传感器网络环境监测时间序列数据的高斯过程建模与多步预测

doi:10.11959/j.issn.1000-436x.2015247

摘要/Abstract

摘要：

针对传感网环境监测应用采集的时间序列数据，提出了一种新的基于高斯过程模型的多步预测方法，实现了对未来时刻的环境监测数据的预测。高斯过程模型通过核函数描述数据的特性，通过对环境监测数据的经验模态分解，以及对其内在物理特性的分析，构建了针对环境监测数据的高斯过程核函数，实现了对数据变化模式的描述。在基于3个数据集的5个种类、20 000多个环境监测数据上进行了性能对比实验，结果表明，与对比预测方法相比，提出的高斯过程多步预测方法对未来时刻的环境监测数据的平均预测精度可以提高20%，可以应用于环境参数未来趋势分析、异常环境事件预警等场景。

关键词: 传感网环境监测, 时间序列, 高斯过程, 多步预测

Abstract:

For time series data collected from WSN environmental monitoring applications,a novel multi-step prediction method based on Gaussian process model was proposed.The method could make prediction for future environmental monitoring data.Kernel functions were used to describe data properties in the Gaussian process model.Kernel functions for environmental monitoring data were constructed through the EMD(empirical mode decomposition)technique and analysis of data inherent physical properties.And the constructed kernel functions were capable of describing the data change mode.Extensive experiments for multi-step prediction performance comparison test were performed on three kinds of data sets using over 20 000 environmental monitoring data records.Experimental results show that the average prediction accuracy of the Gaussian process multi-step prediction method can be increased by 20% than compared prediction methods.The prediction method can be applied to future environmental parameters trend analysis,early warning for abnormal environmental events and other scenes.

Key words: WSN environmental monitoring, time serie, Gaussian process, multi-step prediction

陈艳,王子健,赵泽,李栋,崔莉. 传感器网络环境监测时间序列数据的高斯过程建模与多步预测[J]. 通信学报, 2015, 36(10): 252-262.

Yan CHEN,Zi-jian WANG,Ze ZHAO,Dong LI,Li CUI. Gaussian process modeling and multi-step prediction for time series data in wireless sensor network environmental monitoring[J]. Journal on Communications, 2015, 36(10): 252-262.

