基于遗传BP神经网络的地下水位预测模型

地下水水位动态预测对农田土壤盐渍化防治、地下水地表水资源的合理调度具有十分重要的意义。以新疆和静县某地下水观测井为研究对象,选择月均蒸发量、气温和灌溉量3个因素作为BP神经网络模型的输入量,利用遗传算法优化神经网络的权值与阈值,建立地下水水位的遗传BP神经网络预测模型。结果表明:遗传BP神经网络模型能较好表达地下水位与主控因素之间的非线性关系,预测结果与实测值之间的平均绝对百分比误差为0.040 3,测试样本的网络输出值与网络目标值的相关系数达0.967 3,模型预测效果较佳。研究结果为区域地下水的开发利用与保护提供参考依据。     

软土物理力学性质指标与微结构参数的灰色关联-神经网络模型

通过对广州市南沙地区大量软土物理力学试验和微结构分析,获取了40组软土试样的物理力学性质指标和微观结构参数。综合运用灰色关联分析的数据分析能力和人工神经网络的非线性映射功能,建立了软土物理力学性质指标与微结构参数的灰色关联-径向基神经网络模型。该模型利用灰色关联分析方法对数据进行预处理,提取重要因子作为网络的输入,而径向基神经网络充分利用样本数据信息,自适应确定隐含层节点个数、径向基函数中心、宽度以及网络的权系数。克服了传统RBF网络隐层节点数为样本个数,当数据较多时导致网络结构庞大、学习速度慢的缺点。通过模型A和模型B的实例研究表明,该方法简化了网络结构,提高了训练速度和预测精度,为软土物理力学性质与微结构参数关系的定量研究提供了一条有效途径。     

径向基函数神经网络需水预测研究

在分析山西省历年用水量和人均用水量的基础上，建立径向基函数神经网络需水预测模型，采用最近邻聚类学习算法确定径向基函数的宽度、选取聚类中心和权值。采用丰富的需水预测因子作为模型的输入，网络输出需水预测值。预测结果表明，径向基函数神经网络需水预测模型运算速度快，有较高的预测精度。需水预测可为水资源规划和配置提供依据。     

利用BP网络技术预测碳酸盐岩储层孔隙度

利用径向基函数(RadialBasisFunction,RBF)神经网络的原理对储层参数(孔隙度、渗透率)进行了预测,运用有效测井数据和岩心测试资料作为网络模型的学习样本,通过网络的学习、训练,建立了测井解释的RBF神经网络模型。应用此模型定量计算了鄂尔多斯盆地杭锦旗地区多口井的碎屑岩层的孔隙度和渗透率。与用传统的统计方法比较,神经网络的方法显示出了更好的精度和更强的实用性。     

基于改进的数量化理论和RBF神经网络组合方法的地下水水质预测

文中首先运用了一种改进的数量化理论I模型作为预处理工具,对影响地下水水质的20个因子进行定性数据转换、数据降维,随后将8个重要特征因子作为RBF（径向基函数）神经网络模型的输入,进一步对监测井的采样数据进行学习、训练,揭示地下水污染质迁移转化规律。尝试用经过改进的数量化理论与RBF神经网络方法二者结合,对沈阳李官水源地研究区监测井地下水水质变化进行模拟与预测,其仿真结果覆盖了现有的绝大部分实测数据,适用范围广泛,具有一定的推广价值。     

采空区移动变形的径向基函数神经网络概率积分法(RBF)反演

借助部分矿区地表移动实测数据,选用径向基函数(RBF)神经网络对概率积分法的计算参数进行反演,采用K层交叉验证对模型精度进行优化,并与BP神经网络模型和SVM模型预测结果进行对比,发现RBF模型精度均优于BP模型和SVM模型,且其稳定性较好。可为概率积分法预测评估采空区地表移动变形范围提供一种可靠的方法。     

用径向基函数配点法求解河渠间地下水问题

本文采用径向基函数配点法建立了河渠间地下水非承压稳定流问题的数值模拟模型。径向基函数配点法的计算结果与形状参数的取值密切相关。将计算所得的近似解与解析解对比产生的误差很小,说明径向基函数配点法是一种既有效又有较高精度的求解方法。     

