Documentation

This is the documention of the High-Resolution Energy demand model (HIRE) of the ITRC-MITRAL framework.

1. Introduction

HIRE allows the simulation of long-term changes in energy demand patterns for the residential, service and industry sector on a high temporal and spatial scale. National end-use specific energy demand data is disaggregated on local authority district level and a bottom-up approach is implemented for hourly energy demand estimation for different fuel types and end uses.Future energy demand is simulated based on different socio-technical scenario assumptions such as technology efficiencies, changes in the technological mix per end use consumptions or behavioural change. Energy demand is simulated in relation to changes in scenario drivers of the base year. End-use specific socio-technical drivers for energy demands modelled where possible on a household level.

The methodology is fully published in Eggimann et al. 2018: TBD, DOI.

2. Overview

2.1 Simulated end uses and sectors

Within HIRE, residential, service and industrial energy demands are modelled for three main sub-modules. For the United Kingomd, each sub module modelles end uses and sectors taken from BEIS (2016). A simplified overview is given in Figure 1.

_images/Fig_01_model_overview.jpgModel overview Figure 1: Simplified overview of modelled end uses and sectors

A complete list of all modelled end uses and sectors is provided in Table 2.1. As HIRE is highly modular, all end uses or sectors can be replaced with different end uses depending on available energy consumption statistics.

More information on sectors and their fuel inputs can be found here.

Residential Service Industry
     
End use End use End use
space heating space heating space heating
water heating water heating drying and separation
lighting lighting lighting
cooking catering motors
cold cooling and humidification refrigeration
home computing ICT equipment compressed air
consumer electronics fans high temperature processes
wet small power low temperature processes
  other gas other
  other electricity  
  cooled storage  
     
Sector Sector Sector
  Community, arts and leisure Other mining and quarrying
  Education Manufacture of food products
  Emergency Services Manufacture of beverages
  Health Manufacture of tobacco products
  Hospitality Manufacture of textiles
  Military Manufacture of wearing apparel
  Offices Manufacture of leather and related products
  Retail Manufacture of wood related products
  Storage Manufacture of paper and paper products
    Printing and publishing of recorded media
and other publishing activities
    Manufacture of chemicals and chemical products
    Manufacture of basic pharmaceutical products
and pharmaceutical preparations
    Manufacture of rubber and plastic products
    Manufacture of other non-metallic mineral products
    Manufacture of basic metals
    Manufacture of fabricated metal products,
except machinery and equipment
    Manufacture of computer, electronic and optical products
    Manufacture of electrical equipment
    Manufacture of machinery and equipment n.e.c.
    Manufacture of motor vehicles,
trailers and semi-trailers
    Manufacture of other transport equipment
    Manufacture of furniture
    Other manufacturing
    Water collection, treatment and supply
    Waste collection, treatment and disposal
activities; materials recovery

Table 2.1: Complete overview of modelled submodels, enduses and sectors for the United Kingdom.

2.2 Work flow

The main working steps as outlined in Figure 2 are explained in full detail in Eggimann et al. (2018).

_images/Fig_02_model_flow.jpgModel overview Figure 2: Overview of modelled end uses and sectors

2.2.1 I Step - Spatial disaggregation

As a starting point, national based energy demand consumption statistics are disaggregated to regional energy demands. This is performed with disaggregation factors such as population, heating degree day calculations or employment statistics. In case regional data is available, this step can be omitted.

2.2.2 II Step - Demand simulation

Total energy demand of a (simulation) year (D^{tot}_{y}equation) is calculated over all regions (r), sectors (s), end-uses (e), technologies (t) and fuel-types (f) as follows:

D^{tot}_{y} = D^{tot}_{by} + \sum _{r} \left( \sum _{s} \left( \sum _{e} \left( \sum _{t} \left( \sum _{f} \left( D^{sd}_{y} + D^{\eta}_{y} + D^{tecs}_{y} + D^{climate}_{y} + D^{behaviour}_{y}  \right) \right) \right) \right )\right)equation

D^{sd}_{y}equation: Demand change related to change in scenario driver (SD)

D^{\eta}_{y}equation: Demand change related to change in technology efficiency

D^{tecs}_{y}equation:Demand change related to change in technology mix

D^{climate}_{y}equation: Demand change related to change in climate

D^{behaviour}_{y}equation: Demand change related to change in behaviour (e.g. smart meter, base temperatures)

Energy demand change in relation to the base year (by) for end-use specific scenario drivers is defined as follows:

D^{sd}_{y} = D^{tot}_{by} * \frac{sd_{y}}{sd_{by}}equation

The individual terms are fully described in Eggimann et al. (2018).