图/表 11

图1

图2

图3

图4

表1

表2

图5

图6

图7

表3

表4

参考文献 28

[1]	崔莉, 鞠海玲, 苗勇 ,等. 无线传感器网络研究进展[J]. 计算机研究与发展, 2005,42(1): 163-174. CUI L , JU H L , MIAO Y ,et al. Research overview of wireless sensor network[J]. Journal of Computer Research and Development, 2005,42(1): 163-174.
[2]	孙利民, 李建中, 陈渝 ,等. 无线传感器网络[M]. 北京: 清华大学出版社, 2005. SUN L M , LI J Z , CHEN Y ,et al. Wireless Sensor Network[M]. Beijing: Press of Tsinghua UniversityPress, 2005.
[3]	BROCKWELLL P J , DAVIS R A . Introduction to Time Series and Forecasting[M]. New York: SpringerPress, 1994.
[4]	JIN L , JU J M , MIAO Q L . Study on Ann-based multi-step prediction model of short-term climatic variation[J]. Advances Atmosphere Sciences, 2000,17(1): 157-164.
[5]	CHEVILLON G . Direct muti-step estimation and forecasting[J]. Journal of Economic Surveys, 2007,21(4): 746-785.
[6]	COX D R . Prediction by exponentially weighted moving averages and related methods[J]. Journal of the Royal Statistical Society Seri B, 1961,23(1): 414-442.
[7]	BONTEMPI G . Long term time series prediction with multi-input multi-output local learning[A]. Proc of the 2nd European Symposium on Time Series Prediction[C]. Helsinki,Finland, 2008. 145-154.
[8]	MCELROY T , WILDI M . Multi-step-ahead estimation of time series models[J]. International Journal of Forecasting, 2013,29(3): 378-394.
[9]	TAIEB S B , SORJAMAA A , BONTEMPI G . Multiple-output modeling for multi-step-ahead time series forecasting[J]. Neurocomputing, 2010,73(10): 1950-1957.
[10]	BAO Y , XIONG T , HU Z Y . Multi-step-ahead time series prediction using multiple-output support vector regression[J]. Neurocomputing, 2013,129(10): 482-493.
[11]	PROIETTI T . Direct and iterated multistep AR methods for difference stationary processes[J]. International Journal of Forecasting, 2011,27(2): 266-280.
[12]	HUANG Z F , SHYU M L . K-NN based LS-SVM framework for long-term time series prediction[A]. Proc of the IEEE International Conference on Information Reuse and Integration[C]. Las Vegas NV,USA, 2010. 69-74.
[13]	AIGUO S A , BO Z . K-nearest neighbor LS-SVM method for multi-step prediction of chaotic time series[A]. 2012 IEEE Symposium on Electrical and Electronics Engineering[C]. Kuala Lumpur,Malaysia, 2012. 407-409.
[14]	CHEN P A , CHANG L C , CHANG F J . Reinforced recurrent neural networks for multi-Step-ahead flood forecasts[J]. Journal of Hydrology, 2013,497(8): 71-79.
[15]	SHEN H Y , CHANG L C . Online multistep-ahead inundation depth forecasts by recurrent NARX networks[J]. Hydrology and Earth System Sciences, 2013,17(3): 935-945.
[16]	PILKA F , ORAVEC M . Multi-step ahead prediction using neural networks[A]. Proceedings of the 53rd International Symposium ELMAR[C]. Zadar,Croatia, 2011. 269-272.
[17]	LIANG L J , YU D , CHANG X . Research on multi-step prediction of the short-term information by empirical model decomposition to abnormal state road traffic information[A]. Proc of 2012 International Conference on Computer Science and Service System[C]. Nanjing,China, 2012. 2034-2037.
[18]	GUO Z H , ZHAO H Y , LU H Y ,et al. Multi-step forecasting for wind speed using a modified EMD-based artificial neural network model[J]. Renew Energy, 2012,37(1): 241-249.
[19]	XIONG T , BAO Y K , HU Z Y . Beyond one-step-ahead forecasting:evaluation of alternative multi-step-ahead forecasting models for crude oil prices[J]. Energy Economics, 2013,40(6): 405-415.
[20]	王秋香, 于德介 . 设备状态的多项式神经网络迭代多步预测法[J]. 计算机仿真, 2010,27(3): 179-181,262. WANG Q X , YU D J . Iterative multi-step based PFANN for condition prediction of equipment[J]. Computer Simulation, 2010,27(3): 179-181,262.
[21]	谢景新, 程春田, 周桂红 ,等. 基于经验模式分解与混沌分析的直接多步预测模型[J]. 自动化学报, 2008,34(6): 684-689. XIE J X , CHENG C T , ZHOU G H ,et al. A new direct multi-step ahead prediction model based on EMD and chaos analysis[J]. Acta Automatica Sinica, 2008,34(6): 684-689.
[22]	李军, 张友鹏 . 基于高斯过程的混沌时间序列单步与多步预测[J]. 物理学报, 2011,60(7): 143-152. LI J , ZHANG Y P . Single-step and multiple-step prediction of chaotic time series using Gaussian process model[J]. Acta Physica Sinica, 2011,60(7): 143-152.
[23]	RASMUSSEN C E , WILLIAMS C K I . Gaussian Processes for Machine Learning[M]. London: The MIT PressPress, 2006.
[24]	OSBORNE M A , ROBERTS S J , ROGERS A ,et al. Towords real-time information processing of sensor network data using computationally efficient multi-output gaussian process[A]. Proc of the International Conference on Information Processing in Sensor Networks[C]. St.Louis MO,USA, 2008. 109-120.
[25]	WILLIAMS , C K I , BARBER D . Bayesian classification with Gaussian processes[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1998,20(12): 1342-1351.
[26]	DUVENAUD D , LLOYD J R , GROSSE R ,et al. Structure discovery in nonparametric regression through compositional kernel search[A]. Proc of the 30th International Conference on Machine Learning[C]. Atlanta GA,USA, 2013. 1166-1174.
[27]	Intel Berkeley Lab[EB/OL]. .
[28]	MANDIC D , REHMAN N , WU Z ,et al. Empirical mode decomposition-based time-frequency analysis of multivariate signals[J]. The Power of Adaptive Data Analysis, 2013,30(6): 74-86.

数据		方法
数据	GPI	BPI	WNI
DS_3-H	0.393 5	0.433 3	1.805 2
DS_3-T	0.179 3	0.203 5	1.064 0
DS_1-OXY	0.350 3	0.392 4	0.466 5
DS_1-PH	0.030 0	0.032 2	0.044 7
DS_1-WT	0.170 6	0.214 9	0.305 9
DS_2-H	0.333 6	0.353 8	0.395 5
DS_2-T	0.162 8	3.972 7	1.129 2

数据			方法
数据	GPD	BPD		WND	BPM
DS_3-H	0.386 1	0.388 9		0.461 9	0.407 0
DS_3-T	0.179 2	0.187 4		0.626 6	0.192 9
DS_1-OXY	0.313 4	0.320 0		0.377 6	0.322 6
DS_1-PH	0.029 8	0.034 2		0.035 0	0.033 3
DS_1-WT	0.180 3	0.201 5		0.222 3	0.178 1
DS_2-H	0.288 3	0.313 8		0.327 9	0.415 4
DS_2-T	0.153 5	0.184 7		0.211 39	0.214 5

数据		方法
数据	GPNI	GPND	GPI
DS_3-H	1.157 8	0.592 0	0.393 5
DS_1-OXY	0.532 2	0.472 8	0.350 3
DS_1-PH	0.051 5	0.049 9	0.030 0
DS_1-WT	0.213 6	0.203 2	0.170 6
DS_2-T	0.368 9	0.286 7	0.162 8

数据		方法
数据	GPD	GPLI	GPLD
DS_3-H	0.386 1	0.506 2	0.502 0
DS_1-OXY	0.323 4	0.534 4	0.472 8
DS_1-PH	0.029 8	0.048 4	0.050 0
DS_1-WT	0.180 3	0.201 0	0.205 1
DS_2-T	0.153 5	0.265 7	0.223 7