模糊神经网络与SARIMA结合的时间序列预测模型

模糊神经网络和SARIMA模型分别对非线性和线性时间序列有很好的预测能力,但在实际应用中大多数序列并非稳定、单纯线性或非线性的。为了提高预测精度,提出了一种基于T-S模糊神经网络与SARIMA结合的时间序列预测模型。针对悉尼航班乘客收入数据给出了三种混合模型,并与模糊神经网络、支持向量机、SARIMA和BP神经网络四种单独模型进行比较。实验结果表明,从预测精度和参数选择方面来看,所给模型是有效的。     

基于多模态特征子集选择性集成建模的磨机负荷参数预测方法

如何融合球磨机系统研磨过程所产生的多模态机械信号构建磨机负荷参数预测(Mill load parameter forecasting, MLPF)模型是当前研究的热点. 针对上述问题, 本文提出一种基于多模态特征子集选择性集成(Selective ensemble, SEN)建模的MLPF方法. 首先, 对多模态机械信号进行时频域变换得到高维频谱数据; 接着, 采用相关系数法和互信息法对多模态频谱进行线性和非线性特征子集的自适应选择; 最后, 采用优化和加权算法对上述特征子集的候选子模型进行自适应地选择与合并, 得到基于SEN机制的MLPF模型. 采用磨矿过程实验球磨机的机械信号仿真验证了所提方法的有效性.     

Model validation for structural dynamic analysis: An approach to the Sandia Structural Dynamics Challenge

The validation of mathematical models constructed for the dynamic analysis of critical structures is a very important, but complex, process. The essential requirement is to provide confirmation, using independent and more reliable data than that presented by the model in question, that the subject model is capable of describing the essential physics of the structure’s behaviour within the required accuracy. In this paper, the procedures of model validation using experimental data on a structure are summarised and applied to a structural dynamics validation problem developed by Sandia National Laboratories. One of the essential issues is to separate out any non-linear features of the system and to construct an appropriate linear model that is as accurate as possible to cope with variability of the subsystem structures. The linear model, which is constructed using simulated test data from an assembly of sample subsystems, is expressed as a mean model with a standard deviation. It is further used in the system response prediction for system accreditation and target application under specified excitation loads. The influence of the weak non-linearity features are neglected in the system response prediction because the experimental method used to derive the test data obscured the non-linear effects and precluded their identification. Further consideration of identification and modelling of the non-linear element for the Sandia 3DOF calibration system is discussed to evaluate its influence on the accuracy of the spatial model.     

Recursive transformation matrices for linear dynamic system models

In time series analyses and forecasting, dynamic linear models of canonical form are quite often used, specially in Generalised Exponentially Weighted Regression (GEWR) type models, introduced by Harrison and Akram (1982) and Akram (1984). This form is convenient, but usually not attractive for operation since the meaning of the parameters is not clear. To overcome this problem one needs a system of dynamic linear models in diagonal or any other meaningful form similar to canonical form. This is obtained by reparameterisation of a model to a desired form by similarity transformation of one system to another system through a transformation matrix. A general analytical expression for the elements of this matrix and its inverse, specially in case of canonical to diagonal transformation, is not known. One needs to compute these elements each time a transformation is sought. To ease this cumbersome problem a general recursive form of transformation matrix along with its inverse is presented for canonical to diagonal transformation. This general result apart from simplifying the process of transformation of linear dynamic systems helps us to reparameterise the models to a desired form.     

Constructing NARMAX models using ARMAX models

This paper outlines how it is possible to decompose a complex non-linear modelling problem into a set of simpler linear modelling problems. Local ARMAX models valid within certain operating regimes are interpolated to construct a global NARMAX (non-linear NARMAX) model. Knowledge of the system behaviour in terms of operating regimes is the primary basis for building such models, hence it should not be considered as a pure black-box approach, but as an approach that utilizes a limited amount of a priori system knowledge. It is shown that a large class of non-linear systems can be modelled in this way, and indicated how to decompose the systems range of operation into operating regimes. Standard system identification algorithms can be used to identify the NARMAX model, and several aspects of the system identification problem are discussed and illustrated by a simulation example.     

