22
The value and challenges of micro- component domestic water consumption datasets Jo Parker Working as part of the ESPRC - ARCC water project with the support of Anglian Water Services (AWS)

The value and challenges of micro- component domestic water consumption datasets Jo Parker Working as part of the ESPRC - ARCC water project with the support

Embed Size (px)

Citation preview

The value and challenges of micro-component domestic water

consumption datasets

Jo Parker

Working as part of the ESPRC - ARCC water project with the support of Anglian Water Services (AWS)

Jo Parker

Study aim

• Examine the sensitivity of long-term water demand micro-components to climate variability and change.

What are micro-components?

Source: Ofwat

Jo Parker

Estimating/forecasting household water demand?

• Traditionally water into supply.

• Complexity of household water demand.

• Micro-component data provides us with the ability to investigate water use at the household scale.

The ‘Golden 100’

Micro-components Socio-economic variables Meteorological variables Other variables

Bath Occupancy rate Minimum temperature (oC) Day of week

Shower Region Maximum temperature (oC) Month of year

Basin Billing type Rainfall (mm) Bank holiday

WC ACORN classification Sunshine (hours per day)

Kitchen sink Rateable value

Washing machine

Dishwasher

External tap

• More than 22million data points.• Too large to handle in excel.• 100 households.

The ‘Golden 100’

Error checking algorithm

1. Basic error checks.

2. Remove large outliers percentile approach.

3. Stratification.

4. Second screening.

5. Apply transformation.

6. Regression analysis.

1. Basic error checks

• Remove gross errors.• Completeness checks.• Dummy variables.• Remove 0l/d PCC.

Sunday 0 0 0 0 0 0

Monday 1 0 0 0 0 0

Tuesday 0 1 0 0 0 0

Wednesday 0 0 1 0 0 0

Thursday 0 0 0 1 0 0

Friday 0 0 0 0 1 0

Saturday 0 0 0 0 0 1

2. Percentile approach

• Remove PCC outliers (0.05% threshold determined via sensitivity testing).

• e.g., one rogue entry purported 98,020 litres/day for a single occupancy household.

3. Stratification

4. Second Screening

• User defined threshold.

• e.g., secondary screening (250l/d threshold) removed values such as 131218l/d in bath usage for a 3 occupancy household.

• Excluding external usage.

5. Transformation

• The Kolmogorov-Smirnov normality test.

• Box-Cox transformation.

Jo Parker

6. Regression – One approach doesn’t fit all

Metered households, East region, single occupancy.

Basin Bath

Jo Parker

Bath (non-zero)

Metered households, East region, single occupancy.

Jo Parker

6. Regression

• Analyse the frequency of usage and non-usage (Logistic regression)• Is this weather, bank holiday, day of the week etc.

sensitive?

• Analyse the volume used (Multiple linear regression)• Is this weather, bank holiday, day of the week etc.

sensitive?

Variables modelled

Observed data input

(subpopulation)

Micro-components

modelled

Explanatory variables used

Metered Bath Mean temperature (oC)

Unmetered Shower Temperature range (oC)

Basin Sunshine (hr)

WC Rainfall (mm)

Kitchen sink 7 day rainfall (mm)

Washing machineRegional soil moisture deficit

index (mm)

Dishwasher Day of week

External tap Month of year

Year

Bank holiday

Occupancy rate

ACORN category

Basin water usage vs. Daily mean Temp.

• Relatively insensitive to Mean T

• What is causing striations?

• Understand peak users (>40l/d)?

Bath water usage vs. Daily mean Temp.

• Relatively insensitive to Mean T

• What is causing striations between 20-60l/d?

• Understand peak users (>80l/d)?

Dishwasher water usage vs. Daily mean Temp.

Metered • Relatively insensitive to

Mean T• Understand peak users

(2 uses per day)?

Unmetered• Slight negative

correlation with Mean T

Metered households

Unmetered households

Shower water usage vs. Daily Mean Temp.

• If we look at peak cluster positive correlation with Mean T.

External water usage vs. Mean Temp.

• Non-linear sensitivity to Mean T

• Where is the tipping point?

Metered households Unmetered households

Jo Parker

Thank you