A robust optimization approach for real-time multiple source drinking water blending problem

Although many optimization methods can be applied to real-time multiple source drinking water blending problems, the field still lacks an approach to rapidly produce a robust optimal solution by simultaneously optimizing multiple objectives. This paper develops a fuzzy multiple response surface methodology (FMRSM) to achieve this objective. In the FMRSM, experimental data are fitted to mean response surface models while the residuals (the error between the predicted response of the mean model and the measured data of the real system) are fitted to standard deviation models. Fuzzy linear programming using the min-operator approach is applied to optimize the multiple objectives. Six scenarios are designed based on data from a real-time multiple source drinking water blending operation. The results show the FMRSM is a robust, computationally efficient optimization approach. The FMRSM could be extended to other real-time multi-objective non-linear optimization problems.