反向建模方法在汽轮机热耗率建模中的应用

针对热耗率与其影响因素之间存在的复杂非线性关系,提出了基于自适应混沌反学习万有引力算法(ACOGSA)和最小二乘支持向量机(LSSVM)的汽轮机热耗率反向建模方法.利用某600 MW超临界汽轮机组运行数据,采用基于LSSVM的反向建模方法建立热耗率预测模型,采用ACOGSA算法解决LSSVM的模型参数优化问题,并与GSA-LSSVM模型和BP神经网络模型的预测结果进行比较.结果表明:所建立的模型比传统模型具有更好的泛化能力,更能准确地预测汽轮机的热耗率.     

基于FOA和LS-SVM的风电场风速预测研究

针对目前最小二乘支持向量机选取核参数和惩罚因子的各种方法尚存在着一定的局限性,文章采用果蝇优化算法对参数进行优化选择,提出了基于果蝇优化算法与最小二乘支持向量机结合的风速混合预测方法。对新疆某风电场为期5天的240个(采样间隔0.5 h)实测风速值进行了仿真测试,利用建立的预测模型,对第5天的风速值进行预测,预测结果的平均绝对百分比误差仅为8.32%。将其与单纯的LS-SVM模型和基于网格搜索优化的LS-SVM模型的预测结果作了对比,仿真结果验证了基于果蝇优化算法和最小二乘支持向量机混合预测模型的可行性和果蝇算法对最小二乘支持向量机参数优化的有效性。     

基于数值模拟与试验运行数据的电站锅炉燃烧系统复合建模

建立电站锅炉燃烧系统的数值模拟模型,通过该模型提供的训练样本建立最小二乘支持向量机模型(LS-SVM),利用试验数据对所建模型进行修正,并通过实时运行数据更新模型,提出了一种电站锅炉燃烧系统的复合建模方法.结果表明:复合建模方法有效地提高了模型的精度和运算速度,通过对初始LS-SVM模型的修正和更新可以进一步提高模型精度,为燃烧优化打下良好的基础.     

基于PSO优化LS-SVM的短期风速预测

提出了一种基于粒子群(PSO)算法优化最小二乘支持向量机(LS-SVM)的风电场风速预测方法。以相关性较高的历史风速序列作为输入,建立预测模型,并用粒子群算法优化模型参数。在对未来1 h风速进行预测时,文章所提出的模型比最小二乘支持向量机模型及BP神经网络模型具有较高的预测精度和运算速度。算例结果表明,经粒子群优化的最小二乘支持向量机算法是进行短期风速预测的有效方法。     

模糊树鲁棒回归算法的研究及其应用

针对实际工程中噪声难以避免和预测的问题,提出了鲁棒性较强的加权模糊树(W-FT)算法,采用基于局部异常因子(LOF)的加权最小二乘法代替最小二乘法学习模糊规则的后件参数,通过2个典型的非线性例子验证了该算法的有效性.应用W-FT算法建立了电站锅炉NOx排放特性模型,并与其他建模方法所建模型进行了对比.结果表明:所提出的W-FT算法能够有效地辨识噪声和异常值,具有较强的鲁棒性,所建立的模型预测精度较高,泛化能力较强.     

基于BPNN和LSSVM的燃油工业锅炉热效率建模分析

采用BPNN(误差反向传播神经网络)和LSSVM(最小二乘支持向量机)分别对燃油工业锅炉热效率进行智能建模,并比较此2种模型的准确性和泛化性,结果表明,BPNN更具有推广性。     

基于NWAPSO参数全局寻优的LS-SVM电站风机性能预测

风机的性能曲线是风机选型和优化运行的重要依据.通常该曲线通过试验试验数据和性能图表上的数据进行曲线拟合获得.由于该曲线非线性很强,传统方法复杂昂贵,而且拟合精度不高。针对以上不足,提出了一种基于非线性权重自适应粒子群优化(NWAPSO)参数全局寻优的最小二乘支持向量机(LS-SVM)风机性能预测方法。通过最小二乘支持向量机建模,并应用非线性权重自适应粒子群优化算法对模型参数进行全局寻优,得到具有较高精度的风机性能曲线。计算结果表明,根据本文方法建立的模型很简洁,只需要知道少量的训练样本就能建立,可以比较精确的预测风机性能,具有较显著的工程应用价值。     

