摘要
针对砷盐净化除钴中影响钴离子浓度预测的因素影响程度不同和数据噪：占造成的钴离子浓度预测精度较低的问题，提出一种新的钴离子浓度预测模型——灰色特征加权最小二乘支持向量机（LSSVM）模型。该模型采用灰色累加的方法削弱原始数据中的噪声，并通过对砷盐除钴反应影响较为明显的因素的分析，根据影响因素的重要程度分别赋予其不同的特征权重，以提高LSSVM的预测精度；利用动态分级微粒群算法的快速收敛性和全局优化能力，优化选取LSSVM模型的两个关键参数——惩罚因子a和核参数盯，以避免参数选择的盲目性。基于选取的砷盐净化除钴过程生产数据进行了仿真，结果表明，灰色特征加权LSSVM的预测精度高，预测值能很好地跟踪实际钴离子浓度的变化趋蛰．满足砷盐除钴讨程钴离子浓度预测的要求。
To solve the problem in removal of cobalt with arsenic that the cobalt ion concentration prediction accuracy is always low because of the data noise and multiple influence factors, a novel grey feature weighted least squares sup- port vector machine （LSSVM） model for cobalt ion concentration prediction is proposed. The model uses the grey accumulation method to weaken the influences of the noise in primary data, and based on the investigation of the main influence factors to the prediction of the cobalt ion concentration, sets feature weights to every influence fac- tors according to their importance to improve the predictive accuracy. The two parameters of LSSVM model, the penalty factor C and the core parameter tr, are optimized by the dynamic hierarchical particle swarm optimization al- gorithm which has the abilities of fast convergence and global optimization, so that the blindness in choice of model parameters can be avoided. The simulation results show that the grey feature weighted LSSVM model has the higher prediction accuracy, and its predicting values can well follow the changing trend of the real cobah ion concentra- tion. The proposed model satisfies the requirements of on-line prediction of the cobalt ion concentration in the co- balt removal process with arsenic.
出处
《高技术通讯》
CAS
CSCD
北大核心
2012年第12期1322-1328,共7页
Chinese High Technology Letters
基金
国家自然科学基金（61174133,61273159）,国家杰出青年科学基金（61025015）和国家“十二五”科技支撑计划（2012BAF03805）资助项目.
关键词
特征加权
最小二乘支持向量机(LSSVM)
钴离子
微粒群算法
灰色累加
feature-weighted, least squares support vector machine （ LSSVM）, cobah ion, particle swarmoptimization, grey accumulation
