In order to improve the accuracy of distributed photovoltaic output power prediction and enhance the safety and reliability of distribution network operation, a distributed photovoltaic output power prediction model based on Pelican algorithm, variational modal decomposition and deep learning is proposed. Firstly, the weather conditions are clustered into three kinds of fusion weather: sunny weather, cloudy weather and rain and snow. Then, the Pelican Optimization Algorithm (POA) is used to optimize the Variational Mode Decomposition (VMD) algorithm, and POA is used to adaptively determine the optimal parameter combination (k,a) in VMD, so as to adaptively decompose the original data sequence of photovoltaic power generation and reduce data noise. Finally, the Long Short-Term Memory (LSTM) model is used to predict each sub-mode component and the predicted values of each sub-mode component are superimposed to obtain the final photovoltaic power prediction result. The 2022 power data of a 50kW distributed photovoltaic power plant is selected as a sample for case study, and the results show that the model improves the photovoltaic output power prediction accuracy, which can provide valuable reference for the distributed photovoltaic output power prediction research in rural fuzzy meteorological areas.