ARIMA模型与BP神经网络模型的比较研究——基于供电企业电费现金流入预测

文章基于供电企业对电费现金流入预测的现实需要,分别建立了基于时间序列的ARIMA模型和BP神经网络预测模型,并对两种预测方法进行了对比分析,最终确定了以ARIMA模型为主、BP神经网络为辅的综合预测手段,有效地提高了电费现金流的预测精度,增强了供电企业的现金流管理水平。     

基于BP神经网络改进算法在地铁隧道施工中的沉降预测

为使地铁隧道在施工中沉降监测数据具有一定的预见性,分别采用了BP神经网络改进算法的预测模型、传统BP神经网络预模型以及基于时间序列的三次指数平滑法预测模型对地铁隧道施工中的沉降监测数据进行了预测。对其预测结果进行分析,得出了BP神经网络改进算法模型预测精度优于传统BP神经网络模型以及基于时间序列的三次指数平滑法模型预测精度的结论。     

面向区块链节点负载预测的ARIMA组合预测方法

准确预测服务器的剩余负载率可以合理地分配系统资源,提高系统的资源使用率。为了能有效提高区块链应用中各节点剩余负载率序列预测的准确度,提出了一种差分自回归移动平均(AutoregressiveIntegrated MovingAverage,ARIMA)模型、BP神经网络以及局部异常因子(Local Outlier Factor, LOF)算法的组合预测模型。对比ARIMA模型、ARIMA-BP模型、LOFARIMA-BP模型的预测结果,比较三个模型的预测能力。实验结果表明,LOF-ARIMA-BP组合模型的预测精度优于ARIMA模型以及ARIMA-BP模型。     

基于NAR动态神经网络的桥梁SHM应变预测

基于NAR动态神经网络对桥梁SHM应变进行了分析预测。分别建立了10期、25期、50期的NAR动态神经网络预测模型和ARIMA预测模型,并对比了2个模型预测的稳定性和准确性,通过上海市某座斜拉桥的SHM实测应变数据进行验证。模型结果表明,NAR动态神经网络预测模型相比于ARIMA预测模型具有更高的准确性,工程应用价值更高。     

新息自适应混合卡尔曼滤波算法构建地表沉降预测模型

为解决矿区地表沉降变形预测的问题,提高预测模型的精度,提出了基于自回归综合移动平均模型(Autoregressive Integrated Moving Average model,ARIMA)的新息自适应卡尔曼滤波(Innovation Adaptive Kalman Filter,IAKF)与组合神经网络相结合的混合预测模型。首先,针对沉降变形监测序列的非平稳性与复杂性等特点,ARIMA模型能够将原始数列平稳化,以此构建地表下沉的预测模型,并作为新息自适应卡尔曼滤波的状态方程。然后,将集合经验模态分解(Ensemble Empirical Mode Decomposition,EEMD)、粒子群优化算法(Particle Swarm Optimization,PSO)和BP神经网络结合,构建EEMD-PSO-BP神经网络的组合网络模型,将组合神经网络的沉降预测结果作为观测值引入到卡尔曼滤波观测方程中,以建立混合预测模型。最后针对噪声方差Q与R选取的问题,利用新息自适应卡尔曼滤波估计出噪声方差的协方差阵。混合预测模型能有效减小单一预测机制造成的同一性质误差的累积,将基于ARIMA的新息自适应卡尔曼滤波、EEMD-PSO-BP神经网络模型与混合滤波模型的精度进行对比,新息自适应混合卡尔曼滤波预测模型的均方根误差降低至0. 3194mm,相对百分误差降到1. 42%。实验结果表明,混合滤波模型的各项预测结果要优于传统预测模型,精度相比较传统的预测模型有较大的改善。     

基于SVM的混沌时间序列预测模型应用研究

混沌时间序列预测是混沌理论的一个重要应用领域和研究热点,目前它在信号处理、自动化控制等领域中已得到了广泛的应用。本文联系支持向量机(SVM)和混沌时间序列预测的相关理论,建立基于二者的变形序列预测模型。同时,结合具体实例从变形时间序列的混沌识别、相空间重构以及预测模型的参数优化等方面探讨了模型的具体建立过程。实验结果表明,该模型的预测精度要优于BP神经网络。     

基于改进BP神经网络的基坑支护结构数据处理

针对现有的基坑监测和预测数据处理方法的不足,对BP神经网络预测模型作了研究和改进,应用改进后的BP算法对基坑支护结构水平位移数据进行处理,并将改进的BP算法与传统算法所建立的模型应用于工程实例进行比较,结果表明,改进后的BP神经网络模型在预测精度、训练时间、稳定性等方面均优于传统BP神经网络模型。     

基于时间序列的神经网络建模及边坡位移预测

边坡系统是一个影响因素众多、非常复杂的非线性系统,使得作为边坡内部力学现象外部表现的边坡变形同样具有很强的非线性特征,而神经网络所具有的高度鲁棒性、学习和联想记忆功能及数据挖掘等特性,对诸如存在内在联系的单时间序列的边坡位移预测有着较大的优势.以此为出发点,通过对单时间序列特点的分析,构造了基于单时间序列的神经网络预测模型,并以渝黔高速公路某边坡位移实际监测数据为例进行了计算.研究结果表明,通过挖掘边坡位移序列中的隐含信息,运用单时间序列BP神经网络进行边坡位移预测是完全可行的,预测平均误差仅为2.72%,预测结果与实际情况吻合度较高.最后通过与传统灰色理论预测方法进行对比发现,该方法预测效果明显提高,预测误差平均降低了近8倍.     

