Application of BP Neural Network Approach for Cost Estimation of Wastewater Treatment Plants: A Case Study of Taiwan Region

Reliable cost estimation is crucial to the planning process of a wastewater treatment plant (WWTP). Among the developed methods in literatures, not only the assumption of linearity but the existence of a great deal of uncertainty limits the actual application. In this paper, cost estimation of WWTPs in Taiwan region using BP neural network (NN) was investigated. The correlations between cost related variables and total construction cost and plant construction cost were obtained based on 26 collected data sets of design flow rate, influent BOD5 concentration and cost data etc. The study revealed that the proposed NN outperformed linear regression in respect to performance measures such as mean absolute error rate and coefficient of determination. Results from weight interpretation reflected the relative importance of input variables to costs. The NN-based approach can provide an economical and rapid means of cost estimation of WWTP.