Pressure exchanger alarm prediction using KNN in desalination station

Abstract

The desalination process removes dissolved salts from water, and this can include seawater, water from inland seas, mineralized groundwater, and municipal waste water, where the aim is to convert sea water into fresh water that’s safe for drinking, irrigation, and other uses to help for solving the water crisis. The process involves separating salt from water molecules from water, that several desalination methods can accomplish. For example, membrane processes are a class of water treatment technologies that use permeable membranes to separate salts and other impurities from water as it passes through a two main types of membrane processes are reverse osmosis and electrodialysis. This type of technique contain an Energy Recovery Device called Pressure Exchange (PX) which is a flow based operating meaning that the flows of the process are the only factors affecting its operation when having good pre-treatment plant .where its alarms are not easily detected by the SCADA stuff due to codependency of the process flows from every plant equipments, such that there are percentages of one flow to the other that should be respected to avoid alarm cases from the domination of one on another resulting in unbalanced state. We propose in this project, an alarm prediction model that use standardization scalers for pre-processing and machine learning algorithms to investigate and compare the supervised ML binary classifiers choosing the top three. Experimental evaluation yields the best performance using Min-Max scaler as feature scaling technique, and K-Nearest Neighbors as a classi¿er, with an accuracy of 99.13%.