ESN-ATCN: A Solar Irradiance Prediction Model by ESN-TCN with Attention Mechanism

Accurate prediction of solar irradiance is crucial to fully utilize solar energy. Solar irradiance is often not accurately predicted due to temperature, humidity, seasonal conditions, and other factors. In order to improve the accuracy of solar irradiance prediction, we propose a solar irradiance prediction model that combines Echo State Network(ESN), Temporal Convolutional Networks(TCN) and the attention mechanism. ESN has strong nonlinear feature extraction and modeling capabilities by virtue of its unique reserve pool. TCN performs well in time series forecasting due to its parallel structure and larger acceptance domain. However, for the prediction process of long time series, Echo state networks and TCNs lack consideration of attention to critical information, and some non-critical information may affect the final prediction results. Therefore, in this paper, we propose to combine ESN, attention mechanism, and TCN network and present a new solar irradiance prediction model ESN-ATCN. The introduction of the attention mechanism allows the model to pay more attention to key information, improving the accuracy of the predictions.Experiments are conducted on real solar irradiance data, and the results demonstrate the accuracy of the prediction of the proposed model, which outperforms the ESN, TCN, DeepESN and LSTM.