河流健康评价的回归支持向量机模型及应用

建立河流健康评价指标体系、分级标准及回归支持向量机（ SVR ）河流健康评价模型,并以云南省文山州清水河健康评价为例进行研究。首先,利用层次分析法（ AHP ）从水文水资源、物理结构、水质、水生生物和社会服务功能5个方面遴选出13个评价指标,构建3个层次的河流健康评价指标体系和5个等级的分级标准;其次,基于SVR原理,利用随机生成和随机选取的方法,在等级标准阈值间构造5种不同容量大小的训练样本和检验样本,提出5种不同容量方案的SVR河流健康评价模型,设计合理的输出模式,并构建具有良好性能的RBF（radial basis function neural network ）回归模型作为对比模型,利用模型随机5次运行的平均相对误差绝对值、最大相对误差绝对值和运行时间对各方案模型性能进行评价;最后,利用达到期望精度的SVR模型对实例进行评价分析。结果表明：①无论是训练样本还是检验样本,5种方案的SVR模型的预测精度和泛化能力均优于 RBF模型。在相同参数设置条件下,SVR模型随着样本容量的增加其精度和泛化能力变化不大;而RBF模型随着样本容量的增加其精度和泛化能力均有提高。表明SVR模型具有较高的精度和泛化能力,可以用于河流健康评价,尤其在小样本情况下,SVR模型的精度和泛化能力是RBF模型不可比拟的。②5种方案的SVR模型对清水河2011-2012年3次调查的评价结果均为健康,但已接近于亚健康。     

重庆藕塘滑坡地下水位时间序列混沌性判别与预测

在周期性库水位涨落和季节性强降雨等因素影响下,藕塘滑坡前缘地下水位时间序列呈现混沌特性。以藕塘滑坡前缘水文孔(MZK5)地下水位序列为研究对象,进行混沌分析;并利用饱和关联维数法和最大Lyapunov指数法,在水位序列相空间重构的基础上对混沌特性验证。为克服粒子群算法本身的不足和SVR模型参数选取困难的缺点,采用基于遗传选择策略的粒子群算法(GSSPSO)优化选取SVR模型的参数,再用预测性能较为精准的回归型支持向量机(SVR)模型对其进行预测。对MZK5水文孔地下水位进行了BP神经网络模型预测和优化后模型(GSSPSO-SVR)预测。结果表明:藕塘滑坡前缘地下水位序列具有混沌特征,GSSPSOSVR模型预测结果的均方根误差为0.036 m,拟合优度为0.803,说明GSSPSO-SVR模型预测效果较理想,且预测精度高于BP网络模型,具有较强的实用性。     

基于人工神经网络和支持向量回归机的干旱预测

气候变暖导致南方湿润地区干旱逐渐加剧,干旱预测对于水资源优化管理、缓解旱情具有重要意义。基于不同时间尺度的标准化降水蒸散指数(SPEI)作为干旱评价指标,通过构建人工神经网络(ANN)和支持向量回归机(SVR)模型进行了预见期为1～3个月的干旱预测,并进一步构建了经验模态分解的人工神经网络(EMD-ANN)和支持向量回归机(EMD-SVR)耦合模型,以提高1个月时间尺度SPEI1的预测精度。结果表明:对时间尺度为3个月及以上的SPEI,ANN和SVR模型具有良好预测效果,且SVR模型预测精度略优于ANN模型;预见期越短,干旱预测精度越高,预见期1个月的ANN和SVR模型预测决定系数可达到0.834～0.911;ANN和SVR模型对1个月时间尺度的SPEI1预测效果较差,通过EMD及小波消噪处理后,基于EMD-ANN和EMD-SVR模型的SPEI1预测精度显著提高。     

