Data-driven minimization of pump operating and maintenance cost

A data-driven model for scheduling pumps in a wastewater treatment process is proposed. The objective is to minimize the cost of pump operations and maintenance. A neural network algorithm is applied to model performance of the pumps using the data collected at a municipal wastewater treatment plant. The discrete-state Markov process is utilized to develop a model of maintenance decisions. The developed pump performance and maintenance models are integrated into a scheduling model. A hierarchical particle swarm optimization algorithm is designed to solve the proposed scheduling model. The concepts developed in this paper are illustrated with two case studies.