Demand Forecasting - University of Readingblogs.reading.ac.uk/tsbe/files/2017/04/1.1_J_Caplin... ·...

Place your chosen image here. The four corners must just cover the arrow tips. For covers, the three pictures should be the same size and in a straight line.

Demand Forecasting

Jeremy CaplinEnergy Forecasting ManagerNational Grid


National Grid Overview

3

Facts:

Total employees - 24,000

Largest UK utility

2nd largest US energy utility

FTSE 25 company

National Grid Overview50:50

Transmission

UKUS

Distribution

Electricity Gas

4

National Grid – UK – Electricity

Transmission

7,200 km overhead line 1500 km underground cable 336 substations

5

National Grid – US – Electricity

Transmission

14,355 km line169 km cable520 substations

Distribution andGeneration116,636 km circuit644 substations3.5 million customers50 generation plants

Electricity National Control Centre

TO

TO

TO/SO

GBSO

Owns, builds and maintains assets

Operates the system Provides customer interface

TO/SO Code to define interfaces such as outage planning and investment decisions

GBSO

Balances the GB System Configures the GB

Transmission System Operational planning Connection and Use of System

Agreements with Generators, Suppliers and Distribution companies

GB charging and billing

Electricity Operations - who is responsible for what?

Transmission Owner (TO)

System Operator (SO)

OFFTO’s

System Operator – Two key roles

Managing electricity flows across the networks

Balancing the generation and demand

Demand Generation forced outages / trips Transmission circuit faults Market participants positions Weather events

Whilst ensuring Security Equipment ratings Voltage and frequency tolerance Cost minimisation

Managing uncertainties

Outage Planning and Co-ordination

System not geographically balanced for generation and demand.

Bottle-necks arise that requires power flow to be “constrained”.

Solutions include demand transfers, intertripping, re-switching, QB tapping etc.

2200MW

6000MW

7000MW

11000MW


Energy Forecasting

National Demand

Unmetered Generation

What is Demand

National Demand suppressed by unmetered generation

20000

25000

30000

35000

40000

45000

0

5000

10000

15000

20000

25000

embedded wind Embedded solar 20150413 Demand 20150413 Virtual Demand

15

National Demand: min by min fluctuations

What Affects National Demand?

Time of Year

Time of Day

Day of Week

Weather

Unmetered Generation

Bank Holidays

School Holidays

TV

Special Events

Reaction to power price

Connection charge minimisation

National Demand: time of year effect

National Demand: variability during a week (GMT)

National Demand: variability during a week (BST)

Illumination

Dem

and

(MW

)

Dull: High Demand

Bright: Low Demand

The Impact of Weather: Illumination

Cold: High Demand

Hot: Quite High Demand

Temperature

Dem

and

(MW

)The Impact of Weather: Temperature

Strong Wind: High DemandLow Wind: Low Demand

Wind Speed

Dem

and

(MW

)The Impact of Weather: Cooling Power of the Wind

The Impact of Weather: Rain

25

Temperature(1°C fall in cold conditions)

Cloud cover(clear sky to thick cloud)

Precipitation(no rain to heavy rain)

Temperature(1°C rise in hot conditions)

+ 500 MW

+ 500 MW

+ 1,000 MW

+ 1,500 MW

Cooling power(10 mph rise in cold conditions) + 1,000 MW

The Impact of Weather Some Numbers

GB National Demand: Bank Holiday effect

26

GB National Demand: Xmas effect

27

GB National Demand: Triad avoidance / CDM

28

Statistical Models (standard linear regression)

Updated twice a year

Separate models for each Cardinal Point

Historical data used to build models: Historic Demands

Day of week effect

Historic Weather

Additional effects – School Holidays, Time of Year

Basic Demand

Week Day Component

Weather Component

NG National Demand forecast models

Additional Effects

Weather variables used in modelling

30

Weather variables used in Forecast Models represent

TemperatureTE and TO

IlluminationEI and related ID

Wind SpeedWS

Cooling Power of Wind (Chill Factor)CP

Met Office Forecast Data

31

Data source is Met Office

Forecast data for ~ 106 locations

Forecast arrives 4 times a day. Each forecast is for the next 14 days ahead and at hourly resolution

Weather variables that come through: temperature, solar radiation, wind speed and wind direction

Met Office weather files timestamps

Files processed & ready to use by Goal run weather file used for

GMT BSTYYYYMMDD 02:30 03:30 04:30 04:00, 09:00, Nominal (D+1)YYYYMMDD 08:30 09:30 10:30 12:00, BPS (D+1), Pre-Nom (D+2)YYYYMMDD 14:30 15:30 16:30 19:30YYYYMMDD 20:30 21:30 22:30 23:00

Met Office Forecast Data

32

Renewable generation calculations: the data for the nearest weather station to the generator is used

Weather station name Station code

Station number

Station weighting (% of Nat. Avg.)

