Model predictive control of a fluidized bed dryer with an inline NIR as moisture sensor

Fluid bed drying and near infrared (NIR) spectroscopy are technologies widely used to dry and measure moisture content and other pharmaceutical granular materials’ attributes, respectively. This work focused on controlling a bench top fluid bed dryer using an industrial control system, the model predictive control (MPC) strategy, and NIR measurements of the moisture content of pharmaceutical powders. The MPC was implemented to reach the desired drying end-point while simultaneously manipulating two variables: airflow and inlet air temperature. These two manipulated variables were constrained based on the physical and chemical behavior of the process. The results showed that the use of the MPC with the inline NIR produced an adequate control performance and resulted at the same time in a reduction in energy consumption of as much as 60% in one case when compared with the current industrial practices.     

Comparison of three control strategies for optimization of spray dryer operation

Spray drying is the preferred process to reduce the water content of many chemicals, pharmaceuticals, and foodstuffs. A significant amount of energy is used in spray drying to remove water and produce a free flowing powder product. In this paper, we present and compare the performance of three controllers for operation of a four-stage spray dryer. The three controllers are a proportional-integral (PI) controller that is used in industrial practice for spray dryer operation, a linear model predictive controller with real-time optimization (MPC with RTO, MPC-RTO), and an economically optimizing nonlinear model predictive controller (E-NMPC). The MPC with RTO is based on the same linear state space model in the MPC and the RTO layer. The E-NMPC consists of a single optimization layer that uses a nonlinear system of ordinary differential equations for its predictions. The PI control strategy has a fixed target that is independent of the disturbances, while the MPC-RTO and the E-NMPC adapt the operating point to the disturbances. The goal of spray dryer operation is to optimize the profit of operation in the presence of feed composition and ambient air humidity variations; i.e. to maximize the production rate, while minimizing the energy consumption, keeping the residual moisture content of the powder below a maximum limit, and avoiding that the powder sticks to the chamber walls. We use an industrially recorded disturbance scenario in order to produce realistic simulations and conclusions. The key performance indicators such as the profit of operation, the product flow rate, the specific energy consumption, the energy efficiency, and the residual moisture content of the produced powder are computed and compared for the three controllers. In this simulation study, we find that the economic performance of the MPC with RTO as well as the E-NMPC is considerably improved compared to the PI control strategy used in industrial practice. The MPC with RTO improves the profit of operation by 8.61%, and the E-NMPC improves the profit of operation by 9.66%. The energy efficiency is improved by 6.21% and 5.51%, respectively.     

改进的云粒子群优化算法及其断路器优化应用

针对断路器的能耗优化设计中复杂的多维参数设定常采用经验选取的方式,易导致断路器自身能耗过大。为此,提出一种结合鲶鱼效应与云模型的改进粒子群优化算法对其多维相关参数进行优化选取,先将传统的粒子群优化算法与云模型相结合,对多维寻优粒子加以分类,控制不同粒子群在不同搜索状态下快速寻优;再引入鲶鱼效应扰动机制增加寻优粒子多样性,提高寻优精度;最后采用该改进算法对断路器能耗模型优化仿真以及断路器参数设定。结果表明,提出的改进方法可以实现断路器低能耗设计要求,并能有效提高其设计效率。     

数据和知识驱动的城市污水处理过程多目标优化控制

城市污水处理过程优化控制是降低能耗的有效手段, 然而, 如何提高出水水质的同时降低能耗依然是当前城市污水处理过程面临的挑战. 围绕上述挑战, 文中提出了一种数据和知识驱动的多目标优化控制(Data-knowledge driven multiobjective optimal control, DK-MOC)方法. 首先, 建立了出水水质、能耗以及系统运行状态的表达关系, 获得了运行过程优化目标模型. 其次, 提出了一种基于知识迁徙学习的动态多目标粒子群优化算法, 实现了控制变量优化设定值的自适应求解. 最后, 将提出的DK-MOC应用于城市污水处理过程基准仿真模型1 (Benchmark simulation model No. 1, BSM1). 结果表明该方法能够实时获取控制变量的优化设定值, 提高了出水水质, 并且有效降低了运行能耗.     

