基于蜻蜓-高斯过程回归耦合的居民社区时用水量动态实时区间预测  

Dynamic Real-time Interval Prediction of Residential Community Water Consumption Based on Dragonfly Gauss Process Regression Coupling

作　　者：刘龙龙[1,2] 李文竹[1] 刘心[1] LIU Long-long;LI Wen-zhu;LIU Xin(Hebei University of Engineering, Handan 056038,Hebei Province, China;Handan River Affairs Bureau,Handan 0560001, Hebei Province, China)

机构地区：[1]河北工程大学,河北邯郸056038 [2]邯郸河务局,河北邯郸056001

出　　处：《中国农村水利水电》2019年第4期89-92,95,共5页China Rural Water and Hydropower

基　　金：国家自然科学基金项目(61440001);教育部新世纪优秀人才支持计划项目(NCET-13-0770);河北省高等学校高层次人才科学研究项目(GCC2014062)

摘　　要：准确的居民社区时用水量预测对于水资源的优化调度和管网的稳定运行具有重要意义。为了克服传统的ANN、SVM等预测方法只能得到确定的点预测结果和未来某一时刻的预测结果,而无法给出预测的区间,也不能进行实时预测的缺点,提出了一种基于蜻蜓-高斯过程回归耦合的居民社区时用水量动态实时区间预测方法。用擅长全局搜索和对初始值没有依赖的蜻蜓算法对高斯过程回归的超参数进行寻优。结果表明,本模型预测精度较高,最大的相对误差为仅0.019,具有较强的实用价值,为未来水资源实时调度提供理论依据。Accurate prediction of residential community water consumption is of great significance to the optimal operation of water resources and the stable operation of the network. In this paper, a kind of water dynamic real time based on the dragonfly Gauss process regression coupling is proposed. dragonfly algorithm which is good at global search is used and the super parameters of Gauss process regression are optimized. The results show that the prediction accuracy of this model is high and the maximum relative error is only 0.019. It has a strong practical value and provides a theoretical basis for real-time water resource scheduling in the future.

关 键 词：蜻蜓算法 高斯过程回归 时用水量 区间预测 

分 类 号：TV213.4[水利工程—水文学及水资源]