基于RF-GA-SVM算法的施工升降机输入电压控制模型

Input voltage control model of construction elevator based on RF-GA-SVM algorithm

针对在超高层建筑施工中施工升降机的驱动电机输入电压受多因素影响,导致其调节困难、能耗严重的问题,提出一种基于数据挖掘的RF-GA-SVM输入电压控制方法。首先,在实现控制电压多因素分析的基础上,基于随机森林(random forests,RF)算法进行了影响因素的重要度计算,确定各因素的影响权重值和样本约简集合,并通过ROC(接收者操作特征)曲线和AUC(ROC曲线下的面积)分析验证属性约简集的分类准确率;其次,采用基于遗传算法(genetic algorithm,GA)优化的支持向量机(support vector machine,SVM)算法,构建施工升降机输入电压控制模型;最后,利用施工升降机远程监控平台采集历史数据,对本研究提出的RF-GA-SVM输入电压控制模型进行训练和验证。实例验证结果表明:该模型检测样本的均方根误差(RMSE)为1.493,平均绝对误差(MAE)为0.899,平均绝对百分比误差(MAPE)为0.291%,决定系数(R2)为0.98,说明本方法可以有效选取影响因素,并根据影响因素和实际需求实现施工升降机输入电压的有效控制。

In view of the problems of difficult regulation and serious energy consumption caused by many factors in the input voltage of the driving motor of the construction elevator in the construction of super high-rise buildings,an RF-GA-SVM control method of input voltage based on data mining is proposed.Firstly,based on the multi factor analysis of control voltage,the importance of influencing factors is calculated based on random forests(RF)algorithm,the influence weight value of each factor and sample reduction set are determined,and the classification accuracy of attribute reduction set is verified by ROC(receiver operating characteristic)curve and AUC(area under ROC curve)analysis.Secondly,the input voltage control model of construction elevator is constructed by the support vector machine(SVM)algorithm optimized based on genetic algorithm(GA).Finally,the RF-GA-SVM input voltage control model proposed is trained and verified using the historical data collected by the remote monitoring platform of construction elevator.The results of example verification show that the RMSE(root mean square error)of the model test samples is 1.493,MAE(mean absolute error)is 0.899,MAPE(mean absolute percentage error)is 0.291%and R2(coefficient of determination)is 0.98,indicating that this method can effectively select the influencing factors and realize the effective control of the input voltage of the construction elevator according to the influencing factors and actual needs.