ARIMA-SVR组合方法在水质预测中的应用

针对复杂水域水质变化机理难以掌握、水质预测建模困难且预测精度低的问题,将时间序列分析方法与机器学习方法引入水质预测领域,提出了基于差分自回归移动平均(ARIMA)与支持向量回归(SVR)组合模型的水质预测方法。数据经过预处理后先由ARIMA模型对其进行线性拟合,然后通过SVR模型预测残差以补偿其中的非线性变化。选择巢湖水域2004—2015年间的pH和溶解氧监测数据作为试验样本,通过Hodrick-Prescott(HP)滤波方法分析,结果表明2组数据具有不同的趋势特性和波动特性。根据精度评价指标对比分析模型的预测效果,结果表明组合模型预测精度显著提高,pH和溶解氧预测值与观测值间的相关系数均达到了0.99,均方根误差分别为0.20和0.61,平均绝对百分比误差分别为2.2%和6.6%。本研究所建立的组合预测方法具有较高的预测精度和较强的泛化能力,适用于复杂水域的水质预测。     

基于多元变量组合的回归支持向量机集成模型及其应用

为进一步提高径流预测的精度和泛化能力,提出基于多元变量组合的回归支持向量机(SVR)集成年径流预测模型,以云南省龙潭站年均径流预测为例进行实例研究。首先,以实例1—10月月均流量作为预测因子,采用相关分析法确定预测因子与年均径流量的相关系数,按照相关系数大小顺序依次选取预测因子,构建2维输入变量~10维输入变量的9种SVR模型对实例后12年的年均径流量进行预测。最后,采用简单平均(SA)和加权平均(WA)两种集成方法对具有较高预测精度的7种SVR模型的预测结果进行综合集成。结果表明:①SVR模型的预测精度随着输入变量维数的增加明显提高。②SA-SVR和WA-SVR模型对实例后12年年均径流量预测的平均相对误差绝对值分别为1.73%和1.79%,最大相对误差绝对值分别为6.34%和6.47%,精度和泛化能力均优于各SVR模型。相对而言,由于采用多个SVR模型进行集成,SA-SVR模型预测效果略优于WASVR模型。     

基于改进SVR模型的新疆冰川河流年径流预测研究

文章结合人工蜂群算法的自适应度因子对传统SVR模型的循环求解进行改进和优化,提高传统模型收敛和求解精度,并将改进SVR模型用于新疆喀什某冰川河流的年径流预测中。研究结果表明:在AIC计算准则下,改进的SVR模型最小计算值最小,参数结果更为合理;相比于传统模型,改进模型在区域冰川河流年径流预测精度得到较为明显的改善和提高,其中预测的误差均值减少16.3%,年相关系数提高0.25;建立的自回归方程可对未来冰川径流的变化趋势进行预测。研究成果对于新疆冰川河流年径流的变化预测提供方法参考。     

基于改进支持向量机模型的开都河年径流量预测

针对回归支持向量机(SVR)惩罚因子C和核函数参数g的选取对模型性能有着关键性影响以及在实际应用中存在参数选取等困难,提出基于启发式算法的PSO-SVR和GA-SVR年径流预测模型,以新疆开都河大山口水文站年径流预测为例进行实例研究。首先,利用DPS软件选取年径流影响因子,确定输入向量;其次,基于粒子群算法(PSO)、遗传算法(GA)基本原理,采用PSO、GA优化SVR惩罚因子C和核函数参数g,构建PSO-SVR和GA-SVR年径流预测模型,并构建基于网格划分(GS)与交叉验证(CV)算法相结合的GS-SVR模型作为对比模型。最后,利用所构建的模型对实例进行预测分析。结果表明:PSO-SVR和GA-SVR模型对开都河1996—2012年径流预测的平均相对误差绝对值分别为2.65%、3.22%,平均绝对误差分别为1.055亿m3和1.291亿m3,预测精度和泛化能力均优于GS-SVR模型,表明PSO和GA能有效对SVR惩罚因子C和核函数参数g进行优化,具有预测精度高、泛化能力强以及稳健性能好的特点。相对而言,PSO-SVR模型性能略优于GA-SVR模型。     

