能源互联网下基于HS-Elman的光伏出力预测研究

摘要

Based on elabrating the concept of energy internet, this paper puts forward an improved Elman neural network (HS-Elman) forecasting model for photovoltaic power, considering different weather types, aiming to the photovoltaic output prediction problem. This paper analyzes the impacts of weather types, ambient temperature, air humidity, wind speed and irradiance on photovoltaic output prediction, and optimizes the model parameters, such as weights and thresholds, using harmony search algorithm. The prediction model is trained and tested based on photovoltaic historical data from a university in Shanghai Energy Internet platform.The results show that the model based on the HS-Elman intelligent algorithm can meet the standard of the photovoltaic output forecast, and the advanced model has a faster speed and better prediction precision under the different types of weather,compared with the traditional Elman neural network. Effectiveness and practicability can be improved and tested by the results.%文章在阐述能源互联网概念的基础上,针对光伏出力的预测问题,提出了一种基于天气类型的改进Elman神经网络(HS-Elman)光伏出力预测模型.首先,分析了天气类型、环境温度、空气湿度、风速、太阳辐照度等对光伏出力的影响;然后,利用和声搜索算法对预测模型的权值和阈值等进行优化;最后,利用上海某能源网实验平台的历史数据,对所提出的预测模型进行验证.分析结果表明:基于HS-Elman的光伏出力预测模型的预测结果能够达到光伏出力的预测标准;与传统的Elman神经网络相比,在不同的天气类型条件下,文章所提出的预测模型具有更优的运算速度和预测精度.