This paper studies and predicts the number growth of China＇s mobile users by using the power-law regression. We find that the number growth of the mobile users follows a power law. Motivated by the data on the evolut...This paper studies and predicts the number growth of China＇s mobile users by using the power-law regression. We find that the number growth of the mobile users follows a power law. Motivated by the data on the evolution of the mobile users, we consider scenarios of self-organization of accelerating growth networks into scale-free structures and propose a directed network model, in which the nodes grow following a power-law acceleration. The expressions for the transient and the stationary average degree distributions are obtained by using the Poisson process. This result shows that the model generates appropriate power-law connectivity distributions. Therefore, we find a power-law acceleration invariance of the scale-free networks. The numerical simulations of the models agree with the analytical results well.展开更多
The Brazilian electric sector reform established that the remuneration of distribution utilities must be through the management of their systems. This fact increased the necessity of control and management of load flo...The Brazilian electric sector reform established that the remuneration of distribution utilities must be through the management of their systems. This fact increased the necessity of control and management of load flows through the connection points between the distribution systems and the basic grid as a function of the contracted amounts. The objective of this control is to avoid that these flows exceed some thresholds along the contracted values, avoiding monetary penalties to the utility or unnecessary amounts of contracted flows that overrates the costumers. This question highlights the necessity of forecast the flows in these connection points in sufficient time to permit the operator to take decisions to avoid flows beyond the contracted ones. In this context, this work presents the development of a neural network based load flow forecaster, being tested two time-series neural models： support vector machines and Bayesian inference applied to multilayered perceptron. The models are applied to real data from a Brazilian distribution utility.展开更多
传统时序预测模型通常仅关注捕捉复杂时序中的趋势和模式,而忽略了变量间的相互作用,限制了该模型在复杂时序预测中应用.提出一种Dualformer双模型并联方案,该模型并联iTransformer(inverted transformer)和PatchTST(patch time series ...传统时序预测模型通常仅关注捕捉复杂时序中的趋势和模式,而忽略了变量间的相互作用,限制了该模型在复杂时序预测中应用.提出一种Dualformer双模型并联方案,该模型并联iTransformer(inverted transformer)和PatchTST(patch time series transformer),通过激活函数替代前馈神经网络,并通过多层感知机计算输出结果.Dualformer利用注意力机制同时捕捉复杂时序中的时间维度和变量维度信息,关注时间趋势与多变量交互.实验结果显示,Dualformer在复杂时序预测效果上显著优于对比模型iTransformer、PatchTST和DLinear(decomposition linear),在实际应用中可显著提高复杂时序预测的准确度,具有广泛应用前景.展开更多