基于多变量LSTM神经网络的地下水水位预测

为解决以往模型未考虑地下水位相关影响因素的问题,探讨长短期记忆（LSTM）神经网络在地下水位预测中的应用,利用长短期记忆神经网络,采用多变量输入的方式,构建了基于多变量LSTM神经网络的地下水水位预测模型。以泰安市岱岳区J1号监测井为例,采用2001-2014年地下水水位动态监测资料与相关影响因素数据,利用多变量LSTM神经网络对2015-2016年地下水位进行预测,并与单变量LSTM神经网络和反向传播（BP）神经网络进行对比。研究结果表明：以相关影响变量为输入的BP神经网络无法考虑时序变化规律,预测均方根误差最大,为2.399 3;以地下水位为变量输入的单变量LSTM神经网络仅能根据时序变化作出相应预测,无法考虑相关变量影响,预测均方根误差为2.102 2;基于多变量输入的LSTM神经网络的预测精度显著高于单变量LSTM神经网络和BP神经网络,预测均方根误差最小,仅为1.919 1。总体上,多变量LSTM神经网络地下水位预测模型仅在某些峰值处误差较大,但总体预测效果较为理想。     

用径向基函数神经网络模型预报感潮河段洪水位

径向基函数神经网络方法是一类比较优越的前向式多层神经网络,将其应用于感潮河段的洪水位预报。利用K 均值算法和最小二乘法来确定径向基函数神经网络的参数,并给出了具体计算方法。由于该方法比传统的BP算法有较快的收敛速度,使其具有较大的应用价值。基于感潮河段的具体特点,构建了具有若干个时段预见期的径向基函数神经网络模型。该模型应用于沂河的水位预报,结果表明,该模型运算快速、简便,预报精度较高。     

基于SOM-I-SVM耦合模型的滑坡易发性评价

在使用机器学习模型对滑坡进行易发性评价时,通常会在滑坡影响范围之外随机选取非滑坡样本点,具有一定的误差。为了提高滑坡易发性评价的精度,将自组织映射（self-organizing map,SOM）神经网络、信息量模型（information,I）以及支持向量机模型（support vector machine,SVM）进行耦合,提出一种基于SOM-I-SVM模型的滑坡易发性评价方法,并将SOM神经网络与K均值聚类算法进行对比,验证模型的可靠性。以十堰市茅箭区为例,首先通过对环境因子的相关性及重要性分析,筛选出距水系距离、坡度、降雨量、距构造距离、相对高差、距道路距离、地层岩性等7个因子,建立滑坡易发性评价指标体系,在此基础上计算出各因子的分级信息量值,并作为模型的输入变量进行滑坡易发性评价。分别采用SOM神经网络和K均值聚类算法选取非滑坡样本,然后将样本数据集代入I-SVM模型预测滑坡易发性。将SVM、I-SVM、KMeans-I-SVM、SOM-I-SVM等4种模型预测精度进行对比,其ROC曲线下面积（AUC）分别为0.82,0.88,0.90,0.91,说明SOM-I-SVM模型能...     

Groundwater depth predictions by GSM,RBF, and ANFIS models: a comparative assessment

The potential of grey self-memory model (GSM), radial basis function network (RBF), and adaptive neuro fuzzy inference system (ANFIS) models in forecasting groundwater depths over an unconfined aquifer was compared. GSM, RBF, and ANFIS modeling was carried out at five sites in Jilin City, northeastern China, considering the influential lags of monthly groundwater depth as the inputs. The performance of the models was evaluated using criteria standards (R, RMSE, MARE, NS) and graphical indicators. Results indicate that the performance of all models was satisfactory in the region which lack of hydro-meteorological data. Comparison of the goodness-of-fit statistics in the research indicated that ANFIS was the better technique than the other two at all the sites except for J₂₁, and GSM(1,1) was the worst model at all the sites. However, considering the practical advantages of GSM(1,1) technique, it was recommended as an alternative and cost-effective groundwater modeling tool. Meanwhile, it was found that the modeling prediction for the well with the stable and evenly distributed data series has more accurate fitting results, generally.     

Recent innovations in hazard and risk analysis

Air pollution has been a major transboundary problem and a matter of global concern for decades. Climate change and air pollution are closely coupled. Just as air pollution can have adverse effects on human health and ecosystems, it can also impact the earth’s climate. As we enter an era of rapid climate change, the implications for air quality need to be better understood, both for the purpose of air quality management and as one of the societal consequences of climate change. In this study, an attempt has been made to estimate the current air quality to forecast the air quality index of an urban station Kolkata (22.65°N, 88.45°E), India for the next 5 years with neural network models. The annual and seasonal variability in the air quality indicates that the winter season is mostly affected by the pollutants. Air quality index (AQI) is estimated as a geometric mean of the pollutants considered. Different neural network models are attempted to select the best model to forecast the AQI of Kolkata. The meteorological parameters and AQI of the previous day are utilized to train the models to forecast the AQI of the next day during the period from 2003 to 2012. The selection of the best model is made after validation with observation from 2013 to 2015. The radial basis functional (RBF) model is found to be the best network model for the purpose. The RBF model with various architectures is tried to obtain precise forecast with minimum error. RBF of 5:5-91-1:1 structure is found to be the best fit for forecasting the AQI of Kolkata.     

