Study on Complex-LSTM Photovoltaic Output Prediction Model Considering Reactive Power Distribution

Accurate photovoltaic power generation prediction is key to daily grid dispatch management and secure, stable operation. The reactive power control method of photovoltaic power stations is related to the system operation status, which indirectly affects the relationship between reactive and active power outputs. Therefore, reactive information can be used to further improve the accuracy of photovoltaic active power output prediction. This paper proposes a distributed photovoltaic active-reactive power collaborative prediction method based on Complex Long Short-Term Memory networks (C-LSTM). First, the active power and reactive power are combined into a complex number, which then serves as an input to the C-LSTM. N ext, a photovoltaic power output prediction model based on complex convolution layers and Long Short-Term Memory networks is established, demonstrating stronger temporal feature extraction capabilities. Finally, a variety of methods are used to process and predict photovoltaic power using simulation verification. The results show that this method can improve the prediction accuracy of photovoltaic power output to a certain extent.

I. Introduction

As traditional fossil energy resources are gradually depleted and environmental quality continues to decline, the proportion of global thermal power generation has been gradually decreasing, while renewable energies such as wind and solar power have been rapidly developing. Among these new energy sources, solar energy, which is abundant and low-cost, has led to the widespread promotion and application of photovoltaic power generation worldwide. However, due to the impacts of geographical environment, meteorological elements, and equipment characteristics, photovoltaic power generation exhibits randomness, volatility, and counter-peak characteristics, which bring many adverse effects to the normal operation and planning of the power system. Therefore, improving the prediction accuracy of photovoltaic power generation has become an urgent matter.