基于高分辨率快速更新无缝隙融合集成预报RISE系统(Rapid-refresh Integrated Seamless Ensemble system),采用华北3 km分辨率快速循环更新的中尺度数值模式CMA-BJ、欧洲中心0.125°分辨率全球数值模式ECMWF、常规自动站和冬奥赛道...基于高分辨率快速更新无缝隙融合集成预报RISE系统(Rapid-refresh Integrated Seamless Ensemble system),采用华北3 km分辨率快速循环更新的中尺度数值模式CMA-BJ、欧洲中心0.125°分辨率全球数值模式ECMWF、常规自动站和冬奥赛道加密自动站逐时观测资料,以北京冬奥会复杂山地为研究区域,对比分析了不同模式背景场对100 m网格分辨率的地面2 m温度和10 m风场融合分析场和1~24 h逐小时间隔预报准确性的影响。对比试验结果表明:(1)采用区域模式和全球模式的预报数据作为RISE系统背景场,均可有效形成复杂山地百米级精细化融合产品,但不同模式背景场对不同气象要素分析和预报性能的影响存在明显差异;(2)对于温度分析场,以CAM-BJ和ECMWF模式的预报数据为背景场的RISE温度分析场空间分布基本一致,分析平均绝对误差(MAE)均小于0.2℃;(3)对于风场分析场,采用高分辨率区域模式比粗分辨率全球模式更能提升RISE高精度风场融合产品精细化水平;(4)对于温度预报,以ECMWF模式的预报数据为背景场的RISE格点融合预报性能显著优于CMA-BJ模式的预报数据为背景场,冬奥高山站和所有站平均预报MAE分别减小10.5%和7.0%;(5)对于风场预报,以CAM-BJ和ECMWF模式的预报数据为背景场的RISE冬奥高山站临近1~6 h风速预报MAE分别为1.42 m s^(-1)和1.30 m s^(-1),7~24 h预报MAE则分别为1.52 m s^(-1)和1.54 m s^(-1),而RISE区域内所有站1~24 h平均MAE分别为1.38 m s^(-1)和1.24 m s^(-1)。研究成果有助于深入理解模式背景场在百米级融合预报中的作用,对提升复杂地形下天气预报准确性有重要的科学意义和业务应用价值。展开更多
针对城市电网电压受负荷影响较大、中长期尖峰负荷难以预测等问题,提出一种基于蚁群优化双向长短期记忆神经网络的中长期电网尖峰负荷预测方法。解决了传统双向长短期记忆(bi-directional long short-term memory,BiLSTM)网络处理时间...针对城市电网电压受负荷影响较大、中长期尖峰负荷难以预测等问题,提出一种基于蚁群优化双向长短期记忆神经网络的中长期电网尖峰负荷预测方法。解决了传统双向长短期记忆(bi-directional long short-term memory,BiLSTM)网络处理时间序列数据时的局限性,并利用蚁群算法(ant colony optimization,ACO)在优化神经网络参数上的潜力来提高预测的准确性和可靠性;通过融合ACO和BiLSTM模型,实现自动调整网络参数,寻求最优解决方案,不仅提升了学习效率,还增强了对电网尖峰负荷特性的捕捉能力。通过对某地区电网数据进行测试,所提方法的预测时间和预测精度均超过传统BiLSTM。分析了蚁群优化策略对预测性能的影响,并探讨了预测不同时间尖峰负荷的表现。研究结果表明:利用ACO能有效辅助BiLSTM模型捕捉复杂时间序列特征,在中长期尖峰负荷预测中,显著提升了性能,模型的R2达到0.99,且平均绝对误差降低了7.8%。展开更多
为满足第十四届全国冬季运动会(简称“十四冬”)对复杂地形高精度天气预报服务的需求,结合大涡模拟技术及高精度地形数据,搭建了针对“十四冬”室外赛区的内蒙古次百米级数值预报系统(the Sub-hundred-meter Numerical Forecast System ...为满足第十四届全国冬季运动会(简称“十四冬”)对复杂地形高精度天气预报服务的需求,结合大涡模拟技术及高精度地形数据,搭建了针对“十四冬”室外赛区的内蒙古次百米级数值预报系统(the Sub-hundred-meter Numerical Forecast System of Inner Mongolia,SNFS)。以赛区站点观测资料为参考,根据风速预报检验指标及检验评分等参数,对SNFS在“十四冬”3个赛区的风场预报性能进行检验评估。结果表明:扎兰屯及喀喇沁赛区预报风速总体表现为系统性高估,凉城赛区为低估。对于海拔高度较低的测站,预报与观测风速偏差均以正偏差为主,随测站海拔高度增加,风速偏差转为负偏差,偏差离散程度增大。扎兰屯及喀喇沁赛区在各预报时效内都存在对小于6级实测风预报偏强,而对大于6级实测风预报偏弱的特点,凉城赛区则预报均偏弱,其中扎兰屯赛区对大于6级的实测风预报评分优势明显。进一步分析扎兰屯赛区一次大风天气个例表明,预报风向与观测风向基本吻合,风速变化趋势一致,SNFS可以刻画出局地流场特征,对复杂地形下的风场具有较好的预报能力。展开更多
基金supported by the National Natural Science Foundation of China(Grant No.70871082)the Shanghai Leading Academic Discipline Project,China(Grant No.S30504)
文摘This paper studies and predicts the number growth of China＇s mobile users by using the power-law regression. We find that the number growth of the mobile users follows a power law. Motivated by the data on the evolution of the mobile users, we consider scenarios of self-organization of accelerating growth networks into scale-free structures and propose a directed network model, in which the nodes grow following a power-law acceleration. The expressions for the transient and the stationary average degree distributions are obtained by using the Poisson process. This result shows that the model generates appropriate power-law connectivity distributions. Therefore, we find a power-law acceleration invariance of the scale-free networks. The numerical simulations of the models agree with the analytical results well.