Heathrow LN 772 28Bristol Filton BR 628 18

Birmingham C/H BM 535 16Hawarden MN 321 14Glasgow GW 134 10

Leconfield LF 382 7Leeming LM 257 7

Demand calculation: Forecast data for 7 main stations are used, weighted by population, to give National Average

Met Office Actual Data

33

We receive actual weather data for 45 Met Office weather stations Rest of the forecast stations are “Virtual MetMast” stations with no actual recorded

data Data is received for every hour

arrives at 30 minutes past the hour

GB National Demand: Cardinal Points

BST 2016 CP chart

W/Beginning 00:30 01:00 01:30 02:00 02:30 03:00 04:00 04:30 07:30 08:00 09:00 09:30 10:30 11:00 13:00 13:30 16:30 17:00 18:00 18:30 19:00 19:30 20:00 20:30 21:00 21:30 22:00 22:30 23:00 24:00

28-Mar-16

1F

1S 1A

04 1B 2F 2A 2B 3B

3C

4A 4B

4C

04-Apr-16 4A 4B11-Apr-1618-Apr-16

4A 4B25-Apr-1602-May-1609-May-16

4A 4B

16-May-1623-May-1630-May-1606-Jun-1613-Jun-16

1A

20-Jun-1627-Jun-1604-Jul-1611-Jul-1618-Jul-1625-Jul-16

01-Aug-1608-Aug-1615-Aug-16 4A 4B22-Aug-1629-Aug-16 4A 4B05-Sep-16

1S 1A 3C W/E F

12-Sep-16 4A 4B19-Sep-1626-Sep-16 4A 4B03-Oct-1610-Oct-16 4B17-Oct-16 4B24-Oct-16

Each Cardinal Point has a minimum of two forecast models: At least one conventional model

At least one trend model

Difference between conventional and trend models:


CP Trend model trends on:

1F previous day 1F

1A previous day 1A

1B previous day 1B

2F in day 1B

2A in day 1B

2B in day 1B & in day 2A


3C / DP in day 1B & in day 2A


4C in day 1B & in day 2A

Basic Demand

Week Day Component

Weather Component

Additional Effects

Previous relevant demand that a trend model trends on

Trend models

Basic Demand

Week Day Component

Weather Component

Additional Effects

Conventional models

Example of the 3B statistical model equations for the conventional and trend models:

37


3B conventional model

53501.31 + 633.29 * Y12 + 24.05 * R + -0.17 * R*R + -1211.94 * Fri + -6189.44 * Sat + -5772.22 * Sun + -702.67 * TE15_0 + -17.05 * (EI12_0 + EI15_0) + 48.32 * WS15_0

3B trend model trending on 1B in day demand

32803.04 + 0.71 * L1B_0 + -1108.77 * Tue + -1319.84 * (Wed + Thu) + -2326.37 * Fri + -6238.73 * Sat + -5023.05 * Sun + -274.39 * T015_0 + -23.78 * (EI12_0 + EI15_0) + 47.26 * WS15_0

3B trend model trending on 2A in day demand16004.87 + 0.74 * L2A_0 + -1079.74 * Fri + -2176.56 * Sat + -0.02 * R*R + -196.6 * T015_0 + -14.03 * (EI12_0 + EI15_0) + 57.34 * WS15_0


Wind Generation

Wind Power Forecasting- Feb 17- Metered Wind Capacity ~ 10,100 MW- Unmetered ~ 4,800 MW- Mar 16- Metered Wind Capacity ~ 9,200 MW- Unmetered ~ 4,200 MW

National Demand ForecastMetered Wind Power Forecast

Unmetered Wind FarmsMetered Wind Farms

Wind Generation

40

Load Curve Optimisation Using Actual Wind Speed Data

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 5 10 15 20 25

Current Model Optimised Based on Actual WS

41-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 2 4 6 8 10 12 14 16 18 20

Current Model Optimised Based on Forecast WS

Load Curve Optimisation Using Forecast Wind Speed Data to Remove Bias

42

Alternative Load Curves

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 2 4 6 8 10 12 14 16 18 20

Current Model Optimised Based on Actual WS Optimised Based on Forecast WS

Metered Wind Generation

Unmetered Wind Generation (estimate)