鲁能运河电厂长沟岩溶水源地水位模拟预测

长沟水源地是鲁能运河电厂供水水源地,通过预测水源地水位变化趋势,为电厂制定用水计划、及时调整开采布局提供依据。水位模拟预测是在五年长系列动态监测资料及地下水开采量调查基础上进行的,利用数值模型演练和预测了嘉祥单斜蓄水构造单元内在现状岩溶水开采状态下及增加20%开采量条件下的区域地下水流场变化特点,所模拟的曲线与实测曲线相吻合,所推算出的水文地质参数合理,为保证水源地合理、稳定、持续的开采提供了科学支持。     

基于WorldView-02高分影像的BP和RBF神经网络遥感水深反演

遥感水深反演是水深测量的一种重要技术和手段.以美济礁水深反演为例,选择WorldView-02高分影像为数据源,在辐射定标和大气校正的基础上,构建BP（Back Propagation）和RBF（Radial Basis Function）人工神经网络水深反演模型,以遥感影像8个波段为输入层,通过tansig、logsig、高斯函数和purelin函数变换实现从输入层到隐含层、隐含层到输出层的转换,以便反演水深.最后对反演水深与实测水深采用回归分析,求解决定系数（coefficient of determination,R2）、平均决定误差（Mean Absolute Error,MAE）、均方根误差（Root Mean Square Error,RMSE）等进行比较,评价2种模型的精度.结果表明,RBF神经网络模型结构更简单,对样本要求更低,反演精度达到0.995,更适合遥感水深反演.      

A hybrid artificial neural network-based agri-economic model for predicting typhoon-induced losses

Building a model to rapidly simulate the impact of typhoons on agriculture and to predict agricultural losses is crucial and great help for remedial measure and distributing subvention right after the disaster. The relationship between typhoon-related meteorological factors and agricultural losses was first evaluated, and the Pearson??s test was applied to find consequences of both landfall and non-landfall which can be appropriately used to synthesize the possible coverage to suitably describe how typhoons influence agricultural losses. The self-organizing feature map (SOM) was then used to map similar properties of data into the same cluster and display the distribution of input?Coutput patterns. Then, the clusters were adopted as centroids of radial basis function (RBF) neural networks. Finally, two hybrid self-organizing radial basis (SORB) networks that integrated SOM into RBF were constructed for predicting the event-based agricultural losses by feeding two different meteorological inputs (scenarios 1 and 2). The results indicate that the constructed SORB network has great ability to capture the relationship between meteorological characteristics and agricultural losses. Previously, it always takes several days to investigate and evaluate the agricultural damages after typhoons, which is a time-consuming process. In this study, the proposed agri-economic model also demonstrates its outstanding predictability, in real-time, and therefore effectively accelerates the official decision making on agricultural compensation after a typhoon strike.     

Hydraulic analysis of a radial collector well for riverbank filtration near Nakdong River, South Korea

A radial collector well is used for the extraction of a large amount of groundwater without causing a deep drawdown at the well??s center, and it is appropriate for the supply of municipal water through riverbank filtration (RBF). Flow path changes caused by water extraction through a radial collector well were simulated to estimate the amount of river water induction at a RBF site associated with Nakdong River in South Korea. The structure of the screened horizontal arms of a radial collector well was examined with respect to effective riverbank filtration. The relative ratio of the river water induced to the radial collector well compared to the total groundwater extraction was estimated to be 27?C52%. The amount of induced river water varies with the distance of a horizontal arm from the river, indicating that the location and structure of the collector well is significant for RBF. In all simulation cases, the maximum drawdown of the groundwater level near the collector well was 2.1?m, which is not significant considering the substantial pumping rate at the study site. It was concluded that RBF radial collector wells can be used at the study site for a sustainable water supply.     

A comparative study of shallow groundwater level simulation with WA–ANN and ITS model in a coastal island of south China

Accurate and reliable prediction of shallow groundwater level is a critical component in water resources management. Two nonlinear models, WA–ANN method based on discrete wavelet transform (WA) and artificial neural network (ANN) and integrated time series (ITS) model, were developed to predict groundwater level fluctuations of a shallow coastal aquifer (Fujian Province, China). The two models were testified with the monitored groundwater level from 2000 to 2011. Two representative wells are selected with different locations within the study area. The error criteria were estimated using the coefficient of determination (R ²), Nash–Sutcliffe model efficiency coefficient (E), and root-mean-square error (RMSE). The best model was determined based on the RMSE of prediction using independent test data set. The WA–ANN models were found to provide more accurate monthly average groundwater level forecasts compared to the ITS models. The results of the study indicate the potential of WA–ANN models in forecasting groundwater levels. It is recommended that additional studies explore this proposed method, which can be used in turn to facilitate the development and implementation of more effective and sustainable groundwater management strategies.     

