基于改进最小二乘支持向量机的一次风机状态预测方法研究

Research on State Prediction Method of Primary Air Fan Based on Improved LSSVM

在“双碳”目标下,针对一次风机工况复杂且多变量强耦合特点,提出一种基于改进天牛须搜索算法(IBAS)优化的最小二乘支持向量机(LSSVM)一次风机状态预测模型。首先,基于“系统+部件”的思想多维度构建了原始特征体系,采用皮尔逊相关系数对各维度数据进行降维处理;其次,应用IBAS对LSSVM模型中的超参数进行寻优计算,建立了完整的一次风机状态预测模型;最后,采用国内某电厂的一次风机实测数据进行算例分析。结果表明:所提出的一次风机状态预测方法在精度和收敛速度上具有一定的优越性,其平均绝对百分比误差为2.53%,低于其他模型,可满足风机状态预测的工程实践需求。

To meet the goal of"carbon peaking and carbon neutrality",focusing on the characteristics of complex working conditions and multivariable strong coupling of primary air fan,a primary air fan state prediction model based on least squares support vector machine(LSSVM)using improved longicorn whisker search algorithm(IBAS)optimization was proposed.Firstly,the original feature system was built in multiple dimensions according to the concept of"system+component",then the Pearson correlation coefficient was used to reduce the dimension of all dimensions’data.After that,IBAS was used to optimize the super parameters in LSSVM model,thus a complete primary air fan state prediction model was established.Finally,the measured data of primary air fan in a domestic power plant were used for example analysis.Results show that the proposed method has specific advantages in accuracy and convergence speed,and the mean absolute percentage error is 2.53%,which is lower than other conventional models.The proposed method can meet the practical engineering needs of fan state prediction.