For the residential and service sub model, SD values are calculated based on a dwelling stock where scenario drivers are calculated either for dwellings or a group of dwelling (e.g. dwelling types).

Scenario Driver Enduse(s)
Floor Area Space heating, lighting
Heating Degree Days Space heating
Cooling degree Days Cooling and Ventilation
Population Water heating, cooking, appliances ...
GVA Enduses in industry submodel, appliances

Table 1.1: End-use specific scenario drivers for energy demand

2.2.3 III Step - Temporal disaggregation

In order to disaggregate annual demand for every hour in a year, different typical load profiles are derived from measuremnt trials. Load profile are either used to disaggregate total demand of an end use, sector or technology specific end use energy demand.

Only the profile (i.e. the ‘shape’) is loaded as an input into the model. In case of temperature dependent load profiles, the daily load profils are calculated with help of heating and cooling degree days (HDD, CDD).

For different heating technologies, load shares are derived from the following sources:

  • Heat pumps load profile Based on nearly 700 domestic heat pump installations, Love et al. (2017) provides aggregated profiles for cold and medium witer weekdays and weekends. The shape of the load profiles is derived for a working, weekend and peak day.
  • Boiler load profile Load profiles for a typicaly working, weekend and peak day are derived from data provided by Sansom (2014).
  • Micro-CHP Load profiles for a typicaly working, weekend and peak day are derived from data provided by Sansom (2014).
  • Primary and secondary electric heating The load profiles are based on the Household Electricity Survey (HES) by the Department of Energy & Climate Change (DECC, 2014).
  • Cooling The daily load profiles for service submodel cooling demands are taken from Dunn and Knight (2005).

2.3 Main Classes

Figure 3 provides an overview of how all major classes relate to each other for generating a sub model to simluate different end uses. For every simulated region, the geographically closeset WeatherRegion is searched and linked to the Region class. Every WeatherRegion class contains a technology and load profile stock as depending on temperatures, the efficiencies of the technologies and load profiles may differ.

_images/Fig_03_ULM_main_classes.jpgImage of model integration Figure 3: Interaction of all major classes of HIRE

3 Methods

In this section, selected aspects of the methodology are explained in more detail.

3.1 Generic dwelling stock

A generic dwelling model is implemented in HIRE which can be used in case no external dwelling stock model is available to generate provide the necessary dwelling inputs. An abstracted dwelling representation is modelled with the following assumptions for the base year configuration:

• total population • total floor area • dwelling type distribution • age distribution of dwelling types

However, instead of modelling every individual building, an abstracted dwelling representation of the the complete dwelling stock is modelled based on different simplified assumptions. The modelling steps are as follows for every Region (see Figure XX for the detailed process flow):

  1. Based on base year total population and total floor area, the floor area per person is calculated (floor_area_pp). The floor area per person can be changed over the simulation period.
  2. Based on the floor area per person and scenario population input, total necessary new and existing floor area is calculated for the simulation year (by subtracting the existing floor area of the total new floor area).
  3. Based on assumptions on the dwelling type distribution (assump_dwtype_distr) the floor area per dwelling type is calculated.
  4. Based on assumptions on the age of the dwelling types, different Dwelling objects are generated. The heat loss coefficient is calculated for every object.
  5. Additional dwelling stock related properties can be added to the Dwelling objects which give indication of the energy demand and can be used for calculating the scenario drivers.

Note: Current floor area data can be calculated based on floor area data from MasterMap data in combination with the address point dataset by the Ordonance Survey.

_images/Fig_XX_dwelling_stock.jpgDwelling model Figure 4: Modelling steps of the residential dwelling module

3.2 Demand side management

Dirunal demand side responses per end use are modelled with help of load factors (Petchers, 2003). For every end use, a potential (linear) reduction of the load factor over time can be assumed with which the load factor of the current year is calculated l_{cy}equation. With help l_{cy}equation, and the daily average load of the base year l^{av}_{by}equation, the maximum hourly load per day is calculated as follows:

l_{cy}^{max} = \frac{l_{by}^{av}}{lf_{cy}}equation

For all hours with loads higher than the new maximum hourly load, the shiftable load is distributed to all off-peak load hours as shown in Figure XY.

