An evolving fuzzy neural predictor for multi-dimensional system state forecasting

被引:10
作者
Li, De Z. [1 ]
Wang, Wilson [2 ]
Ismail, Fathy [1 ]
机构
[1] Univ Waterloo, Dept Mech & Mechatron Engn, Waterloo, ON N2L 3G1, Canada
[2] Lakehead Univ, Dept Mech Engn, Thunder Bay, ON P7B 5E1, Canada
基金
加拿大自然科学与工程研究理事会;
关键词
Multi-dimensional data sets; Evolving fuzzy neural network; System state prognosis; Multiple-step-ahead forecasting; Induction motors; TIME-SERIES PREDICTION; VECTOR AUTOREGRESSION; NETWORKS; MODELS; IDENTIFICATION; PRICES;
D O I
10.1016/j.neucom.2014.05.014
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In many applications of system state forecasting, the prediction is performed using multi-dimensional data sets. The traditional methods for dealing with multi-dimensional data sets have some short-comings, such as a lack of nonlinear correlation modeling capability (e.g., for vector autoregressive moving average (VARMA) models), and an inefficient linear correlation modeling mechanism (e.g., for generic neural fuzzy systems). To tackle these problems, an evolving fuzzy neural network (eFNN) predictor is proposed in this paper to extract representative information from multi-dimensional data sets for more accurate system state forecasting. In the proposed eFNN predictor, linear correlations among multi-dimensional data sets are captured by a VARMA filter, while nonlinear correlations of the data sets are modeled by a fuzzy network scheme, whose fuzzy rules are generated adaptively using a novel evolving algorithm. The proposed predictor possesses online learning capability and can address non-stationary properties of data sets. The effectiveness of the proposed eFNN predictor is verified by simulation tests. It is also implemented for induction motor system state prognosis. Test results show that the proposed eFNN predictor can capture the dynamic properties involved in the multi-dimensional data sets effectively, and track system characteristics accurately. (C) 2014 Elsevier B.V. All rights reserved.
引用
收藏
页码:381 / 391
页数:11
相关论文
共 29 条
  • [1] [Anonymous], 2005, NEW INTRO MULTIPLE T
  • [2] [Anonymous], 1994, Time series modelling of water resources and environmental systems
  • [3] Antweiler W., 1996, DATABASE RETRIEVAL S
  • [4] Time series prediction with recurrent neural networks trained by a hybrid PSO-EA algorithm
    Cai, Xindi
    Zhang, Nian
    Venayagamoorthy, Ganesh K.
    Wunsch, Donald C., II
    [J]. NEUROCOMPUTING, 2007, 70 (13-15) : 2342 - 2353
  • [5] Cooperative coevolution of Elman recurrent neural networks for chaotic time series prediction
    Chandra, Rohitash
    Zhang, Mengjie
    [J]. NEUROCOMPUTING, 2012, 86 : 116 - 123
  • [6] FISHWICK PA, 1989, P 1989 WINT SIM C, P702
  • [7] A vector autoregressive model for electricity prices subject to long memory and regime switching
    Haldrup, Niels
    Nielsen, Frank S.
    Nielsen, Morten Orregaard
    [J]. ENERGY ECONOMICS, 2010, 32 (05) : 1044 - 1058
  • [8] Nonlinear time series modeling and prediction using functional weights wavelet neural network-based state-dependent AR model
    Inoussa, Garba
    Peng, Hui
    Wu, Jun
    [J]. NEUROCOMPUTING, 2012, 86 : 59 - 74
  • [9] AN INTEGRATED BAYESIAN VECTOR AUTOREGRESSION AND ERROR-CORRECTION MODEL FOR FORECASTING ELECTRICITY CONSUMPTION AND PRICES
    JOUTZ, FL
    MADDALA, GS
    TROST, RP
    [J]. JOURNAL OF FORECASTING, 1995, 14 (03) : 287 - 310
  • [10] Evolving fuzzy neural networks for supervised/unsupervised online knowledge-based learning
    Kasabov, N
    [J]. IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART B-CYBERNETICS, 2001, 31 (06): : 902 - 918