PSO-SVR模型及其在COD_(Mn)预测中的应用

针对回归支持向量机(SVR)惩罚因子C和核函数参数g的选取对模型性能有着关键性影响,以及在实际应用中存在参数选取等困难,提出基于启发式优化算法的PSO-SVR多元变量COD_(Mn)预测模型,以云南省某断面COD_(Mn)预测为例进行实例研究。利用SPSS软件选取COD_(Mn)影响因子,确定输入变量;基于粒子群算法(PSO)基本原理,采用PSO优化SVR惩罚因子C和核函数参数g,构建PSO-SVR多元变量COD_(Mn)预测模型,并构建GA-SVR、 GA-BP及传统BP模型作为对比模型。利用所构建的模型对实例进行预测分析,结果表明:PSO-SVR模型随机连续运行100次对实例COD_(Mn)预测的平均相对误差绝对值和最大相对误差绝对值分别在2.355 2%～3.200 5%和4.030 3%～4.735 0%之间,平均值分别仅为2.586 7%和4.268 1%,预测精度和泛化能力均优于GA-SVR、 GA-BP及传统BP模型,表明PSO能有效对SVR惩罚因子C和核函数参数g进行优化, PSO-SVR模型具有预测精度高、泛化能力强以及稳健性能好等特点。     

支持向量机回归算法的唐山市降水量空间插值研究

针对传统插值方法在处理小样本和非线性问题上的不足,运用支持向量机回归（SVR）算法对唐山市降水量进行空间插值研究,提出了SVR算法的解决方案,以提高降水预测的准确性。通过收集2020年唐山市32个气象站点观测数据,提取经度、纬度、经纬度乘积、海拔、坡度、坡向和GPM等参数作为特征变量,并利用SVR算法建立降水量空间插值模型。研究结果表明,SVR在解决小样本和非线性问题上具有显著优势,能够有效分析多维变量对降水量分布的影响;验证显示其插值精度R2为0.79, MAE和RMSE分别为33.57mm、39.29mm,该精度在可接受范围内。通过SVR插值法生成精确的降水量分布式信息,可为唐山市旱涝减灾、水资源利用及生态保护提供科学数据支持。     

基于云计算和智能优化SVR的光伏发电功率预测

为提高光伏发电功率的预测精度,针对支持向量机回归(Support Vector Regression,SVR)模型的预测结果易受其惩罚系数C、敏感损失函数的最大误差系数ε和核函数g影响的问题,提出一种基于新型智能算法-蝗虫算法优化SVR模型参数的光伏发电功率预测模型。由于光伏发电功率数据存在随机性和间隙性的特征,Multi-Agent和分布式思想被引入蝗虫算法优化SVR模型,通过将云计算的MapReduce框架和GOA-SVR结合,提出一种基于MapReduce和GOA-SVR并行化的光伏发电功率预测模型(MapReduce and GOA-SVR,MR-GOA-SVR),从而提高海量高维光伏发电数据的处理能力。将影响光伏输出功率的11个气象因素作为GOA-SVR的输入向量,光伏输出功率作为GOA-SVR的输出向量,建立GOA-SVR的光伏发电功率预测模型。研究结果表明:MR-GOA-SVR可以有效提高不同天气类型下的光伏发电功率的预测精度,具有很强的现实性和指导意义。与PSO-SVR、GA-SVR、GOA-SVR和SVR相比,MR-GOA-SVR在晴天、阴天和雨天均可以提高预测精度,且具有优异的并行性能。     

An experimental study for chlorine residual and trihalomethane formation with rechlorination.

