-, medium-, and low-frequency components, representing short-term fluctuations, periodic variations, and long-term trends. TCN extracts local temporal patterns through causal and dilated convolutions, while BiLSTM captures bidirectional temporal relationships within the historical input window. A feature-temporal attention mechanism dynamically fuses key meteorological and historical power features. LightGBM then learns and corrects systematic residual errors in the preliminary forecasts. Pearson correlation analysis is used to provide interpretable feature analysis and preliminary feature screening. Validated with one year of data from a single wind farm in Inner Mongolia, the model achieves lower forecasting errors than the compared benchmark models and shows consistent forecasting performance across the meteorological conditions examined within the studied site. The ablation results support the contributions of the evaluated components. The proposed framework improves forecasting accuracy and interpretability for single-site short-term wind power forecasting, while multi-site spatial dependency modeling is left for future work.

Spatiotemporal Feature Fusion Using U-Shaped Architecture for Accurate Wind Speed Prediction

A U-shaped spatiotemporal feature fusion network named U-STNet is developed, which realizes joint modeling of inter-turbine spatial correlations and multi-period long-range temporal dependencies for wind speed forecasting and verifies the effectiveness of jointly modeling turbine spatial topology and multi-scale temporal dependencies for wind speed forecasting.

MSF-TransPV: a multi-source fusion transformer for short-term photovoltaic power forecasting

Accurate short-term photovoltaic (PV) power forecasting is essential for secure grid operation, economic dispatch, and efficient utilization of renewable energy. However, PV output exhibits strong nonlinearity and nonstationarity due to rapidly varying meteorological conditions, cloud movements, and system uncertainties. Conventional statistical models and single-source deep learning approaches often fail to fully exploit the rich multi-source information available in modern PV plants, such as historical power, on-site meteorological measurements, and clear-sky or numerical weather prediction (NWP) features. In this paper, we propose MSF-TransPV, a Multi-Source Fusion Transformer framework for short-term PV power forecasting. The model adopts a multi-branch temporal encoder that separately processes historical PV output, meteorological variables, and optional clear-sky/NWP-derived features, mapping them into a shared latent space. A cross-variable multi-head attention module is then introduced to explicitly capture the dependencies between PV dynamics and atmospheric conditions, enabling fine-grained interaction across different feature sources. On top of the fused representation, a temporal Transformer encoder models long-range temporal dependencies and feeds a lightweight decoder that supports both point forecasting and probabilistic forecasting via quantile regression. Experimental results on real-world PV datasets demonstrate that MSF-TransPV consistently outperforms persistence, statistical baselines, recurrent neural networks, and vanilla Transformer models in terms of RMSE, MAE, and normalized error metrics, while also providing reliable prediction intervals. These results indicate that explicit multi-source fusion and cross-variable attention are highly effective for PV forecasting under complex and rapidly changing weather conditions.

Optimization and Accuracy Improvement of Power Forecasting Models for Wind Farms Under Transient Weather Conditions

Given the problems of non-stationary power time series, response lag, and amplified prediction errors due to sudden changes in wind direction under transitional meteorological conditions, this study proposes a wind power forecasting model that integrates multiscale time-series features, transition-aware attention mechanisms, physical constraints on turbine operation, and dynamic residual correction. To improve the model's ability to jointly characterize different scales of meteorological evolution and power lag characteristics, a feature extraction network with short-, medium-, and long-term branches based on TCN (Temporal Convolutional Network) is built, which incorporates bidirectional GRU and Transformer architectures; additionally, the feature weights at each scale are dynamically adjusted according to the severity of inflection points, and the prediction output is constrained by air density, power curves, operational status and ramping limits. Residual caching and time-delay gating are employed to correct for lag errors in the range of 1 to 6 sampling steps. 50,842 valid data sets from a wind farm were used for validation. The general MAE and RMSE of the model were 0.258 MW and 0.386 MW, respectively, and these were lower than those of the baseline TCN by 21.58% and 20.90%. In the transition period, the MAE and RMSE were 0.305 MW and 0.448 MW; these had been reduced by 22.19% and 23.29%. Therefore, the developed model can reduce peak deviations caused by abrupt changes in weather and improve the accuracy and stability of wind power forecasting under all operating conditions.