超临界直流锅炉煤水比控制系统机理建模及定量反馈控制

针对超临界直流锅炉在变负荷运行过程中被控对象参数不断发生变化的特点,将超临界直流锅炉看做具有不确定性的非线性系统,并根据部分分散控制方法对某600MW超临界机组简化模型进行了分析,充分论证了燃料量对中间点温度的动态特性单独建模的合理性．以超临界直流锅炉燃料量变化对水冷壁出口温度的影响为研究对象,进行动态机理分析并建立动态模型,通过计算在3个典型工况点处的传递函数,确定其不确定性模型的参数变化范围．将定量反馈理论应用到煤水比控制中,结合给定的设计性能指标计算出煤水比控制系统的反馈控制器和前置滤波器的参数．结果表明：锅炉中间点温度在不同负荷下都能较快地高精度跟踪给定信号,有较强的鲁棒性,控制性能良好且易于实现．     

基于最小二乘支持向量机的水冷壁温度软监测建模

叙述了利用最小二乘支持向量机（LSSVM）对超临界锅炉膜式水冷壁向火侧两个危险点温度进行多目标软建模。指出,该模型可用于向火侧壁温的在线监测,该方法学习速度快、泛化能力强,比GA-BP的软测量建模具有更好的推广能力。     

基于系统辨识的超临界600MW机组给水控制热工对象建模与仿真

超临界机组给水控制的任务是以中间点温度或焓值作为表征量,进而控制汽温和不同负荷下的给水量。中间点焓值概念明确、灵敏度高,并且由于工质大比热特性,故应用中间点焓作为表征量更科学。对于直流锅炉给水的吸热、蒸发和过热之间没有明确分界的复杂对象,机理建模已难于适用。而以输入输出数据提供的信息来建立热工对象模型的系统辨识方法则显出优越性。依据某超临界600MW机组DCS数据图,应用MATLAB系统辨识建立了中间点焓值-给水量热工对象数学模型并降阶简化。最后以一仿真实例验证了模型的有效性、准确性及工程应用价值。     

用于发动机控制模型的基于循环的平均值离散建模方法

讨论了面向控制的发动机模型,并从角度域提出了基于循环的平均值离散建模方法。将其与目前主要用于实时仿真的模型进行对比,分析了这些模型与本模型的不同特点。通过实例说明了在角度域上建立基于循环的平均值模型能够更好地反映发动机的离散本质。     

Modeling cooling coils

Finned-tube heat exchangers commonly used as cooling coils in air conditioning systems undergo complex heat transfer and dehumidification. Due to the presence of water film on the outside surface of the coils, the general approach for an analysis of dry surface is not adequate to predict the performance of such coils. This paper presents a modeling procedure for cooling coils with dehumidification based on the approach of Threlkeld. In order to verify the calculational results of the model, experiments were conducted with an aim to determine the outlet air conditions as well as some other parameters required as the inputs to the model. Comparison between the simulation and experimental results reveals that the model is accurate and suitable for predicting the performance of cooling coils with dehumidification.     

Modeling of Valveless Micropumps

There are two difficulties encountered in modeling valveless micropumps using lumped-element methods. The pressure loss coefficient for fluidic diodes used in valveless pumps to rectify flow depends on the flow direction. A problem arises in choosing the proper loss correlation because the flow direction is not known a priori. Another problem is the quadratic form of the equation for the flow through the fluidic diodes, which brings about multiple solutions. The above problems become even more serious in multi-chamber cases. They are overcome in this study by suitably formulating the flow resistance. In addition, the flow inertia is accounted for in the unsteady model. The steady and unsteady models are evaluated by comparing with CFD simulations, which also serve to illustrate the flow field in more detail. It is shown that, compared with the steady model, the variation of the flow rate and pressure predicted by the unsteady model behaves in a close manner to those obtained by multidimensional calculations.     

Modeling of coal-combustion processes

Modeling of coal-reaction processes has received significant emphasis over the past decade. Yet, model development has not reached the point where significant use is made in process development for coal utilization. One of the key recommendations from a panel of industrial and university professionals to the U.S. Department of Energy on research needs for coal utilization was that “Development of combustion models is needed to the point where they will find application in the management and control of practical systems.” The purpose of this publication is to present a review and evaluation of the development of coal-reaction models.Coal-combustion models are classified and potential uses identified. Then, the state of development of models for fixed (or slowly moving) beds, fluidized beds and suspended (or entrained) beds is presented. Six fixed-bed models, ten fluidized-bed models and thirteen suspended-bed models are reviewed and compared. Selected comparisons of model predictions with measurements are shown and research needs are noted. Emphasis is placed on suspended-bed models.     