_images/Fig_XX_peak_shifting.jpgPeak shiting Figure 5: Shifting loads from peak hours to off-peak hours based on load factor changes.

3.3 Technologies, technological diffusion

For every end use, technologies can be defined. For the UK application, the defined technologies are listed in Table 2.

End use Technologies
Wet Washing machine, tubmle dryer, dishwasher, washer dryer
Cooking Oven, Standard hub (different fueltypes), Induction hob
Cold Chest freezer, Fridge freezer, Refrigerator, Upright freezer
Lighting Light bulb (incandescent), Halogen, Light saving, Fluorescent, LED
Heating Boiler (different fueltypes), Condensing boiler, ASHP, GSHP, Micro-CHP, Fuel cell, Storage heating, Night storage heater, Heat network generation technologies (CHP,...)

Table 2: Technology assignement to end uses

3.4 System-of-system embedding

The main function of HIRE within the MISTRAL modelling framework is to provide energy demands to the energy supply model developped within this framwork (see Figure 6).

_images/Fig_XX_supply_demand_linkage.jpgImage of model integration Figure 6: Visualisation of System-of-system model integration

This can be achieved in two ways, either demands are provided for an optimised or constrained model mode. In the constrained mode, heat demand is allocated to certain heating technologies and converted into respective fuel carriers. For the unconstrained mode, the energy supply model assigns technologies to meet the given heat demand, which is directly provided from the demand model to the supply model (see Figure 7)

_images/Fig_XX_constrained_optimised_modes.jpgTwo modes Figure 7: Interaction of energy demand and supply model

4 Data sets

This section provides an overview of all used datasets in HIRE and necessary data preparation.

4.1 National Energy Statistics

The base year final energy consumption of the UK in terms of fuels and technology shares for the residential, service and industry sectors is taken from the Department for Business, Energy and Industrial Strategy (BEIS, 2016). National final energy data is provided for different fuelt types in the unit of ktoe (tonne of oil equivalents). All energy unit conversions are based on the unit converter by the International Energy Agency (IEA, 2017).

Even though sub-national data for gas and electricity consumption are available from BEIS, they do not provide the same level of differentiation with respect to distinguishing between different endues or sectors.

4.2 Household Electricity Servey (DECC 2014)

The Household Electricity Survey (HES) by the Department of Energy & Climate Change (DECC 2014) is the most detailed monitoring of electricity use carried out in the UK. Electricity consumption was monitored at an appliance level in 250 owner-occupied households across England from 2010 to 2011. From the provided spreadsheet for users (DECC 2017), peak load profiles and averaged monthly data are available for a sample of 250 households on a 10 minutes basis. The data are aggregated to hourly resolution to derive hourly load profiles for the full year.

The load profiles for different residential enduses for different daytypes (weekend, working day) are taken from a 24 hour spreadsheet tool.

4.3 Carbon Trust advanced metering trial

The Carbon Trust (2007) advanced metering trial dataset contains hourly electricity and gas demand measurements for different service (business) sectors from over 500 measurement sites broken down in sectors. The Carbon Trust data does not allow distinguishing between different end uses within each sector. According to the dominant fuel type of each end use,either aggregated gas or sector specific load profiles are therefore assigned. For ‘water heating’, ‘space heating’ and the ‘other_gas_enduse’, all gas measurements across all sectors are used, because the sample size was too small. For all other end uses, sector specific electricity load profiles are assigned. The provided sectors in the data trial do not fully correspond to the ECUK sectors (see Table 5) and where a sector is missing,the aggregated profiles across all sectors are used.

ECUK Data (Table 5.05) Carbon Trust Dataset
Community, arts and leisure Community
Education Education
Emergency Services Aggregated across all sectors
Health Health
Hospitality Aggregated across all sectors
Military Aggregated across all sectors
Offices Offices
Retail Retail
Storage Aggregated across all sectors

Table 5: Matching sectors from the ECUK dataset and sectors from the Carbon Trust dataset

Yearly load profiles are generated based on averaging measurements for every month and day type (weekend, working day). Only the yearly load profile for space heating is generated based on HDD calculations.

