Demonstrating the use of data-driven classification
As a part of the Smart Water 4 Europe project, Thames Water is demonstrating the use of data-driven classification which will group together properties with similar patterns of water usage. Deeper knowledge about our customers’ behaviour will have a number of benefits that can help to give the customer a better experience. These include leakage targeting, through understanding the ways in which our customers use water and identifying corresponding demographics to help us estimate unmeasured customers’ consumption more accurately. In addition, identifying patterns of consumption may greatly assist with the targeting of water efficiency efforts, providing customers with appropriate devices and advice that can help them save water, energy and money. This might even inform the creation of dynamic tariffs. Also, having a better knowledge of our customers can help to predict demand at a more local level providing operational benefits, or even just more intelligently infill missing or invalid time series data.
Initial trials using a pre-cleansed set of data have been successful, with clustering algorithms enabling the identification of households that do or do not irrigate as well as those who commute or stay home during the day. Further research is expected to be done on the larger set of smart meter data produced by the demonstration site in Reading.