CAJ | 학술논문

通过多种预测方法的综合运用,提出一种依据环境监测数据的近海水质组合预测方法,力求在降低计算难度的同时,提高预测精度.首先,依据入海水体的情况,采用BP神经网络法对近海水质进行因果型预测,预测平均误差为26.46%;其次,采用傅立叶8次级数对近海水质历史数据进行拟合,并将其延伸对近海水质进行类比型预测,平均误差为38.33%;最后,确定近海水质数据符合log-logistic的概率密度函数,提出将上述两种预测结果的概率密度作为其组合权重的近海水质组合预测方法,平均误差降低为21.20%.应用表明,该组合预测方法避免了机理性研究对众多基础数据的要求,原理简单、实用性强,能够为环境管理提供决策支持.
통과다충예측방법적종합운용,제출일충의거배경감측수거적근해수질조합예측방법,력구재강저계산난도적동시,제고예측정도.수선,의거입해수체적정황,채용BP신경망락법대근해수질진행인과형예측,예측평균오차위26.46%;기차,채용부립협8차급수대근해수질역사수거진행의합,병장기연신대근해수질진행류비형예측,평균오차위38.33%;최후,학정근해수질수거부합log-logistic적개솔밀도함수,제출장상술량충예측결과적개솔밀도작위기조합권중적근해수질조합예측방법,평균오차강저위21.20%.응용표명,해조합예측방법피면료궤이성연구대음다기출수거적요구,원리간단、실용성강,능구위배경관리제공결책지지.
Through the using of different prediction methods, a new integrated prediction model for coastal water quality based on monitoring data was proposed, which aimed to reducing calculation difficulty and prediction errors. Firstly, the cause-effect prediction was taken using BP NN （ back-propagation neural network）, whose inputs were data about the incoming water. The average prediction error of this method was 26.46 %. Secondly, the monitoring data serial of each monitoring point was fitted by Fourier serial, after which the fitted Fourier serial was also used for prediction and the error was 38.33%. Finally, the integrated prediction was taken based on the above two prediction results, whose weights were calculated by its log-logistic probability density and average prediction error was reduced to 21.20% . Through application it could be found that the extreme demand on basic data in mechanism studies could be avoided, which made the method in this paper simple, practicable and could be the decision support for environmental management.