基于数据处理和深度学习的光伏发电预测模型

Photovoltaic Power Generation Prediction Model Based on DataProcessing and Deep Learning

目前,光伏发电已成为最受欢迎的可再生能源形式之一,然而光伏的接入给电力系统的运行带来了许多挑战,一个准确可靠的光伏发电预测模型将有效提升系统安全和降低系统运行成本.将数据处理与深度学习模型相结合,提出了一种完整的光伏发电预测模型,其中包括数据预处理、特征工程、核主成分分析、XGBoost以及数据后处理等.此模型以历史光伏功率、太阳辐照度和数值天气预报数据作为输入数据集.案例测试以4个光伏电站的测量功率为例,利用归一化均方根误差(NRMSE)和归一化平均绝对百分比误差(NMAPE)来评估预测模型的性能.测试结果表明,与其他深度学习模型相比,所提出的预测模型在准确率上表现出了优越的性能.

At present,photovoltaic power generation has become one of the most popular forms of renewable energy.However,the photovoltaic access has brought many challenges to the operation of the power system.An accurate and reliable photovoltaic power generation prediction model will effectively improve the system security and reduce the system operation cost.Combining data processing with deep learning model,this paper proposes a complete photovoltaic power generation prediction model,including data preprocessing,feature engineering,kernel principal component analysis,XGBoost,and data post-processing.This model takes historical photovoltaic power,solar irradiance and numerical weather forecast data as input data sets.Taking the measured power of four photovoltaic power stations as an example,the performance of the prediction model is evaluated by using normalized root mean square error(NRMSE)and normalized mean absolute percentage error(NMAPE).The test results show that compared with other deep learning models,the proposed prediction model shows superior performance in accuracy.