基于RFOA优化GRNN的水电机组振动预测  

Vibration prediction of hydropower unit based on RFOA-GRNN

作　　者：王继选[1] 胡润志 管一 张少恺 曹庆皎[1] 王利英[1] WANG Jixuan;HU Runzhi;GUAN Yi;ZHANG Shaokai;CAO Qingjiao;WANG Liying(School of Water Conservancy and Hydropower Engineering,Hebei University of Engineering,Handan 056038,China)

机构地区：[1]河北工程大学水利水电学院,河北邯郸056038

出　　处：《振动与冲击》2021年第21期120-126,共7页Journal of Vibration and Shock

基　　金：河北省教育厅重点项目(ZD2020182);河北工程大学博士专项基金项目(20120134);河北省教育厅重点项目(ZD2021021);河北省高等教育教学改革研究与实践项目(2018GJJG631);河北省水资源水环境调控及综合管理协同创新中心;河北省智慧水利重点实验室;河北省水资源高效利用工程技术研究中心。

摘　　要：针对水电机组振动的非平稳、非线性特点,提出利用改进果蝇算法(RFOA)优化广义回归神经网络模型(RFOA-GRNN)。通过改进果蝇算法的搜索步长和气味浓度判定公式,使该算法的局部寻优能力增强,收敛速度提高。通过8种常用的基准函数对FOA算法、DSFOA算法、RFOA算法进行仿真测试,测试结果验证了RFOA算法的有效性。利用三种优化算法优化GRNN的平滑因子,将优化后平滑因子代入GRNN模型对水电机组振动进行预测。结果表明,与FOA-GRNN和DSFOA-GRNN两种预测模型相比,RFOA-GRNN预测模型的预测结果最大相对误差分别降低了99.96%和99.28%。可以得到RFOA-GRNN模型的预测精度和稳定性方面均优于其他两种模型,验证了此模型的有效性。将其应用于水电机组状态趋势预测研究中,可为维护人员提前发现水电机组故障并及时检修进而保证水电机组安全稳定的运行提供保障。Here,aiming at non-stationary and non-linear characteristics of hydropower unit vibration,the revised fruit fly optimization algorithm(RFOA)was proposed to optimize the generalized regression neural network(GRNN)model,and form the new model called RFOA-GRNN.By improving the search step size and odor concentration determination formula of FOA algorithm,the local optimization ability and convergence speed of the algorithm were enhanced.8 common benchmark functions were used to do simulation testing for FOA algorithm,DSFOA one and RFOA one,and the testing results verified the effectiveness of RFOA algorithm.These 3 optimization algorithms were used to optimize the smoothing factor of GRNN,and the optimized smoothing factor was substituted into GRNN model to predict vibration of hydropower unit.The results showed that compared with two models of FOA-GRNN and DSFOA-GRNN,the maximum relative error of RFOA-GRNN decreases by 99.96%and 99.28%,respectively,so RFOA-GRNN is superior to the other two models in prediction accuracy and stability to verify the effectiveness of RFOA-GRNN;applying RFOA-GRNN in state trend prediction of hydropower unit can provide a guarantee for maintenance personnel finding faults of hydropower unit in advance and repairing it in time to ensure safe and stable operation of hydropower unit.

关 键 词：水电机组 改进果蝇优化算法(RFOA) 广义回归神经网络(GRNN) 平滑因子 振动预测 

分 类 号：TV734.1[水利工程—水利水电工程]