电梯交通流预测方法的研究
|
摘要
本论文主要研究了电梯交通流的预测方法。提出了将ARMA（自回归滑动平均模型）数学模型预测法与基于径向基函数的神经网络相结合的混合预测方法。首先分析电梯交通流样本数据，判断电梯交通流可以根据ARMA预测模型进行预测，分析ARMA模型预测方法预测电梯交通流的特点，通过比较RBF神经网络和BP神经网络，由于RBF网络具有良好的逼近任意非线性映射和处理系统内在的难以解析表达的规律性能力，并且具有较快的学习收敛速度，明显优于BP网络和其它方法，因此选用基于径向基函数的RBF的神经网络方法对电梯交通流预测进行改进。提出选用ARMA与RBF神经网络结合的混合预测方法。建立电梯交通流混合预测模型，根据交通流样本数据确定混合模型的参数，包括ARMA模型的参数和RBF神经网络的参数,建立电梯交通流的混合预测模型，并对模型进行学习训练。分析两种预测方法相结合的混合预测方法的预测效果，与ARMA预测和其他方法的预测效果进行比较，分别计算它们的预测误差值，比较混合预测法与其他方法预测的精度和优劣。通过VC编程和数据库技术实现电梯交通流的预测，通过输入的交通流样本数据，预测未来时刻的电梯交通流状况，得出混合预测模型预测方法具有明显的优势。通过对混合预测方法的仿真试验，比较预测值与实际值来验证预测效果的拟和效果，说明混合预测方法可以较好的预测电梯交通流。
建立电梯交通系统预测模型后，由于ARMA模型具有较强的预测能力和RBF神经网络具有较强的学习能力，对于电梯交通流的随机变化具有很好的适应能力。
在论文的最后，给出了混合预测模型与其他预测方法的比较，可以得出混合预测模型预测的效果较好。无论是计算预测误差还是进行预测的速度，混合预测方法都较好。
This thesis is mainly about the research in forecast approach for elevator traffic system. The hybrid forecast approach of combining the ARMA mathematical model forecast approach and Radial Basis Function neural network approach is proposed. The elevator traffic flow sample data is analyzed first, according to which the parameters and orders of the ARMA forecast model could be fixed on, and the forecast model for the elevator traffic can be built. So we can use the model to forecast the elevator traffic. Then the author analyzed the defects of ARMA model forecast approach in forecast the elevator traffic and compared the RBF and BP neural network. Because the RBF network is well capable of approaching arbitrary non-linear mapping and processing the inner-system principle which is hard to be expressed in resolution and has a faster study convergence speed. It is much superior to BP network and other approaches and the RBF network approach is selected to improve the elevator traffic flow forecast. The effect of the hybrid forecast approach and those of the single ARMA forecast and RBF network forecast are compared. Calculating their forecast error respectively, we can compare their precision and performance. By programming in VC and using database technology, we can realize the forecast for the elevator traffic flow. So with the traffic flow sample data, we can forecast the elevator traffic flow in the future. The result is that the hybrid approach outperforms other approaches outlined in the thesis. The similarity between the expected value and real value in simulation can show the effect of forecast.
The elevator traffic system adopts the hybrid approach. Because the ARMA model has better forecast ability while the RBF network has better study ability, so the approach is quite adaptive to the random variance in elevator traffic flow.
After building the forecast model for elevator traffic system, the model can be used to forecast the elevator traffic flow. Further we can select different forecast approach, from which come out different results. At the end of the thesis, the comparison of the hybrid and single forecast approach is given, and the hybrid approach is much superior, no matter whether to calculate the forecast error or the forecast speed.
