Details

Autor(en) / Beteiligte

• Cheng, Tan
• Yichuan, Xu
• Hui, Liu
• Jie, Gu

Titel

Comparison Method for Photovoltaic Power Prediction Effect Based on LSTM in Multiple Scenarios

Ist Teil von

• 2023 4th International Conference on Advanced Electrical and Energy Systems (AEES), 2023, p.743-479

Ort / Verlag

IEEE

Erscheinungsjahr

2023

Quelle

IEEE Electronic Library (IEL)

Beschreibungen/Notizen

• The high penetration of photovoltaics introduces significant instability and randomness to the distribution grid. Understanding photovoltaic power generation is of paramount importance for enhancing grid stability. This paper first constructs a physical model for photovoltaic power prediction and analyzes the performance of this indirect prediction method. Considering the limitations of indirect prediction methods in practical applications, we then discuss direct prediction methods based on artificial intelligence. We primarily employ Long Short-Term Memory (LSTM) for short-term photovoltaic power forecasting. Additionally, we select feature variables using the Pearson correlation coefficient and perform visual analysis on two specific time periods: the summer solstice and the winter solstice. This analysis compares forecasted values with actual values to assess the prediction accuracy. Furthermore, we discuss parameter selection and compare the predictive performance of LSTM with other models. The results indicate that LSTM demonstrates significant advantages in photovoltaic power prediction.