RESULTS
A water meter was installed on the water inlet pipe of the
EWH, as shown in Figure 8, to determine the actual volume
of warm water consumed by usage events to test the thermal
approach. The water meter outputs a pulse for every 0.5 litres
of water used, but requires a flow of more than 2 litres per
minute. The total number of pulses generated in a sampling
interval (typically a minute) is reported to an online server. The
mobile application can typically be used to obtain input from
users in order to classify events. However, in the absence of
user input, the events were classified using water meter data.
Each detected event was classified as small, medium or
large according to the amount of water recorded by the water
meter. Once classified, the relevant flow rate was assigned
to each usage event over the duration detected by the outlet
temperature algorithm. In some instances, events with very
low usage amounts (i.e. less than 0.5 litres) or flow rates (i.e.
less than 2 litres per minute) are detected using the outlet
temperature but are not registered by the water meter. These
events consume little warm water and are therefore classified
as small events.
After the event classification process, a one node model was
used to simulate an EWH with a set temperature of 65 ◦C
for 4 datasets with varying schedule settings spanning several
seasons, as shown in Table IV. Additionally, Table V shows a
summary of the number of each type of usage events included
in the datasets, as well as the total volume of warm water
measured by the water meter (Vtotal Measured) and the
estimated amount of warm water consumed using the outlet
data (Vtotal Estimated). Finally, Table VI contains the results
of the simulations for each dataset using the water meter and
the outlet temperature data.
The overall estimated energy input of the EWH (Einput) was
in good agreement with the measured values for both the water
meter and outlet temperature data. The calculated energy input
error was less than 10% for the first 3 datasets, with dataset 4
yielding inaccurate results. As expected, the water meter data
yielded more accurate results for the energy estimation than
the outlet temperature data for all the datasets. However, the
volumetric estimation of the outlet temperature data was able
to estimate the volume of water consumed within 10 percent
accuracy for the first 3 datasets.