基于直觉模糊线性方程组的IFTS 预测方法

针对模糊时间序列预测理论对不确定性数据集的实时模糊变化趋势研究存在的不足, 规范了直觉模糊时间序列的定义, 提出了基于直觉模糊线性方程组的直觉模糊时间序列预测方法. 所提出的算法将模型的求解转化为一系列带有约束的线性规划问题, 准确地反映了序列数据随时间发展变化的模糊关联规律, 简化了预测模型的复杂度, 提高了时间序列预测的精度, 扩展了直觉模糊时间序列预测理论的应用范围. 最后, 通过仿真实验验证了所提出方法的有效性和优越性.

     

Parameter set estimation for non-linear systems

Models of dynamical systems are instrumental for many purposes: prediction, control, simulation, tracking and so on. In this paper, we will show how parameter set estimation (PSE) can be applied to non-linear systems. Parameter set estimation identifies a set of estimates which are feasible with respect to the measured data and a priori information. This set of parameters, feasible for the given model structure, can then be used for system tracking or robust control designs. For application to robust control, it is important that the size of this set be as small as possible. In order to apply parameter set estimation techniques to a non-linear system, the system function is expressed in a tensor parameterization which is linear in the parameters (LP). Then it is shown how an optimum volume ellipsoid strategy for linear time invariant systems can be extended to this tensor parameterization of a non-linear system. The methodology is illustrated on two examples, the second of which uses data obtained from an operating glass furnace.     

Constrained quadratic stabilization of discrete-time uncertain non-linear multi-model systems using piecewise affine state-feedback

In this paper a method for non-linear robust stabilization based on solving a bilinear matrix inequality (BMI) feasibility problem is developed. Robustness against model uncertainty is handled. In different non-overlapping regions of the statespace known as clusters the plant is assumed to be an element in a polytope whose vertices (local models) are affine systems. In the clusters containing the origin in their closure, the local models are restricted to being linear systems. The clusters cover the region of interest in the state-space. A n affine state-feedback is associated with each cluster. By utilizing the affinity of the local models and the state-feedback, a set of linear matrix inequalities (LMIs) combined with a single non-convex BMI are obtained which, if feasible, guarantee quadratic stability of the origin of the closed loop. The feasibility problem is attacked by a branch-and-bound-based global approach. If the feasibility check is successful, the Lyapunov matrix and the piecewise affine state-feedback are given directly by the feasible solution. Control constraints are shown to be representable by LMIs or BMIs, and an application of the control design method to robustify constrained non-linear model predictive control is presented. In addition, the control design method is applied to a simple example.     

状态空间时间序列的区域物流需求预测研究

区域物流需求是制定区域物流发展政策、基础设施建设和物流系统规划的重要依据,由区域各项相关经济指标共同决定。针对区域物流需求预测中样本数量小的问题,提出了互信息高维度特征降维方法,在保证相关综合信息完整性基础上降低原始数据维度,在此基础上建立了状态空间时间序列预测模型,同时采用局部线性小波神经网络和LIBSVM支持向量回归模型进行对比实验。算例分析及实验结果表明,采用互信息降维后的预测模型相对误差平均减少54.8%,互信息与状态空间时间序列模型相结合的预测方法对于区域物流需求预测问题预测精度较高,相对误差约为0.08。     