建立电梯交通系统预测模型后，由于ARMA模型具有较强的预测能力和RBF神经网络具有较强的学习能力，对于电梯交通流的随机变化具有很好的适应能力。
在论文的最后，给出了混合预测模型与其他预测方法的比较，可以得出混合预测模型预测的效果较好。无论是计算预测误差还是进行预测的速度，混合预测方法都较好。
This thesis is mainly about the research in forecast approach for elevator traffic system. The hybrid forecast approach of combining the ARMA mathematical model forecast approach and Radial Basis Function neural network approach is proposed. The elevator traffic flow sample data is analyzed first, according to which the parameters and orders of the ARMA forecast model could be fixed on, and the forecast model for the elevator traffic can be built. So we can use the model to forecast the elevator traffic. Then the author analyzed the defects of ARMA model forecast approach in forecast the elevator traffic and compared the RBF and BP neural network. Because the RBF network is well capable of approaching arbitrary non-linear mapping and processing the inner-system principle which is hard to be expressed in resolution and has a faster study convergence speed. It is much superior to BP network and other approaches and the RBF network approach is selected to improve the elevator traffic flow forecast. The effect of the hybrid forecast approach and those of the single ARMA forecast and RBF network forecast are compared. Calculating their forecast error respectively, we can compare their precision and performance. By programming in VC and using database technology, we can realize the forecast for the elevator traffic flow. So with the traffic flow sample data, we can forecast the elevator traffic flow in the future. The result is that the hybrid approach outperforms other approaches outlined in the thesis. The similarity between the expected value and real value in simulation can show the effect of forecast.
The elevator traffic system adopts the hybrid approach. Because the ARMA model has better forecast ability while the RBF network has better study ability, so the approach is quite adaptive to the random variance in elevator traffic flow.
After building the forecast model for elevator traffic system, the model can be used to forecast the elevator traffic flow. Further we can select different forecast approach, from which come out different results. At the end of the thesis, the comparison of the hybrid and single forecast approach is given, and the hybrid approach is much superior, no matter whether to calculate the forecast error or the forecast speed.
引文
[1] Rechalet J, et al, Model Predictive Heuristic Control, Applications to Industrial Processes, Automatic,1978,17(5):413~428
[2] Christine Lim, Michael McAleer, Time series forecasts of international travel demand for Australia, Tourism Management 2002 (23)：389-396.
[3] Siu Hung Cheung, Ka Ho Wu, Wai Sum Chan, Simultaneous prediction intervals for autoregressive-integrated moving-average models, Computational Statistics and Data Analysis 1998 (28) ：297-306.
[4] W.W. Recker, B.V. Ramanathan, X-H. Yu, M.G. Mcnally, Markovian Real-Time Adaptive Control of Signal Systems, Mathl comput, Modeling, 1995, Vol .22, No.4-7: 355-375.
[5] Donald K, Wedding II, Krzysztof J, Cios, Time series forecasting by combining RBF networks, certainty factors, and the Box-Jenkins model, Neurocomputing 1996 (10):149-168.
[6] Fang-Mei Tseng, Hsiao-Cheng Yu, Gwo-Hsiung Tzeng, Applied Hybrid Grey Model to Forecast Seasonal Time Series, Technological Forecasting and Social Chang, 2001, 67,291-302.
[7] S.C. Chang, R.S. Kim, S. J. Kim, B. H. Ahn, Traffic-Flow Forecasting Using a 3-stage Model, Proceedings of the IEEE Intelligent Vehicles Symposium 2000.
[8] K.S. Man, Long memory time series and short term forecasts, International Journal of Forecasting 2003 (19): 477-491.
[9] C. Pereira, J. Henriques, A. Dourado, Adaptive RBFNN versus conventional self-tuning: comparison of two parametric model approaches for non-linear control, Control Engineering Practice 2000 (8): 3-12.
[10] J. Nowicka-Zagrajek, R. weron, Modeling electricity loads in California：ARMA models with hyperbolic noise, Signal Processing 2002 (82): 1903-1915.
[11] Dhingra,S.L, Mujumdar,p.p, Grajjar Rajesh H, Application of time series techniques for forecasting truck traffic attracted by the Bombay Metropolitan Region, Journal of Advanced Transportation, v 27, n 3, Winter, 1993, p 227-249.