Evaluating the support vector machine for suspended sediment load forecasting based on gamma test

Due to the various influencing factors on river suspended sediment transportation, determining an appropriate input combination for developing the suspended sediment load forecasting model is very important for water resources management. The influence of pre-processing of input variables by Gamma Test (GT) was investigated on performance of Support Vector Machine (SVM) with two kernels; Radial Basis Function (RBF) and polynomial in order to forecast daily suspended sediment amount in the period between 1983 and 2014 at Korkorsar basin, northern Iran. The best input combination was identified using GT and correlation coefficient analysis. Then, the SVM model was developed and the suspended sediment amount was forecasted with RBF and polynomial kernels. The obtained results in testing phase showed that GT-SVM (RBF kernel) model can estimate suspended sediment more accurately with the lowest RMSE (14.045 ton/day), highest correlation coefficient (0.88) and highest NSEC coefficient (0.88) than SVM (RBF kernel) model (RMSE?=?18.36ton/day, \( {R}^2=0.79, \) \( NSEC=0.73 \)) and had a better performance than the other models. The results indicated that in forecasting the first nine maximum values of suspended sediment load, GT-SVM (RBF) had a higher capability than the other models and could provide a more accurate estimation from the maximum rate of suspended sediment. The results of this study showed the capability of identifying the priority of the input parameters can change GT to a useful and technical test for input variables pre-processing to forecast the amount of suspended sediments.     

Evaluation of operational tropical cyclone intensity forecasts over north Indian Ocean issued by India Meteorological Department

India Meteorological Department (IMD) introduced the objective tropical cyclone (TC) intensity forecast valid for next 24 h over the north Indian Ocean (NIO) in 2003 and extended up to 72 h in 2009. In this study, an attempt is made to evaluate the TC intensity forecast issued by IMD during 2005–2011 (7 years) by calculating the absolute error (AE), root mean square error (RMSE) and skill in intensity forecast in terms of maximum sustained surface wind (MSW). The accuracy of TC intensity forecast has been analysed with respect to basin of formation (Bay of Bengal, Arabian Sea and NIO as whole), season of formation (pre-monsoon and post-monsoon seasons), intensity of TCs (cyclonic storm and severe cyclonic storm or higher intensities) and type of track of TCs (climatological/straight moving and recurving/looping type). The study shows that the average AE (RMSE) in intensity forecast is about 11(14), 14(19) and 20(26) knots, respectively, for 24-, 48- and 72-h forecasts over the NIO as a whole during 2009–2011. The skill of intensity forecast is about 44 %(48 %), 60 %(58 %) and 60 %(65 %) for 24-, 48- and 72-h forecasts during 2009–2011 with respect to AE (RMSE). There is no significant improvement in terms of reduction in AE and RMSE of MSW forecast over the NIO like that over the northwest Pacific and northern Atlantic Oceans during 2005–2011. However, the skill in intensity forecast compared to persistence method has significantly improved by about 6 %(10 %) and 9 %(8 %) per year, respectively, for 12- and 24-h forecasts considering the AE (RMSE) during 2005–2011. There is also significant increasing trend in percentage of 24-h intensity forecasts with error of 10 knots or less during 2005–2011.     

Operational forecasting of daily summer maximum and minimum temperatures in the Valencia Region

Extreme-temperature events have a great impact on human society. Thus, knowledge of summer temperatures can be very useful both for the general public and for organizations whose workers operate in the open. An accurate forecasting of summer maximum and minimum temperatures could help to predict heatwave conditions and permit the implementation of strategies aimed at minimizing the negative effects that high temperatures have on human health. The objective of this work is to evaluate the skill of the regional atmospheric and modelling system (RAMS) model in determining daily summer maximum and minimum temperatures in the Valencia Region. For this, we have used the real-time configuration of this model currently running at the Centro de Estudios Ambientales de Mediterráneo Foundation. This operational system is run twice a day, and both runs have a 3-day forecast range. To carry out the verification of the model in this work, the information generated by the system has been broken into individual simulation days for a specific daily run of the model. Moreover, we have analysed the summer forecast period from 1 June to 31 August for 2007, 2008, 2009 and 2010. The results indicate good agreement between observed and simulated maximum temperatures, with RMSE in general near 2 °C both for coastal and inland stations. For this parameter, the model shows a negative bias around ?1.5 °C in the coast, while the opposite trend is observed inland. In addition, RAMS also shows good results in forecasting minimum temperatures for coastal locations, with bias lower than 1 °C and RMSE below 2 °C. However, the model presents some difficulties for this parameter inland, where bias higher than 3 °C and RMSE of about 4 °C have been found. Besides, there is little difference in both temperatures forecasted within the two daily RAMS cycles and that RAMS is very stable in maintaining the forecast performance at least for three forecast days.