基于深度学习模型的光伏发电负荷预测

Photovoltaic Load Forecasting Based on Deep Learning Model

光伏发电作为分布式能源不断普及,对电力基础设施稳定构成了一系列挑战。本文提出了一个有监督的深度学习模型用于光伏发电预测,该模型利用数值天气预报和高分辨率历史测量来预测时间间隔内的联合概率分布,而不用预测变量的预期值,与常见的基线方法——如完全连接的神经网络和长期短期记忆体系结构相比,这种设计提供了显著的性能改善,使用归一化基于平均平方误差的预测技能得分作为关键绩效指标,将提出的方法与其他模型进行比较,结果表明,新设计的性能高于目前的光伏电源预测技术水平。

The continuous popularity of photovoltaic power generation as a distributed energy source poses a series of challenges to the stability of power infrastructure.This article proposes a supervised deep learning model for photovoltaic power generation prediction.The model uses Numerical weather prediction and high-resolution historical measurement to predict the joint probability distribution within the time interval,rather than the expected value of the prediction variable.Compared with common baseline methods such as fully connected neural network and long-term Short-term memory architecture,this design provides significant performance improvement.The normalized prediction skill score based on mean square error is used as the Performance indicator,Comparing the proposed method with other models,the results show that the performance of the new design is higher than the current level of photovoltaic power prediction technology.