Keywords
hidden markov model, water end use, water micro-component, flow trace disaggregation, water demand management
Start Date
1-7-2012 12:00 AM
Abstract
It is challenging to disaggregate domestic water consumption flow-trace data into end use event categories for urban water management. Currently, domestic end use studies utilise software and analyst experience to disaggregate flow data into end use events (e.g. faucet, dishwasher, toilet, etc.), which often take an excessive amount of time, particularly for separating a combined event (i.e. a group of single events which occur simultaneously) into different isolated events. To tackle this problem, an existing database of end use events for 252 households located in South-east Queensland (SEQ), Australia was utilised with the goal of automating the flow trace analysis process. A newly developed filtering method in conjunction with the Hidden Markov Model (HMM) technique was applied to disaggregate the combined events. The outcome of this practice is a hybrid model which allows great separation accuracy (average of 88%) of a combined event into many different single events. Future work will incorporate this model with existing methods in single event classification to fulfil the development of an automatic system to disaggregate domestic water consumption flow-trace data into end use event categories. Moreover, validation of its accuracy will be examined through independent testing with 20 selected combined samples.
Analysis of simultaneous water end use events using a hybrid combination of filtering and pattern recognition techniques
It is challenging to disaggregate domestic water consumption flow-trace data into end use event categories for urban water management. Currently, domestic end use studies utilise software and analyst experience to disaggregate flow data into end use events (e.g. faucet, dishwasher, toilet, etc.), which often take an excessive amount of time, particularly for separating a combined event (i.e. a group of single events which occur simultaneously) into different isolated events. To tackle this problem, an existing database of end use events for 252 households located in South-east Queensland (SEQ), Australia was utilised with the goal of automating the flow trace analysis process. A newly developed filtering method in conjunction with the Hidden Markov Model (HMM) technique was applied to disaggregate the combined events. The outcome of this practice is a hybrid model which allows great separation accuracy (average of 88%) of a combined event into many different single events. Future work will incorporate this model with existing methods in single event classification to fulfil the development of an automatic system to disaggregate domestic water consumption flow-trace data into end use event categories. Moreover, validation of its accuracy will be examined through independent testing with 20 selected combined samples.