Model-based cardiac diagnosis of pulmonary embolism

A minimal cardiac model has been shown to accurately capture a wide range of cardiovascular system dynamics commonly seen in the intensive care unit (ICU). However, standard parameter identification methods for this model are highly non-linear and non-convex, hindering real-time clinical application. An integral-based identification method that transforms the problem into a linear, convex problem, has been previously developed, but was only applied on continuous simulated data with random noise. This paper extends the method to handle discrete sets of clinical data, unmodelled dynamics, a significantly reduced data set theta requires only the minimum and maximum values of the pressure in the aorta, pulmonary artery and the volumes in the ventricles. The importance of integrals in the formulation for noise reduction is illustrated by demonstrating instability in the identification using simple derivative-based approaches. The cardiovascular system (CVS) model and parameter identification method are then clinically validated on porcine data for pulmonary embolism. Errors for the identified model are within 10% when re-simulated and compared to clinical data. All identified parameter trends match clinically expected changes. This work represents the first clinical validation of these models, methods and approach to cardiovascular diagnosis in critical care.     

Application and comparison of different identification schemes under industrial conditions

Two new identification procedures based on different strategies are presented to determine a mathematical model for the crude iron quality of a blast furnace process. The first algorithm estimates the parameters of a common linear difference equation model based on identification results obtained from separate series of data with given model structure. Prior to complete estimation of the parameters, the second method automatically detects the significant terms of a general model to optimally characterize a linear or non-linear system. The process is described by a general discrete polynomial expansion of past and present input signals and past output signals. This algorithm is based on orthogonalization and application of several information criteria. A comparison of the prediction accuracy of the linear model obtained from the first method with the linear and non-linear model resulting from the second algorithm and with models developed by other identification procedures is presented and some experiences from this application are discussed.     

Evaluation of alternative growth patterns in urban systems

The paper describes the formulation of the general planning model, TOPAZ (Technique for the Optimum Placement of Activities into Zones) and its application to evaluating and optimizing alternative growth patterns in urban systems.TOPAZ identifies an urban system as a set of interacting activities to be allocated to a set of zones to maximize an objective of overall benefit less cost of interaction between activities together with the benefit less cost of establishment of the activities, over a set of time periods. The model takes the form of a non-linear assignment problem with linear constraints, and is solved using iterative linear programming.The model is illustrated by application at a macro level to the evaluation of alternative corridor growth patterns for the city of Melbourne (population 2.4 million) over three time periods, 1970–80–90–2000. Two activities are involved, residential and employment activity, and the interactions consist of work, residential, industrial and commercial trips. The city is divided into forty zones.     

基于GPS实时数据的在线过滤与补遗研究*

为提高GPS数据的有效性与可靠性,使之为实现实时交通流量预测与交通诱导服务,首先基于GPS历史数据,以变异系数极小化为优化目标进行GPS数据过滤模型的优选,并对实时数据缺损提出两种补遗算法;然后给出基于动态GPS实时数据的在线数据过滤与补遗一体化算法;最后结合2008年杭州市GPS历史数据对模型进行应用。结果表明,基于简单算术平均的滤波模型即为GPS数据过滤的最佳模型;基于时间序列缺失数据的快速补遗算法能够很好地满足实时预报要求,实现快速补遗。     

Robust stabilization of uncertain systems by periodic feedback

An uncertain control system described by a family parametrized by an unknown parameter that takes values in a known constraint set is considered. The robust stabilization problem is defined as finding a dynamic output controller that globally stabilizes the uncertain system; that is, no matter what the parameter value chosen by ‘Nature’, the closed-loop system is globally asymptotically stable. It is shown that in the class of dynamic periodic controllers the solution to this problem exists under reasonable assumptions given for a general family of abstract non-linear models. Those assumptions, in the case of a family of linear finite-dimensional models, are equivalent to suitable stabilizability and detectability assumptions. For this case, insensitivity of the design with respect to small errors in the data is proved.     

