Short-Term Photovoltaic Power Prediction Based on IWOA-TCN-BiGRU-MATT

Abstract

: To address the challenges of significant nonlinearity, intricate temporal interdependencies, and the tendency to get stuck in local optima during parameter tuning in short-term photovoltaic (photovoltaic, PV) power forecasting, this paper puts forward a hybrid model called IWOA-TCN-BiGRU-MATT. This model fuses the IWOA (Improved Whale Optimization Algorithm, IWOA) with the TCN (Temporal Convolutional Network, TCN), BiGRU (Bidirectional Gated Recurrent Unit, BiGRU), and MATT. By leveraging IWOA to boost parameter optimization efficiency and integrating TCN’s ability to extract multi-scale features, BiGRU’s bidirectional temporal modeling, and MATT’s emphasis on key features, the model aims to realize high-precision PV power prediction. Based on measured data from a PV power station in Central China, experimental findings show that the proposed model surpasses competing methods under various meteorological scenarios, demonstrating improved prediction accuracy under different weather conditions.