Optimized Bidirectional Long Short-Term Memory Network for Multivariate Solar Radiation Prediction Utilizing Gold Panning Algorithm

This study is dedicated to optimising neural networks for more accurate multivariate solar radiation prediction using the Gold Rush algorithm. In the current research context, existing methods suffer from low accuracy and poor model stability. For this reason, this paper introduced artificial intelligence algorithms, especially neural networks, with a view to improving the prediction performance. This paper collected data on solar radiation and its related influencing factors from multiple data sources, and after data processing, this text used the Gold Rush algorithm to optimise the neural network. This text selected the indicators with high correlation for calculation, used the gold rush algorithm to optimise the algorithm to find the parameters and determine the optimal parameters, and then input the obtained data into the bi-directional long and short memory network for prediction. After testing and error analysis, the neural network model optimised based on the gold rush algorithm has high accuracy and stability in solar radiation prediction. However, the model also has some drawbacks, such as high computational complexity and sensitivity to parameter adjustment.