基于改进果蝇算法优化SVM的凝汽器真空预测  

Prediction of Condenser Vacuum based on SVM Improved by IFOA

作　　者：葛晓霞[1] 赵舒莹 肖洪闯 韩书乐 GE Xiao-xia;ZHAO Shu-ying;XIAO Hong-chuang;HAN Shu-le(School of Energy and Power Engineering,Nanjing Institute of Technology,Nanjing,China,211167)

机构地区：[1]南京工程学院能源与动力工程学院,江苏南京211167

出　　处：《热能动力工程》2020年第11期39-45,共7页Journal of Engineering for Thermal Energy and Power

基　　金：国家自然科学基金项目(51706093);江苏省高等学校大学生实践创新训练计划项目(201911276062Y)。

摘　　要：为了更有效地预测凝汽器真空,提出改进果蝇算法(IFOA)优化支持向量机(SVM)的方法来建立凝汽器真空预测模型。将果蝇算法(FOA)中的固定步长值变为自适应调整步长,使得算法在迭代后期具有较强的局部寻优能力,提高收敛精度,同时引入变异操作,避免了果蝇算法陷入局部最优。通过6种常用的基准函数仿真测试,结果证明了IFOA算法的有效性;以某660 MW机组为例,采用改进果蝇算法优化支持向量机(IFOASVM)对凝汽器真空进行预测,并与粒子群算法(PSO)优化支持向量机模型(PSOSVM)及FOA算法优化支持向量机模型(FOASVM)的凝汽器真空预测结果进行对比。实验结果表明:IFOASVM比另外两种模型的预测精度更高,泛化能力更强,能够实现对凝汽器真空的准确预测。In order to predict the condenser vacuum more effectively,it was proposed to use Improved Fruit Fly Optimization Algorithm(IFOA)to optimize the Support Vector Machine(SVM)for establishing the condenser vacuum prediction model.First,the fixed step size in the FOA algorithm is transformed into an adaptive step size,which makes the algorithm have strong local optimization ability in the later period of the iteration and improve the convergence precision.Beside the mutation operation was introduced to avoid the FOA algorithm falling into local optimum.Then,6 common benchmark functions were used to simulate the test.The results prove the effectiveness of the IFOA algorithm.Finally,with a 660 MW unit as an example,the IFOASVM is used to predict the condenser vacuum,and the prediction results are compared with those of the PSOSVM and FOASVM.The experimental results show that IFOASVM has higher prediction accuracy and generalization ability than the other two models and can accurately predict condenser vacuum.

关 键 词：凝汽器 真空 预测 改进果蝇算法 支持向量机 粒子群算法 

分 类 号：TK242[动力工程及工程热物理—动力机械及工程]