论文部分内容阅读
为解决传统方法水质预测精度低、鲁棒性差等问题,提出了基于小波分析(WA)、人工蜂群(ABC)优化加权最小二乘支持向量回归机(WLSSVR)的工厂化育苗溶解氧组合预测模型(WA-ABC-WLSSVR模型).该模型采用小波分析对原始非平稳溶解氧时间序列数据进行多尺度特征提取,通过加权最小二乘支持向量回归机对不同尺度下的溶解氧数据子序列分别建模,利用改进人工蜂群优化算法(ABC)对各分量序列WLSSVR模型参数进行组合优化,最后叠加各尺度下的预测结果.运用该模型对工厂化育苗溶解氧进行预测,并与BPNN、标准LSSVR、WAACO-LSSVR、WA-PSO-LSSVR等模型对比分析,结果表明,该溶解氧预测模型具有较高的预测精度和泛化能力.
In order to solve the problems of low accuracy and poor robustness of traditional water quality prediction methods, a combination of WLSSVR and Wavelet Analysis (WA) and artificial bee colony (ABC) optimized weighted least square support vector regression (WLSSVR) (WA-ABC-WLSSVR model) .This model uses wavelet analysis to extract multiscale features of the original non-stationary dissolved oxygen time series data. Weighted least square support vector regression (ABC) was used to optimize the WLSSVR model parameters of each component sequence, and then the prediction results at all scales were superposed.Using the model, the model was used to predict dissolved oxygen in plantation nursery and compared with BPNN, LSSVR, WAACO-LSSVR and WA-PSO-LSSVR. The results show that the DO prediction model has high prediction accuracy and generalization ability.