Maintenance of adequate chlorine residuals and control of disinfection byproducts (DBPs) throughout water distribution systems is currently an important issue. In particular, rechlorination can be a powerful tool in controlling adequate chlorine residual in a large distribution system. The patterns of chlorine decay and formation of DBPs due to rechlorination are different from those of chlorination; chlorine decay is slower and trihalomethane (THM) formation is lower with rechlorination. The present study evaluates whether existing predictive models for chlorine residual and THM formation are applicable in the case of rechlorination. A parallel first-order decay model represents the best simulation results for chlorine decay, and an empirical power function model (modified Amy model) with an introduced correction coefficient (phi1, phi2) is more suitable to THM formation.     

给水管壁腐蚀对余氯衰减的影响

在参与供水管网余氯消耗的三种因素中,除了主体水和管壁生物膜的消耗外,管壁腐蚀也是不容忽视的因素,尤其在铸铁管中.从微观电化学反应的角度,得到了管壁腐蚀导致的余氯衰减模型,结果与实际情况符合较好.     

回归支持向量机集成模型在年径流预测中的应用

为进一步提高径流预测精度和泛化能力,根据回归支持向量机(SVR)特性及基本原理,提出考虑不同影响因子(输入向量)的SVR集成预测模型,以云南省南盘江西桥站1961—2007年径流预测为例进行实例研究。首先,利用相关分析法选取年径流预测的若干影响因子,依次构建不同影响因子的SVR单一模型对研究实例进行预测,并构建对应的RBF模型作为对比预测模型;然后,采用加权平均和简单平均2种方法对具有较好预测精度和互补性的单一模型的预测结果进行综合集成。结果表明基于SVR的加权平均和简单平均2种集成模型径流预测的平均相对误差绝对值分别为1.27%和1.54%,最大相对误差绝对值分别为2.99%和2.74%,其精度和泛化能力均大幅优于各单一模型以及基于RBF的加权平均和简单平均集成模型,表明加权平均SVR和简单平均SVR集成模型具有较高的预测精度和泛化能力。相对而言,加权平均集成模型赋予了预测效果好的模型更大的权重,预测精度和泛化能力均优于简单平均集成模型。预测模型和方法可为相关预测研究提供参考和借鉴。     

Integration of Shuffled Frog Leaping Algorithm and Support Vector Regression for Prediction of Water Quality Parameters

Quality of surface water is a serious factor affecting human health and ecological systems. Accurate prediction of water quality parameters plays an important role in the management of rivers. Thus, different methods such as (support vector regression) SVR have been employed to predict water quality parameters. This paper applies SVR to predict eight water quality parameters including (sodium (Na⁺), potassium (K⁺), magnesium (Mg⁺²), sulfates (SO₄ ^?2), chloride (Cl^?), power of hydrogen (pH), electrical conductivity (EC), and total dissolved solids (TDS)) at the Astane station in Sefidrood River, Iran. To achieve an efficient SVR model, the SVR parameters should be selected carefully. Commonly, various techniques such as trial and error, grid search and metaheuristic algorithms have been applied to estimate these parameters. This study presents a novel tool for estimation of quality parameters by coupling SVR and shuffled frog leaping algorithm (SFLA) . Results of SFLA-SVR compared with genetic programming (GP) as a capable method in water quality prediction. Using SFLA-SVR, average of RMSE for training and testing of six combinations of data sets for all of the water quality parameters improved 57.4 % relative to GP. These results indicate that the new proposed SFLA-SVR tool is more efficient and powerful than GP for determining water quality parameters.     

Rules for Optimal Operation of Reservoir-River-Groundwater Systems Considering Water Quality Targets: Application of M5P Model

The aim of this paper is to develop rules for optimal reservoir operation and water withdrawal from river and aquifer considering water supply and pollution control targets. The general approach is making use of an integrated water quantity-quality management (IWQM) modeling in conjunction with accurate data mining techniques. The IWQM model generates data, including; optimal releases and water withdrawal from river and aquifer for different conditions, and M5P and Support Vector Regression (SVR) data mining models utilize the results of the IWQM model for the derivation of rules. The IWQM model minimizes the deviation from water supply and water quality targets during the planning horizon. This method for derivation of operating rules is applied to a real world case study, Zayandehrood system, in Iran, with serious water supply and water pollution problems. The IWQM model is analyzed for different hydrologic and water demands scenarios with total dissolved solids (TDS) as the water quality indicator. Results show that an integrated approach to reservoir-river-aquifer operation in the study area can reduce the TDS by 43 % in the downstream river.     

