基于 ELMR-SVMR 的海水水质预警模型研究

摘要

海水水质预警对海洋环境监控和保护有着重要的意义。极限学习机回归和支持向量机回归(extreme learning machine regression-support vector machine regression,ELMR-SVMR)集成预警模型采用预测与评价相结合的方式,实现对未来海水富营养化状况综合预警的目的。其中 SVMR 方法用于构建海水富营养化评价模型；ELMR 用于对未来一段时间的水质状况进行综合预测,ELMR 的预测结果作为评价模型的输入变量。集成模型的可靠性直接影响预警的有效性,ELMR-SVMR 的可靠性通过分析 ELMR 预测误差对 SVMR 评价模型的灵敏度得到。将各参变量的预测误差结果作为评价模型的灵敏度影响参量,通过灵敏度计算可获得对 ELMR-SVMR 模型的灵敏度评价。通过与其他方法的实验对比及分析,验证了该区域范围内所建 ELMR-SVMR 预警模型的有效性,为探索建立有效的海水水质预警模型提供了一种新途径。%Forewarning of seawater quality has important significance for monitoring and protecting of marine environment.ELMR-SVMR (extreme learning machine regression-support vector machine regression)integrated model takes the ways of combining the prediction and evaluation to achieve the purpose of comprehensive forewarning for the state of seawater eutrophication. Therein SVMR method is used to establish the evaluation model,ELMR method is used to comprehensively predict the seawater quality state for a future time,and the result of the prediction by ELMR is used as the input variable of the evaluation model. The reliability of integrated model directly affects the effectiveness of the forewarning.The reliability of ELMR-SVMR model can be obtained by analyzing the sensitivity regarding the prediction error of ELMR to the evaluation model of SVMR.The prediction errors of all the parameters can be selected as the influence parameters of sensitivity of the evaluation model.The sensitivity evaluation of ELMR-SVMR model can be achieved by the sensitivity calculation.Compared with other methods by experiment and analysis, the effectiveness of the forewarning model of ELMR-SVMR is validated in this scope of sea area.It provides a pathway to establish the forewarning model of seawater quality effectively.