Economics-based NMPC strategies for the operation and control of a continuous catalytic distillation process

Process profitability is an yes or no criterion for the successful long-term operation of industrial processes. This article describes the use of dynamic online economic process optimization to improve the performance of chemical processes. Different model-predictive control techniques have progressively been applied to coupled multivariable control problems and in many cases, especially in the petrochemical industry, the reference values are adjusted infrequently by stationary optimization based upon a rigorous nonlinear stationary plant model (real-time optimization, RTO). In between these optimizations, however, the process may be operated suboptimally due to the presence of disturbances. Nonlinear dynamic model-based optimization has been proposed recently to combine optimal operation and feedback control. In this paper, a model of the complex dynamics of a pilot-scale continuous catalytic distillation process is used to explore the potential benefits of online economics optimizing control strategies. We compare the direct economic optimization scheme with a compromise scheme, the economics-oriented tracking controller. The outcome of this work indicates that by using direct economics optimizing NMPC the plant economics can be handled better while guaranteeing the product specifications which are formulated as explicit constraints.     

Controlling delivery and energy performance of parallel batch processors in dynamic mould manufacturing

Given the mounting concern about service levels and environmental sustainability, mould industry is facing growing pressure to improve delivery reliability and energy efficiency. While heat-treatment operation is a bottleneck that affects related performances in mould manufacturing. Effective production control of this operation is essential to improve the on-time delivery and reduce the energy consumption of the mould. The operation often involves parallel batch processors and incompatible jobs, which allows for simultaneous processing yet with same job family and different weights and due dates. This paper considers the batch process control of parallel processors for dealing with such non-identical jobs in dynamic environments. An event-driven look-ahead batching strategy called MLAB-DE has been proposed. In MLAB-DE, the individual decisions for each family excluding the effects of these decisions on other families are suggested firstly. Then each alternative decision by including its effects on all families is evaluated. MLAB-DE is used to control two kinds of conflicting objectives related to the delivery and energy performances and finally achieve trade-off based on two-level compromise programming model. Simulation study is conducted to verify the effectiveness of the MLAB-DE strategy and show that the results are promising as compared to benchmark rules.     

基于小波神经网络的木材干燥窑内传感器建模研究

针对木材干燥系统具有非线性、强耦合的特性,难以建立准确的数学模型,提出一种基于小波神经网络的建模方法。通过木材干燥窑内木材含水率传感器、温度传感器和湿度传感器采集的数据建立小波神经网络模型,并通过模型预测木材含水率传感器的测量值。小波神经网络将BP神经网络在非线性问题上自学习的能力与小波表征信号局部信息的能力相结合,具有很强的自适应分辨性和容错能力。利用实际木材干燥过程中采集的数据作为训练样本进行仿真实验。结果表明：小波神经网络方法建立的模型能够预测木材含水率传感器的测量值,模型泛化能力强,预测精度高于BP神经网络建立的模型,验证了小波神经网络对木材干燥窑内传感器建模的可行性和有效性。     

基于LSSVM的木材干燥建模研究

针对木材干燥过程的强非线性特点,提出以最小二乘支持向量机LSSVM建立木材干燥基准模型.通过实验用小型木材干燥窑实际干燥过程中采集的数据作为训练样本进行仿真实验,结果表明基于LSSVM的木材干燥模型预测输出能够准确反映干燥过程木材含水率的变化,模型结构简单、预测精度高、泛化能力强,验证了LSSVM对木材干燥过程建模是一种可行而有效的方法.     

基于CEPSO-LSSVM的煤炭消费量预测模型

针对煤炭消费量的时变性、非平稳性特点,为了提高煤炭消费量预测精度,提出了一种鲶鱼粒子群算法优化最小二乘支持向量机（LSSVM）的煤炭消费量预测模型（CEPSO-LSSVM）。将LSSVM参数编码成粒子位置串,并根据煤炭消费量训练集的交叉验证误差最小作为参数优化目标,通过粒子间信息交流找到最优LSSVM参数,并引入“鲶鱼效应”,保持粒子群的多样性,克服传统粒子群算法的局部最优,根据最优参数建立煤炭消费量预测模型,并采用实际煤炭消费量数据进行仿真测试。结果表明,相对于其他预测模型,CEPSO-LSSVM可以获得更优的LSSVM参数,提高了煤炭消费量预测精度,更加适用于复杂非线性的煤炭消费量预测。     

