Seawater desalination energy efficiency and emissions reduction at Gold Coast Desalination Plant

Abstract

This paper investigates the implementation of Artificial Intelligence (AI) and Machine Learning (ML) in the operation of the Gold Coast Desalination Plant (GCDP) to optimise reverse osmosis (RO) membrane efficiency, reduce energy consumption, emissions, and operational costs. The study focuses on using historical and current RO operational data to train and utilise an AI algorithm that generates daily operational setpoints. These setpoints, namely RO Recovery (%) and Permeate Flow (m3/h), directly impact pump power consumption and energy requirements. The collaboration between Veolia and a cleantech start-up involves multiple phases, including system review, performance tests, and iterative optimisation. Three (3) performance tests were conducted at GCDP, revealing unique challenges and findings. Performance Test 1 demonstrated energy savings despite rapidly increasing cartridge filters differential pressure (dP) in the test period. Performance Test 2 encountered underperformance of an energy recovery device (ERD), resulting in increased feed pressure and with it, power consumption. Performance Test 3 highlighted the impact of fluctuating feedwater temperature on energy consumption. Overall, the AI model showcased its ability to recommend setpoints within operational limits and achieve required permeate flow with reduced energy, even under unfavourable conditions. The initiative at GCDP resulted in an estimated 1.1% energy reduction, corresponding to approximately 0.023 kWh/m3 and the potential to save up to 1044 MWh of energy and 762 tonnes of CO2-e per annum. The study suggests further exploring AI and ML applications to optimise treatment plant operations, including varying permeate flow setpoints and improving data transfer and setpoint implementation methods. These advancements hold promise for enhanced energy efficiency and cost reduction in desalination plants worldwide.