中国农业银行股票价格的统计分析及预测

ARIMA模型与指数平滑法是统计应用中非常广泛的两种方法 ,他们可以用来对数据进行拟合并预测。本文对时间序列中的ARIMA与指数平滑法进行了比较,并运用这两种方法对股票收盘价格进行预测,结果显示ARIMA在近期预测中效果较好。     

基于多规则实时学习组合型BP神经网络的城市建筑能耗预测模型

本文在建立城市层面的建筑用电量预测模型时,针对常规BP神经网络模型存在的诸多缺陷,进行了多种优化:组合模型预测取均值以克服随机性;加入统计规则以剔除个别奇异预测值的影响;实时学习以提高长期预测的精度.此怂外,针对影响因素的未来趋势难以准确预测的问题,增加了月份周期变量和月份序列变量这2个变量作为网络的输入,以提高模型的精度.结果显示,该方法预测效果明显优于常规BP神经网络和时间序列等方法.     

Fluvial Flood Forecasting

This paper describes some of the latest developments in real-time fluvial flood-forecasting systems which utilise quantitative weather radar information. Rain-gauge, flow-measuring stations, weather radar and other sources of information are fully integrated by the real-time flood-forecasting system and provide flood warnings for civil protection. The systems have been commissioned in various regions within the UK and are undergoing continuous development in conjunction with the UK water industry and through close involvement with various European partners via the auspices of the European Union.     

灰色马尔柯夫预测模型

概要地介绍了灰色系统理论的研究对象，引入了ＧＭ（１，１）灰色预测模型，并将灰色预测与马尔柯夫预测方法相结合建立了一种对ＧＭ（１，１）模型进行了预测物新方法，克服两种预测法的不足，提高了预测精度。最后用一例子说明灰色马尔柯夫预测方法的应用。     

燃气负荷及其预测模型

在数理统计学和现代优化方法基础上对已有燃气负荷预测文献进行归纳，从多角度对燃气负荷预测进行分类。指出需要按燃气负荷预测的应用要求，燃气负荷变化特性等具体情况选择预测方法。综述了可有效用于燃气负荷预测的方法和模型。     

黄茨滑坡预测预报分析

在介绍了黄茨滑坡地质概况的基础上,简要介绍了滑坡预测预报理论与方法,既而引入基于位移信息的Verhulst灰色模型和日本斋藤的蠕变模型两种方法对黄茨滑坡进行了预测预报系统分析。     

Forecasting future cooling demand in London

Cooling of buildings in the UK is responsible for around 15 TWh per year of energy demand, largely powered by electricity with highly related CO₂ emissions. The Greater London Authority wished to understand the potential impact of London's growing need for cooling on UK CO₂ emissions in the period up to 2030. This paper describes a model developed to analyse the cooling requirements for London's key building stock and assess how these would be affected by change in system mix, improvements in system efficiencies, and by varying degrees of climate change.The analysis showed that, if left unchecked, the growth in active cooling systems in London could lead to a doubling of CO₂ emissions from this source by 2030. This growth will be due to increase in building stock, increase in market share of cooling systems, and climate change. The last of these is difficult to predict, but by itself could add 260,000-360,000 tonnes of CO₂ emissions by 2030. This increase can be strongly mitigated, or even offset, by improvements in system efficiency. The difference between no efficiency improvements, and an assumed 1-3% annual efficiency improvement is around 340,000 tonnes by 2030.     

Neural Network-Based Long-Term Hydropower Forecasting System

Artificial neural networks are alternatives to stochastic models even if the optimization of their architectures remains a tricky problem. Two different approaches in long-term forecasting of potential energy inflows using a feedforward neural network (FNN) and a recurrent neural network (RNN) are proposed. The problem of overfitting, particularly critical for limited hydrologic data records, is addressed using a new approach entitled optimal weight estimate procedure (OWEP). The efficiency of the two models using OWEP is assessed through multistep forecasts. The experiment results show that, in general, OWEP improves the models' performance and significantly reduces the training time on the order of 60 percent. The RNN outperforms the FNN but costs about a factor of 2 longer in training time. Furthermore, the neural network-based models provide more accurate forecasts than traditional stochastic models. Overall, the RNN appears to be the best suited for potential energy inflows forecasting and therefore for hydropower systems management and planning.     

Real-Time Flood Forecasting Using Neural Networks

Real-time forecasting of stream flows during storms provides an essential input to operational flood management. This work is usually very complex owing to the uncertain and unpredictable nature of the underlying phenomena. The technique of neural networks therefore was applied to model it. Forecasting of flood values during storms with a lead time of one and more hours was made using a selected sequence of past flood values observed at a specific location. Training of the network was done with the help of three alternative methods, viz., error backpropagation, conjugate gradient, and cascade correlation. Resulting flood forecasts were found to be satisfactory—especially when warning time was the least.     

岩村滑坡预测预报分析

在介绍了岩村滑坡地质概况的基础上,采用剩余推力法计算滑坡在三种工况下（175m正常水位;175m水位＋雨季;蓄水洪水位）的稳定性系数,继而引入蒙特卡洛概率模型对岩村滑坡进行了破坏概率系统分析。     

城市污水排放量预测模型研究

采用多元线性回归和BP人工神经网络两种方法分别建立城市污水排放量的预测模型，并进行实例计算验证。通过比较分析，发现BP神经网络的非线性映射能力能够更好地反映城市污水量与多个自变量间的复杂关系，具有较高的模拟精度且应用简便。