0

1000

2000

3000

4000

5000

6000

Mar

11

May

11

Jul 1

1

Sep

11

Nov

11

Jan

12

Mar

12

May

12

Jul 1

2

Sep

12

Nov

12

Jan

13

Mar

13

May

13

Jul 1

3

Sep

13

Nov

13

Jan

14

Mar

14

May

14

Jul 1

4

Sep

14

Nov

14

Jan

15

Mar

15

May

15

Jul 1

5

Sep

15

Nov

15

Jan

16

Mar

16

May

16

Jul 1

6

Sep

16

Unmetered Wind Capacity in GB

Impact of PV

Unmetered solar generation (estimate)

0

2000

4000

6000

8000

10000

12000

Mar

11

May

11

Jul 1

1

Sep

11

Nov

11

Jan

12

Mar

12

May

12

Jul 1

2

Sep

12

Nov

12

Jan

13

Mar

13

May

13

Jul 1

3

Sep

13

Nov

13

Jan

14

Mar

14

May

14

Jul 1

4

Sep

14

Nov

14

Jan

15

Mar

15

May

15

Jul 1

5

Sep

15

Nov

15

Jan

16

Mar

16

May

16

Jul 1

6

Sep

16

Unmetered PV Capacity in GB

Location of individual PV sites

Increasing Volatility

Two consecutive Saturdays in June

Saturday 13 Jun 2015

0030

0200

0330

0500

0630

0800

0930

1100

1230

1400

1530

1700

1830

2000

2130

2300


18,000

20,000

22,000

24,000

26,000

28,000

30,000

32,000

34,000

36,000

38,000

0030

0200

0330

0500

0630

0800

0930

1100

1230

1400

1530

1700

1830

2000

2130

2300

Saturdays 06 and 13 Jun 2015

18,000

20,000

22,000

24,000

26,000

28,000

30,000

32,000

34,000

36,000

38,000

0030

0200

0330

0500

0630

0800

0930

1100

1230

1400

1530

1700

1830

2000

2130

2300

Difference between sunny and cloudy SaturdaysSaturday 6th June

Sunny

Windy

Saturday 13th June

Overcast

Still

Difference between sunny and cloudy SaturdaysSaturday 06 Jun 2015

18,000

20,000

22,000

24,000

26,000

28,000

30,000

32,000

34,000

36,000

38,000

0030

0200

0330

0500

0630

0800

0930

1100

1230

1400

1530

1700

1830

2000

2130

2300


0030

0200

0330

0500

0630

0800

0930

1100

1230

1400

1530

1700

1830

2000

2130

2300


18,000

20,000

22,000

24,000

26,000

28,000

30,000

32,000

34,000

36,000

38,000

0030

0200

0330

0500

0630

0800

0930

1100

1230

1400

1530

1700

1830

2000

2130

2300

Embedded Conventional Generation


0030

0200

0330

0500

0630

0800

0930

1100

1230

1400

1530

1700

1830

2000

2130

2300

Embedded Conventional Generation


18,000

20,000

22,000

24,000

26,000

28,000

30,000

32,000

34,000

36,000

38,000

0030

0200

0330

0500

0630

0800

0930

1100

1230

1400

1530

1700

1830

2000

2130

2300

Embedded Wind Generation Embedded Conventional Generation


0030

0200

0330

0500

0630

0800

0930

1100

1230

1400

1530

1700

1830

2000

2130

2300

Embedded Wind Generation Embedded Conventional Generation


18,000

20,000

22,000

24,000

26,000

28,000

30,000

32,000

34,000

36,000

38,000

0030

0200

0330

0500

0630

0800

0930

1100

1230

1400

1530

1700

1830

2000

2130

2300

Solar (PV) Generation Embedded Wind GenerationEmbedded Conventional Generation


0030

0200

0330

0500

0630

0800

0930

1100

1230

1400

1530

1700

1830

2000

2130

2300

Solar (PV) Generation Embedded Wind GenerationEmbedded Conventional Generation


18,000

20,000

22,000

24,000

26,000

28,000

30,000

32,000

34,000

36,000

38,000

0030

0200

0330

0500

0630

0800

0930

1100

1230

1400

1530

1700

1830

2000

2130

2300

Demand on Transmission System Solar (PV) GenerationEmbedded Wind Generation Embedded Conventional Generation


0030

0200

0330

0500

0630

0800

0930

1100

1230

1400

1530

1700

1830

2000

2130

2300

Demand on Transmission System Solar (PV) GenerationEmbedded Wind Generation Embedded Conventional Generation