[12] Pan, ziqang, Altman, Eitan, Basar, Tamer, Robust adaptive flow control in
high speed telecommunication networks, preceedings of the IEEE Conference on Decision and Control, v2, 1996, p 1341-1346.
[13] Nader Kameli, New Britian Conn, Elevator traffic prediction using historical data checked for certainty, United Stated Patent, Us00527228A, Dec.21,1993.
[14] Euan Robertson, Polmont, Scotland, Artifical intelligent traffic modeling and prediction system, United States Patent, Us005304752, Apr,19,1994.
[15] 陈玉祥，张汉亚，预测技术与应用，[M],北京，机械工业出版社，1985.
[16] G. Melard, J.-M. Pasteels, Automatic ARIMA modeling including interventions, using time series expert software, International Journal of Forecasting 2000 (16): 497-508.
[17] 李冬梅，王正欧，基于RBF网络的混沌时间序列的建模与多步预测，系统工程与电子技术，2002，第24卷第6期.
[18] Andreas S, Weigend, Neil A Gershenfeld, Time series prediction: forecasting the future and understanding the past, New York: John Wiley and Sons, Inc,2001.
[19] 夏乐天，朱永忠，工程随机过程，河海大学出版社，2000年6月.
[20] H.Brian Hwarng, Insights into neural-network forecasting of time series corresponding to ARMA(p,q) structures, Omega 2001 (29): 273-289.
[21] Yue Fang, Forecasting combination and encompassing tests, International Journal of Forecasting 19(2003) 87-94.
[22] H.Brian Hwarng, Insights into neural-network forecasting of time series corresponding to ARMA(p,q) structures, Omega 2001 (29): 273-289.
[23] 邵建新，最小二乘法线性拟和参数的确定问题，大学物理，2003.1.
[24] Fang-Mei Tseng, Hsiao-Cheng Yu, Gwo-Hsiuing Tzeng, Combing neural network model with seasonal time series ARIMA model, Technological Forecasting & Social Change 2002(69): 71-87.
[25] Haibo Chen, Susan Grant-Muller, Use of sequential learning for short-term traffic flow forecasting, Transportation Research Part C 9, 2001:319-336.
[26] 张有为，预测的数学方法，国防工业出版社，1991.
[27] Tim Hill·Marcus o’Connor· William Remus, Neural network models for time series forecasts, Management Science, Vol.42,No 7, July 1996 p1082-1092.
[28] Brian L.Smith, Michael J. Demetsky, Short-term traffic flow prediction models-a comparision of neural network and nonparameteric regression
approaches, Journal of Transportation Engineering, 2001.
[29] Nadhir Messai, Philippe Thomas, Dimitri Lefebvre, Abdellah El Moudni, A Neural Network Approach for Freeway Traffic Flow Prediction, Proceedings of the 2002 IEEE International Conference on Control Applications, September：18-22，2002.
[30] Instytut Meteorologii I Gospodar Wodnej, Oddziat Morski, ul, Forecast of storm surge by means of artificial neural network, Journal of Sea Research 49 (2003) 317-322.
[31] Ayca Kumluca Topalli, Iamet Erkmen, A hybrid learing for neural network applied to short term load forecasting, Neurocomputing 51(2003) 495-500.
[32] Holcomb T, et al. Local Training for Radial Basis Function Network: Towords Solving the Hidden Vnit Problem. Proc ACC,1991,2231-2235.
[33] Cybenko G. Approximations by Superposition of a Sigmodial Function, Mathematics of Control, Signals and Systems, 1989,(2): 303-314.
[34] Pank J, et al, Universal Approximain Using Radial Basis Function Network, Neural Computation,1991,(3): 246-257.
[35] Powell M J D, Radial Basis Function Approscimations to Polynomials, Proc of the 12th Biennial Numerial Analysis Conference, Dundee,1987,223-241.
[36] Giros F, et al, Networks and Best Approximation Property, Biological Cybernetics,1990,(63): 169-176.