Data preparation of the raw input files was necessary:

  • Half-hourly data was converted into hourly data
  • Within each sector, only datasets containing at least one full year of monitoring data are used, from which only one full year is selected
  • Only datasets having not more than one missing measurement point per day are used
  • The data was cleaned from obviously wrong measurement points (containing very large minus values)
  • missing measurement points are interpolated

4.4 Temperature data

To calculate regional daily hourly load heating profiles, hourly temperature data are used from the UK Met Office (2015) and loaded for weather stations across the UK.

  • The station ID can be retreived here
  • Metadatda of raw data can be found here

Alternatively, daily minimum and maximum temperature data can be downloaded from the weather@home dataset.

4.5 Census data

Employment statistics from the census (Office for National Statistics 2011) are used to disaggregate industry related energy demands for different end uses and sectors.

  • The data download link can be found here

Literature

BEIS (2016): Energy consumption in the UK (ECUK). London, UK. Retrieved from: https://www.gov.uk/government/collections/energy-consumption-in-the-uk

Carbon Trust (2007). Advanced metering for SMEs Carbon and cost savings. Retrieved from https://www.carbontrust.com/media/77244/ctc713_advanced_metering_for_smes.pdf

DECC (2014) Household Electricity Survey. Retrieved from:https://www.gov.uk/government/collections/household-electricity-survey

Dunn, G. N. and Knight, I. P. (2005) ‘Carbon and Cooling in Uk Office Environments’, in Proceedings of the 10th International Conference Indoor Air Quality and Climate. Beijing, China, p. 6. Available at: http://www.cardiff.ac.uk/archi/research/auditac/pdf/auditac_carbon.pdf.

Love, J., Smith, A. Z. P., Watson, S., Oikonomou, E., Summerfield, A., Gleeson, C., … Lowe, R. (2017). The addition of heat pump electricity load profiles to GB electricity demand: Evidence from a heat pump field trial. Applied Energy, 204, 332–342. https://doi.org/10.1016/j.apenergy.2017.07.026

Petchers, N. (2003) Combined heating, cooling & power handbook: technologies & applications. 2 edition. Lilburn: Fairmont Press.

Sansom, R. (2014). Decarbonising low grade heat for low carbon future. Imperial College London.

UK Met Office (2015): ‘MIDAS: UK hourly weather observation data’. Centre for Environmental Data Analysis. Retrieved from: http://catalogue.ceda.ac.uk/uuid/916ac4bbc46f7685ae9a5e10451bae7c.

Code Overview

This section provides and overview how the model code is stored.

Some model input data used to configure the model is stored in the config_data folder (i.e. load profiles of technologies, fuel input data for the whole UK). All additional data necessary to run the model needs to be stored in a local folder (data_energy_demand).

The python scripts are stored in the following folders:

  • assumptions | Model assumptions which need to be configured
  • basic | Standard functions
  • charts | Functions to generate charts
  • cli | Script to run model from command line
  • config_data | Configuration data (e.e.g technologies, fuel)
  • dwelling_stock | Dwelling stock related functions
  • geography | Space related functions (e.g. weather region)
  • initalisations | Initialisation scripts
  • plotting | All plotting functionality
  • profiles | Load profile related functionality
  • read_write | Reading and writing functions
  • script | Scripts
  • technologies | Technology related funcionality
  • validation | Validation related scripts

Important single python scripts:

  • main.py - Function to run model locally
  • enduse_func - Main enduse function
  • model - Main model function

Within the code, different abbreviations are consistenly used across all modules.

s:         Energy service
rs:         Residential Submodel
ss:         Service Submodel
is:         Industry Submodel

by:         Base year
cy:         Current year
lp:         Load profile
dw:         Dwelling
p:          Decimal (100% = 1.0)
pp:         Per person
lu:         Look-up

e:          Electricitiy
g:          Gas

h:          Hour
d:          Day
y:          Year
yh:         8760 hours in a year
yearday:    Day in a year ranging from 0 to 364

hp:         Heat pump
tech:       Technology
temp:       Temperature

Different endings are appended to variables, depending on the temporal resolution of the data. The following abbreviations hold true:

y: Total demand in a year
yd: ‘Yearly load profile’ - Profile for every day in a year of total yearly demand(365)
yh: ‘Daily load profile’ - Profile for every hour in a year of total yearly demand (365, 24)
dh: Load profile of a single day
y_dh: Daily load profile within each day (365, 25). Within each day, sums up to 1.0