Research on Short-term Prediction Algorithm for Distributed Photovoltaic Power in Regional Distribution Networks

In response to the current lack of intelligent prediction capabilities for distributed new energy in energy management terminals in regional distribution networks, this article relies on the deployment of new energy regulation terminals equipped with artificial intelligence Falcon A200 chips in microgrid distribution radio areas, and introduces Gated Recurrent Unit (GRU) network models to achieve end side ultra short time photovoltaic power generation prediction, And based on this, a combination model with better predictive metrics is proposed by combining the Convolutional Neural Networks (CNN) model optimized with dual attention modules. Due to the deployment of this prediction model on the distribution side, there are certain limitations on the model size. Based on this reason, lightweight convolution is introduced to design the model in a lightweight manner, resulting in a lightweight distributed photovoltaic power prediction algorithm model.