2. Introduction
1. Hourly (electricity) load curves for the future, for using in TIMES and other models.
2. Impacts of adopting different approaches to timeslicing – 1, 6, 16, 192 timeslices.
Case study using the UK TIMES model.
3. Hourly (electricity) load curves for the future
Long-term decarbonisation costs and strategies are sensitive to the shape of future load curves,
because peakier load curves require higher capital investments in flexible technologies.
1. Future load curve projections for dispatch models have been produced by adding on heat and
transport demands to existing load curves. Yet energy system models suggest a range of system
loads, and hence the overall load curve, could change (e.g. LED lights).
0
10
20
30
40
50
60
Electricityload(GW)
2015
2030
2050
2. Energy system models have low temporal
resolutions and tend to flatten the future load
curve in order to reduce costs. It is unlikely that
that this is possible in reality.
3. It is not clear that either approach identifies load
curves that would occur in extreme weather
periods (e.g. 1-in-20 peak winter demand).
4. Approach in this study
Use high-resolution electricity load curves for demands across the economy in 2010 to derive high-
resolution load curves for the future, based on energy system outputs from a 192-timeslice model.
1. Decompose 2010 electricity loads for each energy service demand.
2. Use these to set TIMES load curves for each demand (with other data for heating demands,
etc.).
3. Examine how these demands might change in the future using an energy system model.
4. Create electricity load curves for 2030 and 2050 based on the 2010 loads and the energy
system model scenario projections.
5. Decomposition of the 2010 load curve
• Annual electricity consumption for each sector based on DUKES data.
• Load curves derived primarily using Elexon load factors for each month, days of the week, and
hours in each day.
• Exelon factors were edited to try to improve the fit to 2010 data, when accounting for temperature
errors, etc.
• Some data also taken from algorithms in the Estimo model.
• The actual load curve is estimated by adding decentralised generation to the National Grid
national demand.
6. Actual and decomposed load curve in 2010
0
10,000
20,000
30,000
40,000
50,000
60,000
70,000
01 Jan 31 Jan 02 Mar 02 Apr 02 May 02 Jun 02 Jul 01 Aug 01 Sep 01 Oct 01 Nov 01 Dec 31 Dec
Load(MW)
Actual demand Modelled Demand
7. Discrepancy between the decomposed and the actual load curve
through the year
-15,000
-10,000
-5,000
0
5,000
10,000
15,000
20,000
25,000
30,000
01 Jan 31 Jan 02 Mar 02 Apr 02 May 02 Jun 02 Jul 01 Aug 01 Sep 01 Oct 01 Nov 01 Dec 31 Dec
Differencebetweendecomposedandactualload(MW)
8. Discrepancy between the decomposed and the actual load curve by
time of day
-15,000
-10,000
-5,000
0
5,000
10,000
15,000
20,000
25,000
30,000
Differencebetweendecomposedand
actualLoad(MW)
9. Average daily load in 2010
0
10,000
20,000
30,000
40,000
50,000
60,000
01 Jan 31 Jan 02 Mar 02 Apr 02 May 02 Jun 02 Jul 01 Aug 01 Sep 01 Oct 01 Nov 01 Dec 31 Dec
Load(MW)
Actual Modelled
10. Discrepancy in 2010 as a function of daily air temperature
-15000
-10000
-5000
0
5000
10000
15000
20000
25000
30000
-10 -5 0 5 10 15 20 25 30
Discrepancy(decomposed-actualload)(MW)
Average daily temperature (degC)
11. 0%
10%
20%
30%
40%
50%
60%
-15,000 -10,000 -5,000 0 5,000 10,000 15,000 20,000 25,000 30,000
Fractionoftotalelectricitydemandinsector
Electricity demand discrepancy (model - actual) (MW)
% Res % Ser % Oth Linear (% Res) Linear (% Ser) Linear (% Oth)
12. Decomposition of the load curve in 2010 – average daily load
0
10,000
20,000
30,000
40,000
50,000
60,000
01 Jan 01 Feb 01 Mar 01 Apr 01 May 01 Jun 01 Jul 01 Aug 01 Sep 01 Oct 01 Nov 01 Dec
UKload(MW)
Residential weather Residential non-weather Service weather Service non-weather Transport
Agriculture Industry Oil refineries Hydrogen
13. Decomposition of the load curve in 2010, for an 80% GHG reduction
but no other constraints – peak daily load
0
10,000
20,000
30,000
40,000
50,000
60,000
70,000
01 Jan 01 Feb 01 Mar 01 Apr 01 May 01 Jun 01 Jul 01 Aug 01 Sep 01 Oct 01 Nov 01 Dec
UKload(MW)
Residential weather Residential non-weather Service weather Service non-weather Transport
Agriculture Industry Oil refineries Hydrogen
14. Future load projections in UK TIMES
STEP 2:
• The decomposed load curves for each demand are used to calibrate UK TIMES energy service
demands.
• A scenario of the future is produced using UK TIMES.
STEP 3:
• The changes in each demand are output to an Excel workbook.