Optimization control of a pulverizing system on the basis of the estimation of the outlet coal powder flow of a coal mill

This study aimed to master the operating characteristics of a pulverizing system, improve the output control precision of the system, and reduce the fluctuation amplitude of the main operating parameters of coal-fired units. A nonlinear dynamic model of a direct-fired pulverizing system that considers the effect of coal moisture on the energy balance of a coal mill was established. Then, an estimated signal of the outlet coal powder flow of the coal mill was constructed as a new output control target of the pulverizing system. Finally, an output control optimization method for the pulverizing system was designed on the basis of this signal. Simulation results showed that the model effectively reflects the dynamic characteristics of a pulverizing system. In addition, the results of simulation were concordant with those of online measurements. The control scheme reduced the internal disturbances in the coal feed rate, thereby improving the tracking capability and control precision of the pulverizing system's output and enhancing the disturbance suppression capability of the mill outlet temperature. Thus, the designed control scheme can ensure the safe and stable operation of coal-fired units.     

基于卡尔曼滤波的木材含水率在线估计

为解决木材干燥过程木材含水率检测精度低的问题,提高木材干燥的自动控制水平,针对以含水率为基准的干燥过程,提出了应用卡尔曼滤波进行木材含水率在线估计方法。为验证该方法的有效性：首先建立了基于含水率基准加入高斯噪声数据集的卡尔曼估计模型,并在此模型基础上对实验测得的数值进行了在线估计和比较,结果表明卡尔曼滤波方法具有较好的估计精度。     

基于免疫粒子群算法的水轮机机组供水系统节能控制方法

传统供水系统节能控制方法忽略了对水轮机状态的监控,导致在降低系统能耗的同时,供水过程出现机组运行不稳定问题,为此,设计基于免疫粒子群算法的水轮机机组供水系统节能控制方法。利用T-S模糊模型构建供水系统数学模型。采用模糊神经网络结构作为新型节能控制的设计原理,设计新的节能控制器。通过免疫粒子群算法实现供水系统的整体控制,降低供水系统能源消耗,完成基于免疫粒子群算法的水轮机机组供水系统节能控制方法的设计。仿真实验结果表明：所提控制方法应用后,供水系统的能源消耗明显降低,且水轮机机组供水运行稳定性得到了提升,应用效果较为理想。     

Dynamic Neural-Network-Based Model-Predictive Control of an Industrial Baker's Yeast Drying Process

This paper presents dynamic neural-network-based model-predictive control (MPC) structure for a baker's yeast drying process. Mathematical model consists of two partial nonlinear differential equations that are obtained from heat and mass balances inside dried granules. The drying curves that are obtained from granule-based model were used as training data for neural network (NN) models. The target is to predict the moisture content and product activity, which are very important parameters in drying process, for different horizon values. Genetic-based search algorithm determines the optimal drying profile by solving optimization problem in MPC. As a result of the performance evaluation of the proposed control structure, which is compared with the model based on nonlinear partial differential equation (PDE) and with feedforward neural network (FFN) models, it is particularly satisfactory for the drying process of a baker's yeast.      

A comparison of simultaneous state and parameter estimation schemes for a continuous fermentor reactor

This article proposes a maximum likelihood algorithm for simultaneous estimation of state and parameter values in nonlinear stochastic state-space models. The proposed algorithm uses a combination of expectation maximization, nonlinear filtering and smoothing algorithms. The algorithm is tested with three popular techniques for filtering namely particle filter (PF), unscented Kalman filter (UKF) and extended Kalman filter (EKF). It is shown that the proposed algorithm when used in conjunction with UKF is computationally more efficient and provides better estimates. An online recursive algorithm based on nonlinear filtering theory is also derived and is shown to perform equally well with UKF and ensemble Kalman filter (EnKF) algorithms. A continuous fermentation reactor is used to illustrate the efficacy of batch and online versions of the proposed algorithms.     

