基于PCA-PSO-ELM的道路结冰预测模型

Road Icing Prediction Model Based on PCA-PSO-ELM

路面结冰对道路交通安全产生严重影响,有效预测结冰并采取主动的防冰措施成为解决这一问题的最有效且经济的手段。为提高道路行车安全,实现对路面结冰的精准预测,提出了一种融合主成分分析法(PCA)、粒子群优化算法(PSO)和极限学习机(ELM)的道路结冰预测模型。首先,针对相对湿度、大气压强、空气温度、风速、风向、路面温度、水膜厚度、融雪剂浓度8个影响因素进行主成分分析;然后,提取影响因子的主成分,并设定PSO算法参数、种群规模和迭代次数;通过粒子群优化算法搜索极限学习机模型的最优输入权值和隐含层神经元阈值,从而构建了PCA-PSO-ELM道路结冰预测模型;最后,利用3个路段的路面气象数据对模型进行验证。结果表明:PCA-PSO-ELM结冰预测模型的平均预测准确率达到95.85%,显著优于传统ELM、BP神经网络及SVM;此外,该模型在相同时间的不同路段和不同时间的相同路段上均表现出较高的预测准确率、精确率、召回率及F1分数,表明其具备优秀的泛化能力。PCA-PSO-ELM模型在保证准确率的同时,提高了路面结冰预测结果的稳定性,为有效应对路面结冰问题提供了坚实的理论支持。

Road icing significantly impacts traffic safety.The effective icing prediction and the proactive anti-icing measures are crucial for addressing this issue in a cost-effective manner.To enhance road driving safety and achieve precise prediction on road icing,the integrated model was proposed with principal component analysis(PCA),particle swarm optimization(PSO),and extreme learning machine(ELM).Initially,due to 8 influencing factors,including relative humidity,atmospheric pressure,air temperature,wind speed,wind direction,road surface temperature,water film thickness,and deicer concentration,the PCA was carried out.Subsequently,the principal components of these factors were extracted.The PSO algorithm parameters,population size,and iteration times were set.The PSO algorithm was employed to optimize the input weights and hidden layer neuron thresholds of ELM model;thus the PCA-PSO-ELM road icing prediction model was constructed.Finally,the model was validated by using meteorological data from 3 road sections.The result indicates that the PCA-PSO-ELM icing prediction model achieves the average prediction accuracy of 95.85%,which significantly outperformings traditional ELM,BP neural network,and SVM models.Additionally,the model demonstrates high accuracy,precision,recall,and F1 scores across different road sections and time intervals,illustrating the excellent generalization capabilities.The PCA-PSO-ELM model not only guarantees the accuracy,but also improves the stability of road icing prediction,providing solid theoretical support for effectively addressing road icing issues.