基于小波变换和粒子群小波神经网络组合模型的有效停车泊位短时预测

Short-term forecasting of parking space using particle swarm optimization-wavelet neural network model

基于停车场有效停车泊位短时变化特性,提出了一种小波变换和粒子群小波神经网络组合预测方法。首先,通过选择合适的小波函数对有效停车泊位时间序列进行多尺度的小波分解与重构,然后对重构后的时间序列分别采用小波神经网络进行预测,并利用粒子群算法对神经网络初始参数的选取进行优化,最后将各自外推的预测结果进行合成,得到最终预测结果。实例分析表明：与单独使用小波神经网络模型相比,小波变换-粒子群小波神经网络模型的预测精度提高了5-7倍,且预测稳定性较好。

A forecasting model was proposed based on the short-term changing characteristics of Available Parking Space（APS）.This model integrates the wavelet analysis,Particle Swarm Optimization（PSO）and Wavelet Neural Network（WNN）.First,the APS time series were decomposed and reconstituted by wavelet analysis.Then,WNN model was used to forecast the reconstructed time series respectively.The PSO method was employed to optimize the selection of the initial parameters of the neural network.Finally,the final forecasted ASP was induced by integrating the prediction results.A case study was carried out to verify the applicability of the proposed model.Compared with simple WNN model,the new method enjoys higher accuracy and stable performance that the APS forecasting can be improved by 5to 7times by this new method.