机器学习在光伏发电功率预测中的应用分析

Application of Machine Learning in Photovoltaic Power Generation Prediction

光伏发电存在典型的间歇性、随机性、不确定性特征,准确的光伏发电功率预测模型对于保障电力平衡、优化方式安排、促进新能源消纳十分必要。机器学习通过学习已有的大量数据形成预测及判断,能够挖掘数据内在价值,在光伏发电领域取得丰富成果。给出基于机器学习的光伏发电功率预测框架,重点分析了传统机器学习模型、深度学习模型、组合集成学习模型、在线学习模型、物理数据联合驱动模型5种光伏发电预测方法,并对每种方法给出研究建议。

Photovoltaic power generation has typical characteristics of intermittency,randomness,and uncertainty.An accurate photovoltaic power generation prediction model is essential for ensuring power balance,optimizing mode arrangements,and promoting new energy consumption.Machine learning can form predictions and judgments by learning a large amount of existing data,and can explore the intrinsic value of the data,achieving rich results in the field of photovoltaic power generation.This article provides a framework for predicting photovoltaic power generation based on machine learning,with a focus on analyzing five types of photovoltaic power generation prediction methods:traditional machine learning models,deep learning models,combinatorial ensemble learning models,online learning models,and physical data joint driving models.And research recommendations are provided for each type of method.