多变量时滞GM(1,??) 模型及其应用

针对具有时滞因果关系的小样本系统建模问题, 提出一种灰色多变量时滞GM(1,??) 模型及其求解方法; 考虑到相关变量累加序列变化量较大的情形下, 驱动项不能被视为灰常量的问题, 给出了时滞GM(1,??) 模型的一种派生模型. 在此基础上, 通过数值仿真和实例分析验证了新模型的有效性. 数值结果表明: 时滞GM(1,??) 模型能够较好地描述和预测含时滞特征的小样本数据系统的运行规律, 不考虑相关因素的时滞作用时, 时滞GM(1,??) 模型退化为经典的GM(1,??) 模型.

     

基于虚拟变量控制的GM(1,N)模型构建及其应用

现实中系统行为特征序列常受到虚拟变量的影响,而此时传统GM(1,N)模型不能准确地描述系统特征的变化规律.将虚拟变量引入传统GM(1,N)模型的灰作用量,构建虚拟变量控制的GM(1,N)模型,讨论新模型的参数求解方法;鉴于背景值对模型精度有着重要影响,利用粒子群优化算法对含有插值系数的背景值进行优化求解;从两个角度提出虚拟变量有效性检验方法.最后,通过河南省农民人均收入预测案例表明,新模型能够准确描述虚拟变量影响下系统特征序列的未来变化趋势.     

时滞累积TDAGM($1,N,t)$模型及其在粮食生产中的应用

考虑到社会经济系统中广泛存在时滞因果关系,通过分析驱动因素对系统主行为的时滞累积作用效果以及系统行为线性发展趋势,构建了含时间趋势项的时滞累积型多变量灰色TDAGM($1,N,t$)模型,论证了GM(1, 1)、GM(1,N)、OGM(1,N)、时滞GM(1,N)、TDDGM(1,N)模型均是该模型在不同参数取值下的特殊形式;为避免模型求解过程中微分形式与差分形式转换而产生误差,通过定义TDAGM($1,N,t$)模型的派生形式,给出了TDAGM($1,N,t$)模型时间响应式的直接求解方法;针对模型时滞效应参数的识别和优化问题,基于粒子群优化算法,给出了TDAGM($1,N,t$)模型参数估计的算法框架.时滞系统的数值实验结果表明,TDAGM($1,N,t$)模型能够较好地解决含时滞特征的多变量系统预测问题.将该模型应用于河南省粮食产量预测的实例中,拟合精度较高且预测结果符合河南省粮食生产发展趋势,验证了模型的有效性.     

基于GM(0,N)模型的三元区间数序列预测

在工程和经济领域,很多数据序列具有很强的振荡性,这些振荡序列用区间数表示将包含更多信息.三元区间数不仅包含系统特征的上下界,还在中间增加一个偏好值,对三元区间数序列的预测研究具有很好的应用价值.为了使灰色模型GM(0,N)能够直接对三元区间数序列建模,改进了GM(0,N)模型方程的参数设置,将整体贡献系数和滞后项系数取为精确数,而将线性修正项系数和补偿系数设为三元区间数,从而对三元区间数的不同界点进行线性修正和补偿.进一步,为了提高对振荡序列的预测精度,结合马尔科夫预测和序列转换方法对模型的预测序列进行修正. 通过对我国用电量和社会消费品零售总额的预测,表明了所提出的三元区间数多变量灰色模型比单变量灰色模型和区间数序列转换为精确数序列再预测的方法效果更好.     

基于驱动因素控制的DFCGM$(1,N)$ 及其拓展模型构建与应用

针对一类驱动因素具有复杂变化特征的系统行为预测问题,将驱动因素序列对系统的作用函数引入经典GM$(1,N)$模型的灰色作用量,构建驱动因素控制的DFCGM$(1,N)$模型及其拓展模型,并探讨参数估计方法;从白化信息充分和匮乏两个角度,利用经验分析法和智能优化算法探索驱动因素控制参数的识别方法,并给出模型建模预测步骤;最后,通过对我国粮食产量进行预测,验证了模型的有效性和实用性,表明所提出模型能够有效解决多驱动因素影响的系统预测问题.     

