摘要
在全球气候变化加剧、新能源大规模快速发展背景下,气象已成为影响电网安全和电力保供的关键因素之一。提出一种台风灾害不确定性场景下双层循环风光功率预测模型。首先,上层利用Wasserstein生成对抗网络(Wasserstein generative adversarial network, WGAN)模型建立不确定性场景处理模型,通过WGAN与改进的K-means算法结合,得到各不确定性变量典型场景,实现对可再生能源不确定性的合理刻画。然后,下层建立基于多模型融合Stacking集成学习模型及长短期记忆(long short term memory, LSTM)模型的台风灾害风光功率预测模型,并以预测准确率为迭代参数,通过上下层模型之间的循环迭代,得出准确率最高的风光功率预测值。最后,以2022年登陆浙江舟山港的台风“梅花”为例,验证了所提模型的先进性与可行性。
In the context of global climate change intensification,large-scale and rapid development of new energy,meteorological conditions have become one of the key factors affecting power grid security and power supply.A wind and photovoltaic generation prediction bi-level model based on uncertainty scenarios under typhoon disaster model is proposed.First,the upper layer with Wasserstein generative adversarial network of uncertainty processing model is established.Through the combination of Wasserstein generative adversarial network model and improved K-means clustering,typical scenarios of uncertainties are established to realize the reasonable optimize of renewable energy uncertainty.Second,the lower layer with wind and photovoltaic generation prediction model based on typhoon disaster is established.It has used the Stacking and long short-term memory.Based on the loop iteration between the upper layer and the lower layer,the wind and photovoltaic generation prediction value with the highest accuracy are obtained.Finally,the 2022 super typhoon"Muifa"in Zhoushan port,Zhejiang,is taken as an example.The results verify that the optimization results have an impact on the prediction results and the advanced nature of the proposed model.
作者
侯慧
万屹
王振国
姜凯华
王少华
李特
李正天
林湘宁
HOU Hui;WAN Yi;WANG Zhenguo;JIANG Kaihua;WANG Shaohua;LI Te;LI Zhengtian;LIN Xiangning(School of Automation,Wuhan University of Technology,Wuhan 430070,China;State Grid Zhejiang Electric Power Research Institute,Hangzhou 310014,China;State Key Laboratory of Advanced Electromagnetic Technology,Huazhong University of Science and Technology,Wuhan 430074,China)
出处
《电力建设》
北大核心
2025年第3期146-154,共9页
Electric Power Construction
基金
国家自然科学基金项目(U22B20106)
国家电网公司科学技术项目(5200-202319382A-2-3-XG)
浙江省自然科学基金项目(LZJMY25D050006)
国网浙江省电力有限公司科技项目(B311DS24001A)。
关键词
台风灾害
风光功率预测
不确定性
双层模型
循环迭代
typhoon disaster
wind and photovoltaic generation prediction
uncertainty
bi-level model
loop iteration
