Urban Behavioural Demand

Introduction

The behavioural model applied in the Behavioural Water Use node provides simulation of water use using the Behavioural End-use Stochastic Simulator (BESS) of Thyer et al. (2011). BESS stochastically simulates individual end-uses (outdoor, shower, washing machine, toilet, tap etc) at the household scale at sub-daily time steps using algorithms that probabilistically simulate an individual household’s use of common household water-using appliances. The conceptual framework for BESS as applied in the Urban Developer Plugin, is as follows:

Figure 1. BESS conceptual framework (after Thyer et al, 2009)

For indoor water uses, the water use simulations for each household are based on the type of water-using appliance and the occupancy of that household. The difference from the average daily model is that instead of specifying an average daily volume and percentages for each end use, the user specifies the water-using appliance and household occupancy and BESS simulates the water for each individual end-use using the in-built parameters for the water use event dynamics. The types of water-using appliances for each end-use are configured in the appliance types. This enables users to simulate the effects of changes in the uptake of water efficient appliances.

The appliances and occupancy for each household can be specified in several different ways, configured under Edit>>Urban Developer Options:

1. Appliances – where the type of appliances and occupancy for each house are fixed by the user
2. Sampled appliances and occupancy – where the user inputs probability distributions for the occupancy and water-using appliance. At the start of an Urban Developer Scenario run the occupancy and appliance type is randomly sampled for each house in the Behavioural Water Use nodes.
3. Average appliance demand –  rather than sampling from the probability distribution of water consumption for a usage event for each appliance, the average water use is selected. 
   

Further details on these configurations are given below in Behavioural Model Configuration.

The statistical distributions used by BESS to generate the sub-daily indoor water use are based on the end-use study of Roberts et al. (2005). Users are encouraged to check that the predicted water use statistics provided by BESS match their expectations.

BESS uses a diurnal pattern based on Roberts et al. (2005) to vary the probability of water use events occurring throughout the day. 

For outdoor water use, the behavioural model uses a time series or monthly varying pattern of average daily values, which can be input by the user - similar to the average daily model. For the sub-daily outdoor water use, the daily values are evenly distributed throughout the day. 

Behavioural Model Configuration

The appliances and occupancy for each household can be specified in several different ways, configured under Edit>>Urban Developer Options.

Appliances

The behavioural water use model configuration with Appliances allows the user to set appliance types for each indoor (showers, washing machines, dishwashers, taps and toilets) and outdoor (pool) end use under the Appliances option in Urban Settings for each water use node. Household occupancy is specified on the water use node interface. The water consumption for individual water use events is still sampled from probability distributions, as described in Capturing Temporal Variability in Indoor Water Use Events.

Using the behavioural model with the Appliances option is the recommended approach when simulating a smaller number of houses (< 400), and allows the user to specify exactly the type of appliances and occupancy for each house. The indoor end-use appliance types and occupancy is set specifically for each water use node. Refer to the Node properties for details.

The Appliances model is enabled by disabling the Sampled Appliances & Occupancy menu.

Sampled appliances and occupancy

The behavioural water use model with Sampled appliances and occupancy allows you to specify random sampling of occupancy and appliance types, for each household, from statistical distributions. 

Using the behavioural model with Sampled appliances and occupancy is only recommended when you are simulating a larger number of households (> 400) and you do not want to set the appliance types /occupancy for each house.

When using the Sampled appliances and occupancy model, the occupancy and appliance type node properties do not need to be configured. Other properties, such as the number of households, supply and discharge preferences do need to be configured. Refer to the Node properties for details.

The Sampled appliances and occupancy model is enabled and disabled using the Sampled Appliances & Occupancy menu.

Details of the Sampled appliances and occupancy algorithm are provided in Thyer et al. (2009) and references cited therein, a summary is provided below.

Capturing Spatial Variability in Occupancy and Appliances

The household size and the type of water using appliances varies from household to household. To capture this spatial variability the household size and appliance type for each type of water use event is randomly sampled for each household from a probability distribution based on the proportion of household sizes/appliance types. The probability distributions are specified in the Sampled Appliances & Occupancy Menu.

Capturing Temporal Variability in Indoor Water Use Events

Individual end-use events are generated in a two-stage process. The first stage determines whether a given water use starts in a particular time step, and is based on the probability of occurrence, for a particular end-use. The second stage tracks the subsequent behaviour of that water use over the following time steps, and is dependent on the end-use event volume and water usage pattern.

In general, the probability of occurrence for each event is a function of the frequency of events per person per day, the household size and a diurnal factor. The diurnal factor converts the frequency of events per day into the probability of an event occurrence in a given minute. The water use event volume and the water usage pattern is dependent on the type of water use event and the appliance type. In general, the end-use event volume is sampled from a probability distribution, and the water usage pattern is applied to this volume to produce a water use time series.

For each appliance, the frequency of events per person per day, and the probability distributions of water consumption for an event, are defined in the Appliances menu.

Sample size requirements

If you specify sampled appliances and occupancy, you must specify a minimum number of houses in order to provide a statistically-reliable sample of the household occupancy and appliance type.

Due to random sampling in any particular Urban Developer run, the percentage of houses actually sampled for each household occupancy and appliance type (the sampled percentage) will not be exactly the same as what you specify (the true percentage).

For example - if you specify only 10 houses, with a true probability of 0.2 for 3-Star showers and 0.8 for 1-Star showers, the chance of the sampled percentage being very different (eg: 0.5 3-Star and 0.5 2-Star) is much greater than with a large sample size.

The recommended minimum number of 400 houses ensures that for true percentages greater than 10%, the sampled percentage of houses is within 30% of the true percentage for 90% of the time.

You can specify a number of houses smaller than 400, but should be aware that the sampled percentage of household occupancy and appliance types may be different than the percentage you originally specified.

Average Demand

This mode can be used in conjunction with Fixed or Sampled appliances and occupancy and is activated by the 'Use Average Demand' checkbox available in the Behavioural Water Use node user interface.

References

Micevski, T., Thyer, M., Kuczera, G. (2011) A Behavioural Approach for Household Outdoor Water Use Modelling. Paper submitted to Water Resources Research (April 2011).

Roberts, P. (2005) 2004 Residential End Use Measurement Study, Final Report: Yarra Valley Water, Victoria.

Thyer, M. A., Duncan, H., Coombes, P., Kuczera, G., & Micevski, T. (2009) A probabilistic behavioural approach for the dynamic modelling of indoor household water use. In H2009: 32nd Hydrology and Water Resources Symposium: Adapting to Change, 30 November - 3 December 2009, Newcastle, Australia (p. 1059).

Thyer, M., Micevski, T., Kuczera, G., and Coombes, P. (2011) A Behavioural Approach to Stochastic End Use Modelling. Paper presented at Oz Water, 9-11 May 2011, Adelaide.

Acknowledgements

This material has been adapted from:

eWater Cooperative Research Centre (2011) Urban Developer User Guide: Urban Developer v1.0.0, eWater Cooperative Research Centre, Canberra, 29 June 2011. ISBN 978-1-921543-40-1