• The decomposed load curves for 2010 are multiplied by the future demands to estimate the future
load curve.
• Micro-CHP and industrial CHP generation is assumed to be heat-led. Other CHP is assumed flat
(i.e. with boilers or other heat-only devices, and heat storage, coping with peaks.
• Some EV demand is assumed flexible. An algorithm first allocates this to overnight low periods,
then any low period, then to flatten the load curve.
15. Decomposition of the load curve in 2050, for an 80% GHG reduction
but no other constraints – average daily load
0
20,000
40,000
60,000
80,000
100,000
120,000
01 Jan 01 Feb 01 Mar 01 Apr 01 May 01 Jun 01 Jul 01 Aug 01 Sep 01 Oct 01 Nov 01 Dec
UKload(MW)
Residential weather Residential non-weather Service weather Service non-weather Transport
Agriculture Industry Oil refineries Hydrogen
16. Decomposition of the load curve in 2050 – peak daily load
0
20,000
40,000
60,000
80,000
100,000
120,000
01 Jan 01 Feb 01 Mar 01 Apr 01 May 01 Jun 01 Jul 01 Aug 01 Sep 01 Oct 01 Nov 01 Dec
UKload(MW)
Residential weather Residential non-weather Service weather Service non-weather Transport
Agriculture Industry Oil refineries Hydrogen
17. 2. IMPACTS OF ADOPTING DIFFERENT APPROACHES TO TIMESLICING
19. Modelling approach
• Base model is annual only.
• Timeslicing data is introduced only by SysSettings and by scenario files.
• Hourly energy demands are aggregated to produce consistent data for each timeslicing approach.
• Different electricity reserve capacity fractions are calculated for timeslicing approach.
• Approach partly made possible by the release of TIMES Version 4.3.2 (March 2019) – which
added support for varying timeslice cycles. Previously, processes defined at timeslice levels that
were not in SysSettings were converted to ANNUAL processes. Now, they move to the next most
aggregated timeslice level.
• For UK TIMES, WEEKLY SEASON.
21. Daily load curve variability by season in 2010
• In each season, each representative day has
the same demands.
• These change between seasons.
• Hence the model ignores demand variability
within seasons 0
10
20
30
40
50
60
70
Spring
0
10
20
30
40
50
60
70
Winter
0
10
20
30
40
50
60
70
Summer
0
10
20
30
40
50
60
70
Autumn
Load(MW)Load(MW)
22. Objective function differences between the four versions
Timeslices ObjZ Difference
1 6,942,001
6 6,903,674 -0.6%
16 6,872,681 -1.0%
192 6,891,317 -0.7%
Lower because a high
winter demand stress
period is not represented?
23. CO2 emissions by sector in 2050
-100
-50
0
50
100
150
200
1 TS 6 TS 16 TS 192 TS
MtCO2e Non-energy use
Upstream
Processing
Hydrogen
Transport
Residential
Industry
Electricity
Services
Agriculture
25. Electricity generation in 2050
0
500
1,000
1,500
2,000
1 TS 6 TS 16 TS 192 TS
PJ/year
Year
Storage output
Net Imports
Fuel cells
Hydrogen
Nuclear
Hydro
Geothermal
Wave
Wind
Tidal
Solar
Biomass CCS
26. Residential heat in 2050
0
200
400
600
800
1,000
1,200
1,400
1 TS 6 TS 16 TS 192 TS
PJ/year
District heat
Standalone water heater
Standalone air heater
Night storage heater
Fuel cell micro-CHP
Stirling engine micro-CHP
Solar heating and heat pump
Solar heating and boiler
Hybrid heat pumps
Heat pumps
Hydrogen boiler
Biomass boiler
Electric boiler
Coal boiler
Oil boiler
Gas boiler
27. Road transport end-use technologies in 2050
0
100
200
300
400
500
600
700
800
900
1 TS 6 TS 16 TS 192 TS
Bvkm
Van Petrol
HGV Hybrid
Hydrogen
HGV EV
Car Petrol
Car EV
28. Comparison of peak daily load curves in 2050
0
20,000
40,000
60,000
80,000
100,000
120,000
01 Jan 01 Feb 01 Mar 01 Apr 01 May 01 Jun 01 Jul 01 Aug 01 Sep 01 Oct 01 Nov 01 Dec
Peakdailyelectricitydemand(MW)
TS1 TS6 TS16 TS192
29. Conclusions
• Finding credible load data is challenging.
• The approach in this study is not ideal, but produces reasonable load curves for higher-resolution
dispatch models. It’s straightforward to explain and the error can be gauged.
• Using higher temporal resolutions does not greatly change the model outputs for this scenario.
• Even with 192 timeslices that are optimised for different renewable output days, the model does
identifies a much lower need for energy storage than in some hourly studies.
• The next step is to repeat the storage analysis, with the same input data, in the hourly highRES
model.
1 TS 6 TS 16 TS 192 TS
1.37 1.37 2.26 1.38
Electricity storage capacity (GW)