文摘The Brazilian electric sector reform established that the remuneration of distribution utilities must be through the management of their systems. This fact increased the necessity of control and management of load flows through the connection points between the distribution systems and the basic grid as a function of the contracted amounts. The objective of this control is to avoid that these flows exceed some thresholds along the contracted values, avoiding monetary penalties to the utility or unnecessary amounts of contracted flows that overrates the costumers. This question highlights the necessity of forecast the flows in these connection points in sufficient time to permit the operator to take decisions to avoid flows beyond the contracted ones. In this context, this work presents the development of a neural network based load flow forecaster, being tested two time-series neural models： support vector machines and Bayesian inference applied to multilayered perceptron. The models are applied to real data from a Brazilian distribution utility.
文摘传统时序预测模型通常仅关注捕捉复杂时序中的趋势和模式,而忽略了变量间的相互作用,限制了该模型在复杂时序预测中应用.提出一种Dualformer双模型并联方案,该模型并联iTransformer(inverted transformer)和PatchTST(patch time series transformer),通过激活函数替代前馈神经网络,并通过多层感知机计算输出结果.Dualformer利用注意力机制同时捕捉复杂时序中的时间维度和变量维度信息,关注时间趋势与多变量交互.实验结果显示,Dualformer在复杂时序预测效果上显著优于对比模型iTransformer、PatchTST和DLinear(decomposition linear),在实际应用中可显著提高复杂时序预测的准确度,具有广泛应用前景.
文摘基于高分辨率快速更新无缝隙融合集成预报RISE系统(Rapid-refresh Integrated Seamless Ensemble system),采用华北3 km分辨率快速循环更新的中尺度数值模式CMA-BJ、欧洲中心0.125°分辨率全球数值模式ECMWF、常规自动站和冬奥赛道加密自动站逐时观测资料,以北京冬奥会复杂山地为研究区域,对比分析了不同模式背景场对100 m网格分辨率的地面2 m温度和10 m风场融合分析场和1~24 h逐小时间隔预报准确性的影响。对比试验结果表明:(1)采用区域模式和全球模式的预报数据作为RISE系统背景场,均可有效形成复杂山地百米级精细化融合产品,但不同模式背景场对不同气象要素分析和预报性能的影响存在明显差异;(2)对于温度分析场,以CAM-BJ和ECMWF模式的预报数据为背景场的RISE温度分析场空间分布基本一致,分析平均绝对误差(MAE)均小于0.2℃;(3)对于风场分析场,采用高分辨率区域模式比粗分辨率全球模式更能提升RISE高精度风场融合产品精细化水平;(4)对于温度预报,以ECMWF模式的预报数据为背景场的RISE格点融合预报性能显著优于CMA-BJ模式的预报数据为背景场,冬奥高山站和所有站平均预报MAE分别减小10.5%和7.0%;(5)对于风场预报,以CAM-BJ和ECMWF模式的预报数据为背景场的RISE冬奥高山站临近1~6 h风速预报MAE分别为1.42 m s^(-1)和1.30 m s^(-1),7~24 h预报MAE则分别为1.52 m s^(-1)和1.54 m s^(-1),而RISE区域内所有站1~24 h平均MAE分别为1.38 m s^(-1)和1.24 m s^(-1)。研究成果有助于深入理解模式背景场在百米级融合预报中的作用,对提升复杂地形下天气预报准确性有重要的科学意义和业务应用价值。
文摘为满足第十四届全国冬季运动会(简称“十四冬”)对复杂地形高精度天气预报服务的需求,结合大涡模拟技术及高精度地形数据,搭建了针对“十四冬”室外赛区的内蒙古次百米级数值预报系统(the Sub-hundred-meter Numerical Forecast System of Inner Mongolia,SNFS)。以赛区站点观测资料为参考,根据风速预报检验指标及检验评分等参数,对SNFS在“十四冬”3个赛区的风场预报性能进行检验评估。结果表明:扎兰屯及喀喇沁赛区预报风速总体表现为系统性高估,凉城赛区为低估。对于海拔高度较低的测站,预报与观测风速偏差均以正偏差为主,随测站海拔高度增加,风速偏差转为负偏差,偏差离散程度增大。扎兰屯及喀喇沁赛区在各预报时效内都存在对小于6级实测风预报偏强,而对大于6级实测风预报偏弱的特点,凉城赛区则预报均偏弱,其中扎兰屯赛区对大于6级的实测风预报评分优势明显。进一步分析扎兰屯赛区一次大风天气个例表明,预报风向与观测风向基本吻合,风速变化趋势一致,SNFS可以刻画出局地流场特征,对复杂地形下的风场具有较好的预报能力。