A Novel Hybrid Method for River Discharge Prediction

Accurate prediction of river discharge is essential for the planning and management of water resources. This study proposes a novel hybrid method named HD-SKA by integrating two decomposition techniques (termed as HD) with support vector regression (SVR), K-nearest neighbor (KNN) and ARIMA models (combined as SKA) respectively. Firstly, the proposed method utilizes local mean decomposition (LMD) to decompose the original river discharge series into sub-series. Next, ensemble empirical mode decomposition (EEMD) is employed to further decompose the LMD-based sub-series into intrinsic mode functions. Further, the EEMD decomposed components are used as inputs in three data-driven models to predict river discharge respectively. The prediction of all components is then aggregated to obtain the results of HD-SVR, HD-KNN and HD-ARIMA models. The final prediction is obtained by taking the average prediction of these models. The proposed method is illustrated using five rivers in Indus Basin System. In five case studies, six models were built to compare the performance of the proposed HD-SKA model. The data analysis results show that the HD-SKA model performs better than all other considered models. The Diebold-Mariano test confirms the superiority of the proposed HD-SKA model over ARIMA, SVR, KNN, EEMD-ARIMA, EEMD-KNN, and EEMD-SVR models.

     

天津市给水管网动态水质模拟系统

通过实验室试验和现场测试,建立了生长环作用下的实际管网水质数学模型,在水力模拟的基础上,进行模型求解与分组校核,实现了天津市给水管网水质动态模拟,模拟参数包括水龄、余氯和三卤甲烷。针对管网中可能发生的污染事件,实现了水质路径诊断。     

基于主要驱动因子筛选法和深度学习算法的浙江省动态需水量预测

收集了浙江省2000—2020年各用水行业需水量数据,采用基于Spearman秩相关分析的主要驱动因子筛选法筛选了影响各行业需水量的主要驱动因子,进而构造了改进的长短时记忆(LSTM)神经网络需水量预测模型,对各行业需水量进行动态滚动预测,并将改进LSTM模型的预测结果与传统单变量LSTM预测模型、卷积神经网络模型、支持向量回归模型的预测结果进行了对比。结果表明,基于主要驱动因子筛选法改进的LSTM模型能实时动态滚动预测各行业每年需水量,且预测结果精度高于其他3种模型。     

RESEARCH ON MATHEMATICAL SIMULATION OF RESIDUAL CHLORINE DECAY AND OPTIMIZATION OF CHLORINATION ALLOCATION OF URBAN WATER DISTRIBUTION SYSTEM

The concentration of Residual Chlorine (RC) frequently violates the standard in situations of urban water distribution system with large water supply area and long time of distribution. If chlorine dosage increases within water treatment plant, although RC in distribution system could meet water quality standard, Disinfection By-Products (DBPs) such as hydrocarbon halide rises. In the paper, a mathematical model of chlorine allocation optimization was presented based on reaction kinetics mechanism and optimization theory to solve the problem. The model includes the objective function of minimizing annual operation cost and constraints of RC standard and rational chlorination station distribution, and solving by 0-1 Integer Programming (IP). The model had been applied to a real water distribution system. The simulation results of the model showed that adding chlorine in‘ water distribution system remarkably improved water quality and reduced the operation cost by 49.3% per year less than chlorine dosed only in water treatment plant to meet RC standard. The results prove adding chlorine in water distribution system based on the model can bring both technological and economic advancement.