基于混沌和BP神经网络的有效停车泊位预测

The Prediction of Effective Parking Space Based on Chaos and BP Neural Network

随着停车需求的迅速扩大,停车难题已成为城市交通中最为棘手的问题之一。有效停车泊位预测技术是实现停车诱导信息实时、准确的重要因素之一,对实施停车诱导信息系统具有重要的意义和作用。针对有效停车泊位变化的特点,采用混沌时间序列法对停车泊位的历史数据进行相空间重构,通过建立BP神经网络模型对有效停车泊位的变化趋势进行预测。相关验证分析显示,该方法对有效停车泊位进行短时预测时的相对误差小于6%,具有较高的预测精度。

With the rapid expansion of parking demand, parking problem has become one of the thorniest issues of urban transport. The technology of effective parking space prediction is an important factor to realize the parking guidance information system （PGIS） in real time and accurately, it also has great significance and effect to PGIS. For the characteristics of effective parking space changes, this paper performs phase space reconstruction for historical data of parking space by chaotic time series, and BP neural network model is constructed to predict the trend of the effective parking space. Finally, with the example verification, the average relative error of this method does not exceed 6% in the short-term prediction of effective parking space; it has high prediction accuracy.