摘要
以风能和太阳能为代表的新能源具有随机性、间歇性和波动性,对新能源发电功率进行预测是有效解决以上问题的途径。为解决可再生能源(Renewable Energies,REs)在电力网络中不断整合对电力供应系统去中心化带来的问题,提出了一种基于神经计算的预测模型。首先,获取不同的数据集,并介绍了数据使用方法;其次,对不同的人工神经网络(Artificial Neural Network, ANN)架构进行了深入研究,同时探讨了不同的输入特征,以及对预测结果不确定性的评估方法;最后,通过模拟实验表明在光伏发电方面,本模型的预测相对平均绝对误差(Relative Mean Absolute Error,RMAE)可以在夏季达到2.5%,在冬季达到0.5%。而在整个年度内,通过使用输入类别特征的减小均值,光伏和风力发电的RMAE分别可以达到1.7%和4.9%。这一方法的实施为地区可再生能源的集成提供了有力支持,同时也为应对可再生能源的波动性和不可控性提供了新的解决思路。
The new energy represented by wind energy and solar energy is random,intermittent and fluctuating.The prediction of new energy power generation is an effective way to solve the above problems.In order to solve the problem of decentralization of power supply system caused by the continuous integration of renewable energies(REs)in power network,this paper proposes a prediction model based on neural computing.Firstly,different data sets are obtained and the data usage method is introduced.Secondly,different artificial neural network(ANN)architectures are studied in depth,and different input characteristics and evaluation methods for uncertainty of prediction results are discussed.Finally,through simulation experiments,it is shown that in terms of photovoltaic power generation,the relative mean absolute error(RMAE)of the prediction of this model can reach 2.5%in summer and 0.5%in winter.In the whole year,the RMAE of photovoltaic and wind power can reach 1.7%and 4.9%respectively by using the reduced mean of input category features.The implementation of this method provides strong support for the integration of regional renewable energy,and also provides a new solution to deal with the volatility and uncontrollability of renewable energy.
作者
杨强
颜宗辉
杜秀举
吴胜安
牟景艳
李培霞
YANG Qiang;YAN Zonghui;DU Xiuju;WU Shengan;MU Jingyan;LI Peixia(Anshun Power Supply Bureau of Guizhou Power Grid Co.,Ltd.,Anshun 561000,China;Dongfang Electronics Co.,Ltd.,Yantai 264000,China)
出处
《电气应用》
2024年第2期76-82,共7页
Electrotechnical Application
基金
中国南方电网有限责任公司科技项目(GZKJXM20220060)。
关键词
可再生能源
人工神经网络
预测模型
平均绝对误差
renewable energys
artificial neural network
prediction model
relative mean absolute error
