(1. 中南大学 信息科学与工程学院，长沙 410083；
2. 湖南人文科技学院 通信与控制工程系，娄底 417000)
摘 要: 针对锌净化除钴过程生产数据存在噪声和系统参数缓慢变化的问题, 提出一种基于灰色模糊LSSVM的钴离子浓度预测模型。对样本数据进行灰色累加，削弱原始数据序列中的噪声，使数据规律性增强，灰色累加后数据作为LSSVM输入，提高模型抗干扰能力和预测能力；由于锌净化除钴工序的系统参数随时间发生变化，提出对不同时期的样本赋予不同的模糊加权值；利用改进PSO的全局优化能力和快速收敛性, 优化LSSVM模型的惩罚因子和核函数参数，避免人为选择参数的盲目性。对硫酸锌溶液净化除钴过程生产数据的仿真结果表明，灰色模糊LSSVM预测值能很好地跟踪实际值的变化趋势，满足钴离子浓度预测要求。
关键字: 最小二乘支持向量机；微粒群算法；模糊加权；灰色累加
（1. School of Information Science and Engineering, Central South University, Changsha 410083, China;
2. Department of Communications & Control Engineering, Hunan Institute of Humanities Science and Technology,
Loudi 417000, China）
Abstract:To solve the problems that noises exist in the data of cobalt removal from zinc electrolyte and the system parameters change slowly, a cobalt ion forecasting model was proposed based on the grey Fuzzy-LSSVM. Grey accumulation is carried out, which weakens the influences of the random disturbance factors in the primary data sequence and strengthens the regularity of the data. Therefore, the anti-interference ability and the predictive ability of the LSSVM model are strengthened when using the grey-accumulated data as the inputs. The system parameters of the purification process have the characteristic of time-varying. So, different fuzzy weighted values are assigned to different samples collected at different times. The two parameters of LSSVM model are optimized by PSO which has the abilities of fast convergence and global optimization, so that the blindness of artificial choice of model parameters can be avoided. The model is applied to the industrial purification process. The experiment on process data in cobalt removal from zinc electrolyte shows that, the grey fuzzy LSSVM algorithm can commendably predict the cobalt ion concentration, which meets the requirement of cobalt ion concentration prediction.
Key words: LSSVM; PSO; fuzzy weighted; grey accumulation