基于非等间距的多变量MGM(1,m)模型

针对相互影响的多变量非等间距原始数据序列的模拟颅测问题,对多变量MGM(I,m)模型在非等间距情形下进行建模,建立了非等间距MGM(1,m)模型,以提高模拟颅测精度.心用实例表明了非等间距MGM(1,m)模型较各个变量建立单变量非等间距GM(1,1)模型的模拟预测精度有显著提高.该模型适于等间距和非等间距建模,使非等间...     

改进灰色GM(1,m)模型在变压器故障预测中的应用

针对灰色模型在预测变压器故障时对波动数据序列的预测误差较大的问题,提出了一种灰色GM(1,m)预测模型改进方案:对原始数据序列进行处理,使其具有更好的指数规律,以满足预测模型对光滑性的要求;对处理过的原始数据序列进行灰关联度分析,以得到各变量之间的关系;优化预测模型的背景值并用其建模;采用等维新息模型预测数据。采用改进的灰色GM(1,m)模型预测某变压器油中7种特征气体的体积分数,所得预测数据的平均残差和后验相对误差均小于GM(1,1)模型和传统GM(1,m)的预测结果,表明其具有更好的预测精确度。     

自适应GM(1,1,λ)模型及其适用范围

为解决传统GM(1,1)模型存在的问题,在运用积分中值定理证明含有自适应因子λ∈(0,1)的背景值构造方法可行性的基础上,将该方法引入传统GM(1,1)模型的定义型,推导出了GM(1,1)定义型预测公式,构造了具有自适应能力的GM(1,1,λ)模型.通过理论证明和数据模拟实验两方面的研究结果表明自适应GM(1,1,λ)模型能够克服现存问题,并将模型的适用范围扩大为发展系数a∈(-1/λ,1/1-λ),大于传统GM(1,1)模型的适用范围a∈(-2,2),且可用于高增长序列建模,比传统GM(1,1)模型具有更高的拟合和预测精度.     

灰色GM(1,1)模型及其在电力负荷预测中的应用

讨论了灰色模型GM(1,1)及其改进模型在电力负荷预测中的应用.从灰色理论建模特点出发,提出使用加权均值生成原始数据序列的数据生成方法,在进行平滑的非负电力负荷数据序列的预测中取得了较好的效果.通过后验差检验,对传统的灰色系统GM(1,1)模型和加权均值的GM(1,1)模型进行了比较分析.实例证明,加权均值生成数据的方法进行建模具有较好的精度,在实际电力预测系统中有很好的应用价值.     

基于组合插值的GM(1,1)模型背景值的改进*

针对 GM(1,1) 模型预测误差偏大的问题,对GM(1,1)模型背景值的构造形式进行了研究。为了能够更加有效地降低GM(1,1)模型的预测误差,提出了基于辛普森3/8公式和牛顿插值公式的组合插值方法来构造出新的GM(1,1)模型的背景值。在GM(1,1)模型的建模过程中,由于原始建模数据序列中的第一个数据没有参与建模, 导致原始数据序列的数据资源利用效率降低,影响了GM(1,1)模型预测精度。因此,可以通过把灰色协调系数b加在原始建模数据序列前面的方法,使第一个数据能够参与到GM(1,1)模型的建模过程中。为了检验模型的改进效果,进行了原始建模数据类型分别为纯指数型数据序列、稳定型数据序列和缺失型数据序列的三组实验。对每组测试实验的预测结果进行对比分析,可以发现,基于组合插值方法对GM(1,1)模型的背景值进行改进,可以极大地降低GM(1,1)模型的模拟和预测误差。改进后的模型具有比较好的预测稳定性,增强了GM(1,1)模型的适用性。     

振荡型GM(1,1)幂模型及其应用

针对现实世界广泛存在的小样本振荡序列建模和预测问题,提出含有系统延迟和时变参数的振荡型GM(1,1)幂模型。给出最小二乘准则下的两级参数包计算公式,在此基础上构建非线性优化模型以寻求最佳幂指数和时间作用参数,以此识别原始数据所蕴含的振荡特征。将该模型应用于应急资源需求预测,并将建模结果与传统GM(1,1)幂模型、ARIMA和EMD-ARIMA方法进行比较,结果表明振荡型GM(1,1)幂模型具有较高的精度。     

