基于BP神经网络和模糊推理系统的短时交通流预测

Short-term Traffic Flow Prediction based on BP Neural Network and Fuzzy Inference System

本文研究短时交通流预测。短时交通流预测是智能交通系统研究和实践的必要基础。本文提出和建立了一个短时交通流量预测模型,该模型利用一个基于规则的模糊系统,非线性地组合BP神经网络模型和自适应卡尔曼滤波模型的交通流量预测结果,使得短时交通流量的预测结果更加准确可靠。该模型将传统方法和人工智能方法有机结合,一方面,利用人工神经网络强大的动态非线性映射能力,从而提高预测精度;另一方面,充分发挥卡尔曼滤波的静态线性稳定性,解决了单独使用BP神经网络进行预测时识别率不理想和可信度不高的问题。实验结果表明,本文提出的短时交通流预测模型具有较高的准确度和可靠度。

For the research and practice of modern intelligent transportation systems,short- term traffic flow prediction is an essential element. The main content of this paper is to establish a traffic prediction model for short- term traffic flow forecasting,using a rule- based fuzzy system,nonlinearly combine traffic flow forecasts resulting from an adaptive Kalman filter（ KF） and BP neural network model,which is referred as KBF model. Organic combination of traditional methods and artificial intelligence methods,on one hand,makes use of the powerful dynamic nonlinear mapping ability of artificial neural network,so as to improve the prediction accuracy; On the other hand,takes full advantages of the static linear stability of the Kalman filter to solve the problem that the forecasts recognition rate is not satisfactory and the credibility is not high while using a BP neural network only. Verified by experiments,this model is useful for traffic flow forecasting with high accuracy and high reliability.