Short-term PV Power Prediction Based on TCN-Transformer Model

Photovoltaic power generation has become a key means of energy reform in order to realize the transition to a new power system under the "dual-carbon" goal. However, due to the influence of weather and other factors, PV power generation is fluctuating, nonlinear and intermittent, which makes its power difficult to be predicted accurately, thus affecting the stable operation of the power grid. For this reason, this paper proposes a short-term PV power prediction method based on the TCN-Transformer model. The model combines the advantages of Temporal Convolutional Network (TCN) and Transformer, where TCN extracts global spatial features through causal convolution and inflationary convolution, and Transformer captures the long-term dependencies of sequences through the self-attention mechanism. In this paper, we validate the model with a real dataset of PV power plants in a region and evaluate the model performance using MAE, RMSE and R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> metrics. The results show that the model proposed in this paper outperforms other comparative models in terms of prediction accuracy and generalization ability, provides an efficient method for PV power prediction, and helps to improve grid stability and economic dispatch capability.