Powerful AI optimizes desalination RO plants
With the emergence of Artificial Intelligence capabilities, the legitimate question “could reverse osmosis (RO) plants be better piloted with AI?” is now in mind of engineers around the globe. In recent years, RO systems have been highly utilized in industrial and municipal processes. One of the most important operational issues of these systems is membrane fouling/scaling, which leads to high operating costs and environmental impacts. Researchers since 2016-2017 started to dig the topic. The purpose of such research is simply to optimize RO systems’ operation to reduce fouling/scaling events, increase membrane life span, and minimize the system and operational costs. In 2020, the Associate Professor Sara Nazif, used a general regression neural network (GRNN) model. Traditional parameters affecting the performance of a RO unit were optimized by the application of a single-objective optimization model with the total operating cost minimization as an objective function. Pragmatically, the operation’s parameters of a car manufacturer RO unit could be optimized accordingly (inflow rate, inlet pressure, recovery rate). The AI data revealed that the RO unit could work optimally during 208 days (5,000 hrs) without the need for cleaning. The number of search research projects is exploding at the moment, and no doubt that AI will profoundly influence the way desalination technology is run. To discover more:
Earlier in 2016, AI was incorporated into a control system to finetune a RO plant performance with the electrical power available on the island of Gran Canaria (Spain). The target was to enable the RO function with fluctuating power input. Their artifical neural network models gave pretty good results with the restriction of keeping the recovery rate within a certain range. To discover more:
More recently, different AI models were used and compared in the desalination technology field. In 2022, Australian and Spanish engineers looked at optimizing the use of AI for the optimization and prediction of low salinity reverse osmosis performance. They revealed that models can not be straightforward applied, and that finetuning is needed, by even adding additional tools aside of Artificial Neural Network. No doubt that such preliminary research works will guide future steps and stimulate more investigations.