混部数据中心负载特征及其任务调度优化分析

随着现代互联网数据中心的规模越来越大,数据中心面临着能耗、可靠性、可管理性与可扩展性等方面的挑战。同时,数据中心承载的服务多样,既有在线Web服务,也有离线批处理任务。在线任务要求较低的延迟,而离线任务要求较高的吞吐量。为了提高服务器利用率,降低数据中心能耗,当前数据中心往往将在线任务和离线任务混合部署到同一个计算集群中。在混部场景下,如何同时满足在线和离线任务的不同要求,是目前面临的关键挑战。分析了阿里巴巴于2018年发布的含有4034台服务器的混部计算集群在8天内的日志数据(cluster-trace-v2018),从静态配置信息、动态混部运行状态、离线批处理作业DAG依赖结构等出发,揭示其负载特征,包括任务倾斜与容器部署的相关关系等,根据任务依赖关系与关键路径,提出了相应的任务调度优化策略。     

非线性粒子滤波目标跟踪应用仿真

针对非线性环境中存在的机动目标跟踪问题,对基于贝叶斯估计的粒子滤波器进行研究,为解决混合退火粒子滤波重要密度函数构造的问题,在混合退火粒子滤波的基础上,通过对系统状态和观测粒子方差的研究,提出了非线性环境下动态退火参数粒子滤波的改进算法,在混合退火粒子滤波中引入动态退火参数来构造高效的重要密度函数,提高了混合退火粒子滤波的跟踪精度,应用该滤波方法对机动目标模型进行仿真,并对多种滤波跟踪算法进行性能测试和比较,仿真实验结果表明,在非线性环境下该粒子滤波方法可行有效.     

流程工业能耗系统多维子模型及其开放性集成框架

流程工业组成因素多,运行结构复杂,针对单一模型无法合理、全面地描述其能源消耗系统各因素及其关联关系的问题,提出了流程工业能耗系统多维子模型及其集成化构建方法。首先,综合设备、能量、信息、人员等多方面因素,从静态结构描述、能源消耗的动态行为以及企业能耗系统运作的目的性3个层面建立流程工业能耗系统多维子模型;然后,分析了不同视角子模型关联关系;最后,详细研究了能耗系统多维模型的开放性集成框架,从而为流程工业能耗系统建立集信息流、能源流和物料流为一体,并同时反映内、外部关键因素的多维集成化模型提供了方法支持。这种内外关联、高度集成的流程工业能耗系统模型的构建将为流程工业综合性能效评估奠定基础。     

基于时间序列自回归模型的绿色建筑供暖能耗短期预测

为实时了解绿色建筑供暖能耗的变化趋势,提升能耗预测效果,设计基于时间序列自回归模型的绿色建筑供暖能耗短期预测方法。利用增强迪基-福勒检验法,检验绿色建筑历史供暖能耗时间序列平稳性;对非平稳的历史能耗时间序列进行差分平稳化处理,获取平稳的历史能耗时间序列;在时间序列自回归模型内添加移动平均模型,并考虑能耗的气温影响因素,建立时间序列自回归移动平均模型;利用赤池信息准则确定模型阶数,通过粒子群算法确定模型参数;在模型阶数与参数确定后的模型内,输入平稳的历史能耗时间序列,输出供暖能耗短期预测值。实验证明：该方法可精准预测不同类型绿色建筑的短期供暖能耗;在不同绿色建筑渗透量时,该方法短期供暖能耗预测误差较小;在不同室外温度时,该方法短期供暖能耗预测的可决系数较高,即预测精度较高。     

基于ESN和PSO的非线性模型预测控制

针对传统的控制理论对实际的工业生产过程中的被控系统,特别是具有强非线性的系统控制效果不是很理想,而应用非线性模型预测控制算法能够较好解决非线性系统的控制问题,提出了一种基于回声状态网络(Echo State Network,ESN)模型进行非线性系统辨识和粒子群优化(Particle Swarm Optimizatio...     

电熔镁砂产品单吨能耗混合预报模型

产品的单吨能耗是反映电熔镁砂熔炼过程产品产量和能耗的综合生产指标. 通过分析炉内电热转换关系,利用能量守恒原理建立了产品单吨能耗模型. 针对模型的未知非线性和参数时变等综合复杂性提出了由基于机理分析的单吨能耗主模型和 基于神经网络的补偿模型组成的产品单吨能耗混合预报模型. 其中神经网络补偿模型用于补偿模型的未知非线性和参数不确定性对于预报模型准确性的影响. 采用某电熔镁砂熔炼过程实测数据验证了所建立的混合预报模型是有效的.