Genetic Taboo Search Algorithm Optimized TCN-LSTM for Ultra-short-term Photovoltaic Power Prediction

Solar energy is converted into electricity through photovoltaic power generation. This process is highly volatile and subject to a variety of factors. Characterizing this volatility is a crucial step in improving the accuracy of PV power prediction. This research suggests combining Genetic Taboo Search Algorithm (GATS) optimized Temporal Convolutional Networks (TCN) and Long Short-Term Memory Networks (LSTM) to optimize the ultra-short-term PV power forecast model in order to overcome this challenge. Firstly, the crucial meteorological factors affecting the PV power are selected for correlation analysis. Secondly, the optimal weight thresholds for the TCN and LSTM prediction models are optimized applying the genetic taboo search algorithm. In order to obtain the final optimized prediction results, the information entropy weight allocation method is used to linearly combine the two models of the TCN and the LSTM. The research results demonstrate that the prediction model is capable of accurately describing the ultra-short-term PV power volatility. Furthermore, the power prediction ac-curacy and reliability of the model outperform those of the conventional prediction approaches, demonstrating good adaptability and efficacy.