Modeling drop size distributions

There is abundant literature discussing the prediction of a representative drop diameter in a spray. However, there are relatively few publications discussing prediction of a drop size distribution in sprays. In the present paper, we review the three available methods for modeling drop size distributions: the maximum entropy method, the discrete probability function method, and the empirical method.     

Modeling of vented deflagrations

A mathematical model of vented gas-phase deflagrations is presented. By introducing several empirical parameters, account is taken of initial turbulence in the gases, flame acceleration due to hydrodynamic instabilities prior to vent opening, and increased burning velocity due to turbulence generated by the venting process. Additionally, a mixture of burned and unburned gases is vented. Essential information needed to compute the pressure development during vented deflagrations (or in large closed vessels) is the rate of increase of flame area due to cell formation in the flame front prior to the vent opening.The model has been tested against methane/air mixtures at initial pressures of 45 psia in vessels up to 3.8 m³ in volume. Good agreement has been obtained.Further work is underway to gather data on vented deflagrations for gases such as propane, ethylene, and hydrogen (which represent a series of increasing burning velocities) and to investigate more fully the effect of initial turbulence and elevated pressures.     

网络热力学的建模方法

为进行热经济学的能量系统兼顾生态环境的第三层次优化，引入了网络热力学建模，进而形成了一种热经济学的新模式——网络模式。简要回顾网络热力学的起源和发展，提出了网络热力学的建模理论和方法。进行了某生态系统的网络热力学方法分析，表明网络热力学可把许多复杂问题转化为非电问题的电气网络，实现了复杂问题的描述离散化，最终演化成矩阵分析的简洁处理。因而，网络模式热经济学不但扩大了网络热力学的运用范围和规模，而且顺应了计算机发展的新形势，网络热力学必将成为分析各种复杂的包括生态系统在内的物理与工程系统的强有力的通用工具。     

Modeling of Flow System Dynamics

This paper presents non-linear and linear models of flow system of laboratory stand (semi-industrial scale) with pressurized fluidized bed boiler. The non-linear model belongs to the class of lumped parameter models. The linear model was obtained experimentally using a novel method for design of control systems in industrial plants - MIKOZ. This study has connection with comparison of various methods of mathematical modeling of flow systems frequently encountered in power equipment.     

Modeling of smoldering combustion propagation

Smoldering combustion of various natural and synthetic solid materials constitutes a substantial fire hazard; the process itself produces copious toxic gases and it can lead to flaming combustion. This review focuses on the coupled chemical and physical processes involved in self-sustained propagation of smoldering. The potential heat sources (gas-phase oxidation, oxidative polymer degradation. char oxidation) are examined, along with the heat sinks (polymer pyrolysis, water vaporization). It is concluded that, even for the most-studied case of cellulose, the chemical mechanisms involved in these processes are both too complex and too poorly understood to be included in a smolder propagation model. Greatly simplified kinetic schemes are currently inevitable; the kinetic parameters are empirical and of restricted usefulness. A general model of the thermophysics of propagation (for arbitrary chemistry) is presented as a benchmark against which to compare existing models. The general case, with gradients both on the scale of the particles that comprise the fuel and on the scale of the overall combustion wave moving through the fuel bed, is too complex to be tractable. The general model equations are non-dimensionalized: simplifications are noted in the limit of small or large values for certain dimensionless groups. Existing smolder propagation models in the literature are reviewed; all represent great simplifications of the general case and none closely describes a realistic smoldering fire hazard. The principal usefulness of existing models is in rationalizing certain experimentally-observed trends. Truly useful predictive models of realistic hazard situations remain to be developed.     

Modeling PV fleet output variability

This paper introduces a novel approach to estimate the maximum short-term output variability that an arbitrary fleet of PV systems places on any considered power grid. The paper begins with a model that demonstrates that the maximum possible variability for N identical, uncorrelated PV systems equals the total installed capacity divided by $\sqrt{2N}$. The paper then describes a general methodology that is applicable to arbitrary PV fleets. A key input to this generalized approach is the correlation, or absence thereof, existing between individual installations in the fleet at the considered variability time scale. In this respect, the article includes a presentation of new experimental evidence from hourly satellite-derived irradiances relating distance and fluctuation time scales in three geographic regions in the United States (Southwest, Southern Great Plains, and Hawaii) and from recent high density network measurements that both confirm and extend conclusions from previous studies, namely: (1) correlation coefficients decrease predictably with increasing distance, (2) correlation coefficients decrease at a similar rate when evaluated versus distance divided by the considered variability time scale, and (3) the accuracy of results is improved by including an implied cloud speed term.