Distributed photovoltaic power prediction based on Solar-LLM

Distributed photovoltaic power generation has volatility and intermittently, and its power generation is usually difficult to predict accurately. Previous studies have mostly focused on physical or mathematical modeling methods, and it is difficult to grasp the complexity and variability of historical data, and the prediction accuracy is limited. This paper presents a distributed PV power prediction method Solar-LLM, firstly, introduced the technical principles of the model, and then introduced the Solar-LLM model architecture and multistep photovoltaic power prediction method, the large language model, by reprogramming the input and output layer, can effectively consider the complexity and versatility of historical data. Finally, the performance of the model is demonstrated by example analysis and compared with existing methods. The experiment is based on the datasets of five photovoltaic power stations in Hebei Province, China, and the results show that the performance of Solar-LLM is better than that of GRU, LSTM, TCN, Transformer and Informer in different prediction periods, which is a feasible and effective method for power generation prediction.