Efficient Energy Management In Water Distribution Through Applying Case-Based Reasoning (CBR).
Aitor Rodriguez, Gabriel Anzaldi, Edgar Rubion
Friday 3 july 2015
14:15 - 14:30h
at Central America (level 0)
Themes: (T) Water resources and hydro informatics (WRHI), (ST) Management support systems and serious gaming
Parallel session: 16H. Water resources - Serious gaming
Currently, Decision Support Tools (DST) for the energy management in water distribution systems have to deal with complex and variable interrelation between large quantities of variables, making difficult the operational usage of empirical models to perform competitive results. Most of the explored solutions are based on complex optimization models designed for controlled situations or simplified water networks due to difficulties in obtaining needed data and the adaptability of the model to changes that frequently undergo in the water network. The work presented tackles this situation by the application of a totally different technological strategy based on a cognitive intelligence that learns from operational situations to achieve incremental energy savings. Specifically, the learning is performed by finding past situations (cases) based on the similarity between past and forecasted daily demands, accomplishing minimum energy consumption. This inference engine, known as CBR, is aimed at recommending suitable pumping scheduling by using a combination of machine learning techniques (clustering, classification and windowing) to accomplish suitable case retrieval according to current water distribution situation. Case retrieval considers lesser energy consumption produced by the expected demand curves to return the case. After this, water manager performs modifications according to his experience and simulations in hydraulic tools to revise the returned case. Then, revised cases are evaluated by the CBR and learning is performed by those cases that improve the stored cases’ solution without compromising the water distribution network. The developed CBR takes part of a DST aimed at using decisional processes knowledge, existing in the water supply distribution chain, to improve daily operations by avoiding water resource mismanagement and inefficient energy strategies. DST takes advantage of the ICT-WatERP architecture to ensure interoperability, information sharing, contributing to address water management issues by being part of the multiple-inference engines core. As a result, the DST shows potentially energy savings around 20% over 2011 demands, against the applied strategies in the Karlsruhe water distribution network by adjusting the energy consumption to the demand curve.