[37] Chen S, et al, Orthogonal Least Squares Learning Algorithm for Radial Basis Function Networks, IEEE Trans on Neural Networks, 1991,2(2):302-309.
[38] Chen S, et al, Orthogonal Least Squares Learning Algorithm for Training Multioutput Radial Basis Function Networks, IEEE Proceedings F, 1994,139(6): 378-384.
[39] J.T. Yokum, J.S. Armstrong, Beyond accuracy : comparison of criteria used to select forecasting methods, Int. J. Forecast, 1995(4): 591-597.
[40] 诸静等著，智能预测控制及其应用，浙江大学出版社，2002.4.
[41] R.E.Cumby, D.M. Modest,Testing for market timing ability: a framework for forecast evaluation, J.Financ. Econ, 1987(2): 169-189.
[42] 岳有军，《基于遗传算法的电梯群控系统多目标控制策略的研究》，天津大学硕士学位论文，2000.1.
[2] Christine Lim, Michael McAleer, Time series forecasts of international travel demand for Australia, Tourism Management 2002 (23)：389-396.
[3] Siu Hung Cheung, Ka Ho Wu, Wai Sum Chan, Simultaneous prediction intervals for autoregressive-integrated moving-average models, Computational Statistics and Data Analysis 1998 (28) ：297-306.
[4] W.W. Recker, B.V. Ramanathan, X-H. Yu, M.G. Mcnally, Markovian Real-Time Adaptive Control of Signal Systems, Mathl comput, Modeling, 1995, Vol .22, No.4-7: 355-375.
[5] Donald K, Wedding II, Krzysztof J, Cios, Time series forecasting by combining RBF networks, certainty factors, and the Box-Jenkins model, Neurocomputing 1996 (10):149-168.
[6] Fang-Mei Tseng, Hsiao-Cheng Yu, Gwo-Hsiung Tzeng, Applied Hybrid Grey Model to Forecast Seasonal Time Series, Technological Forecasting and Social Chang, 2001, 67,291-302.
[7] S.C. Chang, R.S. Kim, S. J. Kim, B. H. Ahn, Traffic-Flow Forecasting Using a 3-stage Model, Proceedings of the IEEE Intelligent Vehicles Symposium 2000.
[8] K.S. Man, Long memory time series and short term forecasts, International Journal of Forecasting 2003 (19): 477-491.
[9] C. Pereira, J. Henriques, A. Dourado, Adaptive RBFNN versus conventional self-tuning: comparison of two parametric model approaches for non-linear control, Control Engineering Practice 2000 (8): 3-12.
[10] J. Nowicka-Zagrajek, R. weron, Modeling electricity loads in California：ARMA models with hyperbolic noise, Signal Processing 2002 (82): 1903-1915.
[11] Dhingra,S.L, Mujumdar,p.p, Grajjar Rajesh H, Application of time series techniques for forecasting truck traffic attracted by the Bombay Metropolitan Region, Journal of Advanced Transportation, v 27, n 3, Winter, 1993, p 227-249.
[12] Pan, ziqang, Altman, Eitan, Basar, Tamer, Robust adaptive flow control in
high speed telecommunication networks, preceedings of the IEEE Conference on Decision and Control, v2, 1996, p 1341-1346.
[13] Nader Kameli, New Britian Conn, Elevator traffic prediction using historical data checked for certainty, United Stated Patent, Us00527228A, Dec.21,1993.
[14] Euan Robertson, Polmont, Scotland, Artifical intelligent traffic modeling and prediction system, United States Patent, Us005304752, Apr,19,1994.
[15] 陈玉祥，张汉亚，预测技术与应用，[M],北京，机械工业出版社，1985.
[16] G. Melard, J.-M. Pasteels, Automatic ARIMA modeling including interventions, using time series expert software, International Journal of Forecasting 2000 (16): 497-508.
[17] 李冬梅，王正欧，基于RBF网络的混沌时间序列的建模与多步预测，系统工程与电子技术，2002，第24卷第6期.