GM（1，N）和QSIM结合的复杂系统的定性仿真建模方法

针对复杂系统仿真中系统信息缺乏、QSIM建模方法可使用微分方程的特点,在复杂系统仿真方法中,提出了一种GM（1,N）和QSIM相结合的定性建模方法.首先给出了相关研究的现状,然后提出了GM（1,N）和QSIM相结合的仿真建模方法的基本原理和主要过程.最后通过一个系统仿真建模实例验证了该方法的可行性.结果表明,该方法具有充分利用系统较少信息,能将定量和定性信息有效地融合与复杂系统的仿真建模之中的特点.     

Examining the effects of field dependence–independence on learners' problem-solving performance and interaction with a computer modeling tool: Implications for the design of joint cognitive systems

An investigation was carried out to examine the effects of cognitive style on learners' performance and interaction during complex problem solving with a computer modeling tool. One hundred and nineteen undergraduates volunteered to participate in the study. Participants were first administered a test, and based on their test scores they were classified into three groups, namely field-dependent, field-mixed, and field-independent learners. Participants then received the same set of integrated-format materials and were asked to use a computer modeling tool to solve a complex problem about immigration policy. A multivariate analysis of variance was performed with field type as the independent variable, and cognitive load, problem-solving performance, and learner interaction with the computer tool as the dependent variables. The results indicated that there was no significant difference in terms of the amount of cognitive load reported. However, there was a significant difference in terms of learner problem-solving performance. Specifically, field-independent learners outperformed field-dependent learners, and field-mixed learners outperformed field-dependent learners. The results also indicated significant differences in computer interaction between field-independent and field-dependent learners, and between field-mixed and field-dependent learners. The qualitative findings of the study showed that students who interacted poorly with the software were unsure about how to systematically use the affordances of the computer tool to solve the problem, did not have a goal-directed plan or strategy in mind about how to investigate the issue at hand, and had difficulty with testing the immigration policies by appropriately controlling variables in order to collect data to inform decision making. Implications are discussed in terms of designing computer systems that scaffold learners' complex problem solving by considering the cognitive demands of the task.     

时变参数GM(1, 1) 幂模型及其应用

为了进一步增强灰色预测模型对原始数据的适应能力,提出一种时变参数GM(1,1)幂模型,通过引入多项式函数描述GM(1,1)幂模型的结构参数随时间的动态变化规律。根据建模样本量的不同,分3种情形给出了模型的参数辨识算式,同时给出了时变参数GM(1,1)幂模型白化方程的解析解,利用积分复合梯形公式将其转化为可用于预测的离散时间响应式,并提出了参数优化方法。应用实例表明,时变参数GM(1,1)幂模型比固定参数GM(1,1)幂模型具有更高的模拟和预测精度。     

Perspectives on the synthesis of plant-wide control structures

The synthesis of plant-wide control structures has resurfaced as the most important design problem in process control. In this paper we provide a comparative analysis of various approaches, with an emphasis on how well they address the inherent theoretical and practical issues associated with the design of such control systems. Starting with a formal statement of the problem, which corresponds to a multi-objective optimization problem, this paper argues that the central issue to be resolved is the translation of implicit operating objectives to sets of feedback-controlled variables. The principle of the Optimizing Feedback Control Structures is proposed as the formal medium for the identification of controlled variables. Furthermore, it is shown that the selection of the best sets of input (manipulated) and output (measured) variables for the formation of the controllers' structures is governed by classical control-theoretical aspects, such as; open-loop gains, model uncertainty, and nonminimum-phase process characteristics. Hierarchical viewing of a plant is proposed as an effective mechanism to contain the complexities of the problem by streamlining the (i) specification of control objectives at different time-scales, (ii) modeling needs and model uncertainties, (iii) selection of measured and manipulated variables, and (iv) formation of the control structures.