以多變數迴歸與類神經網路模式模擬水庫水質

Application of Multivariate Linear Regression and Neural Network Models for Simulating Water Quality in Reservoir

陳偉柏(Wei-Bo Chen) ； 柳文成(Wen-Cheng Liu)

《農業工程學報》 61卷3期 (2015/09) Pp. 73-87

https://doi.org/10.29974/JTAE.201509_61(3).0006

摘要

水質為水庫操作與管理重要考量的因子，而卡爾森優養指標考量的水質變數包括溶氧、總磷、葉綠素a及透明度為決定水庫水質優養程度的指標。本研究以類神經網路模式（包括幅射基底類神經網路及調適性模糊類神經網路）與多變數線性迴歸模式以預測台灣中部德基水庫之溶氧、總磷、葉綠素a及透明度。類神經模式與多變數線性迴歸模式之輸入水質變數則由線性迴歸後之相關係數決定。為評估水質預測之優劣，以統計誤差包括平均絕對誤差、均方根誤差及相關係數判定。水質預測結果顯示，調適性模糊類神經網路模式優於多變數線性迴歸模式及幅射基底類神經網路模式，調適性模糊類神經網路模式可以較準確的預測水庫之溶氧、總磷、葉綠素a及透明度；因此，調適性模糊類神經網路模式可被應用於水庫水質管理之較佳工具。

關鍵字

多變數線性迴歸 ； 幅射基底類神經網路 ； 調適性模糊類神經網路 ； 優養 ； 水質 ； 德基水庫

並列摘要

The water quality is one of the key factors in operation and management of reservoirs. The Carlson index of dissolved oxygen (DO), total phosphorus (TP), chlorophyll a (Chl a), and secchi disk depth (SD) are commonly used as the indicators for determining the status of eutrophication in the reservoirs. In this study, various artificial neural network models (i.e. radial basis function neural network, RBFN and adaptive neuro-fuzzy inference system approach, ANFIS) and multilinear regression (MLR) model were developed to predict DO, TP, Chl a, and SD in the Techi Reservoir of Central Taiwan. The input variables of the neural network and MLR models are determined using the linear regression. The performances using RBFN, ANFIS, and MLR models were evaluated based upon the statistical errors including mean absolute error, root mean square error, and correlation coefficient computed from the measured and model simulated DO, TP, Chl a, and SD values. The results indicate that ANFIS model's performances are superior to those of MLR and RBFN models. Study results show that the neural network using ANFIS model is suitable to predict water quality variables with reasonable accuracy, suggesting that the ANFIS model can be used as a valuable tool for reservoir management in Taiwan.

並列關鍵字

Multivariate linear regression ； Radial basis function neural network ； Adaptive network-based fuzzy inference system ； Eutrophication ； Water quality ； Techi Reservoir