A new hybrid correction method for short-term load forecasting based on ARIMA,SVR and CSA

Accurate load-forecasting problem is a significant and vital issue, especially in the new competitive electricity market. The models that are employed for forecasting purposes would determine how reliable the last forecasted results are. Therefore, this paper proposes a new hybrid correction method based on autoregressive integrated moving average (ARIMA) model, support vector regression (SVR) and cuckoo search algorithm (CSA) to achieve a more reliable forecasting model. The proposed method gets use of the autocorrelation function (ACF) and the partial ACF to search the stationary or non-stationary behaviour of the investigated time series. In the case of non-stationary data, it will be differenced one or more times to become stationary. After that, Akaike information criterion is utilised to find the appropriate ARIMA model such that the linear component of the data would be captured. Therefore, the ARIMA residuals would contain the non-linear components that should be modelled by use of the SVR model. The role of CSA as a successful optimisation algorithm is to find the optimal SVR parameters for more accurate forecasting. Meanwhile, a novel self-adaptive modification method based on CSA is proposed to empower the total search ability of the algorithm effectively. The proposed method is applied to the empirical peak load data of Fars Electrical Power Company in Iran.     

Multi-model predictive control of an industrial C3/C4 splitter

In this paper, the application of a linear predictive controller to an industrial distillation column that presents a nonlinear behavior is described. The system is represented by a set of linear approximating models, where each model corresponds to a possible operating point of the system. The control sequence computed by the control algorithm is based on a min–max optimization problem where the controller cost is minimized for the worst process model. The control algorithm makes use of a particular form of the state-space model, which preserves the structure of conventional model predictive control controllers that are based on the step response model. The performance of the proposed controller applied to an industrial system is illustrated with results of the real system at typical plant conditions with the controller performing as a regulator and as an output reference tracker.     

Fixed-structure Gaussian process model

This article describes a method for modelling non-linear dynamic systems from measurement data. The method merges the linear local model blending approach in the velocity-based linearisation form with Bayesian Gaussian process (GP) modelling. The new Fixed-Structure GP (FSGP) model has a predetermined linear model structure with varying and probabilistic parameters represented by GP models. These models have several advantages for the modelling of local model parameters as they give us adequate results, even with small data sets. Furthermore, they provide a measure of the confidence in the prediction of the varying parameters and information about the dependence of the parameters on individual inputs. The FSGP model can be applied for the extended local linear equivalence class of non-linear systems. The obtained non-linear system model can be, for example, used for control-system design. The proposed modelling method is illustrated with a simple example of non-linear system modelling for control design.     

数据仓库的多维数据模型的研究

多维数据模型是数据仓库和联机分析处理研究中的一个重要问题,该文根据电力负荷数据集的特点,提出了一种新模型,解决不同维公用一个底层层次属性,把系统中不完全的低粒度数据集和完全的粗粒度数据集在逻辑上无缝地结合起来支持联机分析处理的问题,这是其他多维数据模型所没有解决的。     

Predictive control based on neural network models with I/O feedback linearization

This paper presents an approach for the constrained non-linear predictive control problem based on the input-output feedback linearization (IOFL) of a general non-linear system modelled by a discrete-time affine neural network model. Using the resulting linear system in the formulation of the original non-linear predictive control problem enables to restate the optimization problem as the minimization of a quadratic function, which solution can be found using reliable and fast quadratic programming (QP) routines. However, the presence of a non-linear feedback linearizing controller maps the original linear input constraints onto non-linear and state dependent constraints on the controller output, which invalidates a direct application of QP routines. In order to cope with this problem and still be able to use QP routines, an approximate method is proposed which simultaneously guarantees a feasible solution without constraints violation over the complete prediction horizon within a finite number of steps, while allowing only for a small performance degradation.     

秦淮河流域东山站洪水位预报模型研究

为了提高秦淮河流域洪水预报的水平,对东山站洪水位过程预报模型进行深入研究。采用线性动态系统模型与BP人工神经网络模型建立东山站洪水位逐时段预报模型,采用2010—2015年及2016—2017年汛期秦淮河流域实测雨量和东山站水位资料对模型进行率定和验证。结果表明:东山站洪水位逐时段预报的BP人工神经网络模型相对于线性动态系统模型具有较高的精度;相对于一维河网水动力模型,简单实用。