[18] Andreas S, Weigend, Neil A Gershenfeld, Time series prediction: forecasting the future and understanding the past, New York: John Wiley and Sons, Inc,2001.
[19] 夏乐天，朱永忠，工程随机过程，河海大学出版社，2000年6月.
[20] H.Brian Hwarng, Insights into neural-network forecasting of time series corresponding to ARMA(p,q) structures, Omega 2001 (29): 273-289.
[21] Yue Fang, Forecasting combination and encompassing tests, International Journal of Forecasting 19(2003) 87-94.
[22] H.Brian Hwarng, Insights into neural-network forecasting of time series corresponding to ARMA(p,q) structures, Omega 2001 (29): 273-289.
[23] 邵建新，最小二乘法线性拟和参数的确定问题，大学物理，2003.1.
[24] Fang-Mei Tseng, Hsiao-Cheng Yu, Gwo-Hsiuing Tzeng, Combing neural network model with seasonal time series ARIMA model, Technological Forecasting & Social Change 2002(69): 71-87.
[25] Haibo Chen, Susan Grant-Muller, Use of sequential learning for short-term traffic flow forecasting, Transportation Research Part C 9, 2001:319-336.
[26] 张有为，预测的数学方法，国防工业出版社，1991.
[27] Tim Hill·Marcus o’Connor· William Remus, Neural network models for time series forecasts, Management Science, Vol.42,No 7, July 1996 p1082-1092.
[28] Brian L.Smith, Michael J. Demetsky, Short-term traffic flow prediction models-a comparision of neural network and nonparameteric regression
approaches, Journal of Transportation Engineering, 2001.
[29] Nadhir Messai, Philippe Thomas, Dimitri Lefebvre, Abdellah El Moudni, A Neural Network Approach for Freeway Traffic Flow Prediction, Proceedings of the 2002 IEEE International Conference on Control Applications, September：18-22，2002.
[30] Instytut Meteorologii I Gospodar Wodnej, Oddziat Morski, ul, Forecast of storm surge by means of artificial neural network, Journal of Sea Research 49 (2003) 317-322.
[31] Ayca Kumluca Topalli, Iamet Erkmen, A hybrid learing for neural network applied to short term load forecasting, Neurocomputing 51(2003) 495-500.
[32] Holcomb T, et al. Local Training for Radial Basis Function Network: Towords Solving the Hidden Vnit Problem. Proc ACC,1991,2231-2235.
[33] Cybenko G. Approximations by Superposition of a Sigmodial Function, Mathematics of Control, Signals and Systems, 1989,(2): 303-314.
[34] Pank J, et al, Universal Approximain Using Radial Basis Function Network, Neural Computation,1991,(3): 246-257.
[35] Powell M J D, Radial Basis Function Approscimations to Polynomials, Proc of the 12th Biennial Numerial Analysis Conference, Dundee,1987,223-241.
[36] Giros F, et al, Networks and Best Approximation Property, Biological Cybernetics,1990,(63): 169-176.
[37] Chen S, et al, Orthogonal Least Squares Learning Algorithm for Radial Basis Function Networks, IEEE Trans on Neural Networks, 1991,2(2):302-309.
[38] Chen S, et al, Orthogonal Least Squares Learning Algorithm for Training Multioutput Radial Basis Function Networks, IEEE Proceedings F, 1994,139(6): 378-384.
[39] J.T. Yokum, J.S. Armstrong, Beyond accuracy : comparison of criteria used to select forecasting methods, Int. J. Forecast, 1995(4): 591-597.
[40] 诸静等著，智能预测控制及其应用，浙江大学出版社，2002.4.
[41] R.E.Cumby, D.M. Modest,Testing for market timing ability: a framework for forecast evaluation, J.Financ. Econ, 1987(2): 169-189.
[42] 岳有军，《基于遗传算法的电梯群控系统多目标控制策略的研究》，天津大学硕士学位论文，2000.1.