Distributed PV Power Forecast Based on Public Meteorological Data

In the context of the large-scale deployment of distributed PV power stations, ensuring the accuracy of PV power forecasting is particularly important for the safe and stable operation of the power grid. Therefore, this paper proposes a distributed PV short-term power prediction model based on public meteorological data. Firstly, the time resolution of meteorological data is improved using cubic spline interpolation. Then, in the absence of irradiance data in public meteorological data, convolutional neural networks (CNN) are used to extract features from PV output data and limited meteorological data, and combined with long short-term memory (LSTM) for further learning, which can deeply explore the relationship between meteorological data and distributed PV power output. Finally, the model is tuned using Bayesian optimization. The prediction results show that the CNN-LSTM combination model established in this paper has better prediction performance than using only CNN or LSTM for prediction, and its prediction results are closer to the actual distributed PV power output.