NAR神经网络的应用与检验——以城市居民生活需水定额为例

发布时间：2018-09-03 13:42

【摘要】：NAR神经网络具有反馈和记忆功能,其在时间序列的建模仿真方面具有显著优点。以城市居民生活需水定额为例,采用NAR神经网络建立了贵州省城市居民生活需水定额的时间序列模型,通过试验法、留一法交叉检验讨论了模型相关输入参数的计算与选取,通过相关系数、Nash效率系数、LBQ检验、ROC曲线方法检验了模型的性能和预测结果的精度,进而对贵州省城市居民生活需水定额变化趋势进行了预测。结果表明,(1)NAR模型性能良好并具有较高的预测精度,NAR神经网络的相关系数r、Nash效率系数分别达到0.97、0.87,LBQ检验得出预测结果误差不存在自相关性,采用预测结果绘制ROC曲线,其AUC值达到0.938(处于水平1,有较高准确性);(2)需水定额合理性评价中,预测2020年、2030年需水定额分别为137.72 L/(人·d)、132.94 L/(人·d),满足《室外给水设计规范》(GB50013—2006)的要求,具有较好的适用性。
[Abstract]:NAR neural network has the function of feedback and memory, and it has remarkable advantages in modeling and simulation of time series. Taking the urban residents' living water demand quota as an example, the time series model of the urban residents' water demand quota in Guizhou province is established by using NAR neural network. The calculation and selection of the relevant input parameters of the model are discussed by using the test method and the left one method. The performance of the model and the accuracy of the prediction results were tested by the correlation coefficient and Nash efficiency coefficient and the ROC curve method, and the trend of the change of the water demand quota for the urban residents in Guizhou Province was predicted. The results show that: (1) the correlation coefficient r-Nash efficiency coefficient of the NAR neural network with good performance and high prediction accuracy is 0.97 ~ 0.87%. The prediction error is not autocorrelation. The prediction results are used to draw the ROC curve. Its AUC value reached 0.938 (at level 1, with high accuracy); (2) water demand ration rationality evaluation, forecast 2020, 2030 water demand quota is 137.72 L / (person d) 132.94 L / (person d), meets the requirement of "Outdoor Water supply Design Code" (GB50013-2006), it has better applicability.
【作者单位】：贵州省水利水电勘测设计研究院;贵州省喀斯特地区水资源开发利用工程技术研究中心;武汉大学水资源与水电工程科学国家重点实验室;
【基金】：国家自然科学基金面上项目(51479140) 贵州省科技计划项目(黔科合SY字[2015]3006;黔科合[2016]支撑2903;黔科合重大专项字[2012]6013号
【分类号】：TV213.4

【相似文献】