摘要
针对多晶硅还原生产能耗预测精度较低问题,提出了基于VMD-ASSA-LSSVM模型的多晶硅生产能耗预测方法.首先,采用主成分分析方法对能耗影响因素的数据降维处理,提高模型执行效率.利用变分模态分解(Variational Mode Decomposition,VMD)将能耗序列分解为不同特征尺度能耗分量,降低能耗序列的非平稳性、复杂度.其次,为解决麻雀搜索算法(Sparrow Search Algorithm,SSA)的收敛慢与收敛精度低问题,引入适应性学习因子进行改进.结合改进的自适应麻雀搜索算法寻优最小二乘支持向量机的可调参数,建立了VMD-ASSA-LSSVM的能耗预测组合模型;然后对分解的能耗分量单独预测,叠加子序列预测结果即为最终能耗预测.最后,以某多晶硅企业实际生产数据验证该方法的有效性,证实提高了预测精度.
Aiming at the low prediction accuracy of polysilicon reduction production energy consumption,a polysilicon production energy consumption prediction method based on VMD-ASSA-LSSVM model is proposed.Firstly,principal component analysis is used to reduce the dimension of the data of energy consumption influencing factors to improve the execution efficiency of the model.The energy consumption series is decomposed into energy consumption components with different characteristic scales by using variational mode decomposition(VMD),which reduces the complexity of energy consumption series.Secondly,to solve the problems of slow convergence and low convergence accuracy of sparrow search algorithm(SSA),adaptive learning factor is introduced to improve it.Combined with the improved adaptive sparrow search algorithm to optimize the adjustable parameters of least squares support vector machine,a combined energy consumption prediction model of VMD-ASSA-LSSVM is established;Then,each energy consumption component is predicted separately,and the final energy consumption prediction is obtained by superimposing the prediction results of subsequence.Finally,the actual production data of a polysilicon enterprise are used to verify the effectiveness of the method and improve the prediction accuracy.
作者
赵铁成
谢丽蓉
范协诚
王智勇
邓佑刚
李朋
叶金鑫
ZHAO Tiecheng;XIE Lirong;FAN Xiecheng;WANG Zhiyong;DENG Yougang;LI Peng;YE Jinxin(School of Electrical Engineering,Xinjiang University,Urumqi Xinjiang 830017,China;Xinte Energy Co.Ltd.,Urumqi Xinjiang 830011,China;Xinjiang Yijiahua Industrial Technology Co.Ltd.,Urumqi Xinjiang 830018,China)
出处
《新疆大学学报(自然科学版)(中英文)》
CAS
2022年第4期498-507,共10页
Journal of Xinjiang University(Natural Science Edition in Chinese and English)
基金
国家自然科学基金(62163034).