Increasing Reliance on Weather Forecasts

Solar Radiation is hard to forecast

Physics of cloud formation highly complex

Some metrological conditions particularly challenging

Can see large errors – particularly when clouds form or clear unexpectedly

Impact of PV Growth on Forecast Error

0

2000

4000

6000

8000

10000

12000

400

450

500

550

600

650

700

750

800Ap

r-200

6Ju

l-200

6Oc

t-200

6Ja

n-20

07Ap

r-200

7Ju

l-200

7Oc

t-200

7Ja

n-20

08Ap

r-200

8Ju

l-200

8Oc

t-200

8Ja

n-20

09Ap

r-200

9Ju

l-200

9Oc

t-200

9Ja

n-20

10Ap

r-201

0Ju

l-201

0Oc

t-201

0Ja

n-20

11Ap

r-201

1Ju

l-201

1Oc

t-201

1Ja

n-20

12Ap

r-201

2Ju

l-201

2Oc

t-201

2Ja

n-20

13Ap

r-201

3Ju

l-201

3Oc

t-201

3Ja

n-20

14Ap

r-201

4Ju

l-201

4Oc

t-201

4Ja

n-20

15Ap

r-201

5Ju

l-201

5Oc

t-201

5Ja

n-20

16Ap

r-201

6Ju

l-201

6

Estim

ated

PV

Capa

city

(MW

)

Mea

n Ab

solu

te E

rror

(MW

)

PV Capacity Monthly Average Day Ahead MAE - Afternoon Peak

Impact of Solar Radiation Forecast

0

100

200

300

400

500

600

0700 0800 0900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100

Mea

n Ab

solu

te E

rror

(MW

)

Time of Day

Mean Absolute Error of Demand Forecast Caused by Error in Solar Radiation Forecast

0

100

200

300

400

500

600

0700 0800 0900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100

Mea

n Ab

solu

te E

rror

(MW

)

Time of Day

Mean Absolute Error of Demand Forecast Caused by Error in PV Generation Model

59

Effect of Solar Eclipse on Demand - 11 August 1999

33000

33500

34000

34500

35000

35500

36000

3650009

:00

09:1

0

09:2

0

09:3

0

09:4

0

09:5

0

10:0

0

10:1

0

10:2

0

10:3

0

10:4

0

10:5

0

11:0

0

11:1

0

11:2

0

11:3

0

11:4

0

11:5

0

12:0

0

12:1

0

12:2

0

12:3

0

12:4

0

12:5

0

13:0

0

Time

Dem

and

MW

Tue 10/08/1999Wed 11/08/1999

Start of the eclipse 2200MW dropin demand

Min @11:15

Totality

3000MWincrease indemand

Max @11:44

End of theeclipse

60

35000

36000

37000

38000

39000

40000

41000

42000

43000

44000

45000

07:0

0

07:1

0

07:2

0

07:3

0

07:4

0

07:5

0

08:0

0

08:1

0

08:2

0

08:3

0

08:4

0

08:5

0

09:0

0

09:1

0

09:2

0

09:3

0

09:4

0

09:5

0

10:0

0

10:1

0

10:2

0

10:3

0

10:4

0

10:5

0

11:0

0

Effect of Eclipse

Normal Day

Lost PV

People slow to turn lights offActual Demand

Demand Suppression as people go out to watch

Additional Lighting Load

Lost Embedded Wind

61

Effect of Embedded PV Demand at Indian QueensWednesday 8th April 2015

-50

0

50

100

150

200

250

300

30 130

230

330

430

530

630

730

830

930

1030

1130

1230

1330

1430

1530

1630

1730

1830

1930

2030

2130

2230

2330

62

How Good Are We – PV Forecast 21-28 Feb 17

0

1000

2000

3000

4000

5000

6000

0100

0700

1300

1900

0100

0700

1300

1900

0100

0700

1300

1900

0100

0700

1300

1900

0100

0700

1300

1900

0100

0700

1300

1900

0100

0700

1300

1900

PV Forecast Model with Actual Solar Radiation Forecast PV Generation Best View of Actual PV Generation

In Day 0830 PV

63

National Demand Forecast 21-28 Feb 17

20000

25000

30000

35000

40000

45000

50000

55000

0030

0630

1230

1830

0030

0630

1230

1830

0030

0630

1230

1830

0030

0630

1230

1830

0030

0630

1230

1830

0030

0630

1230

1830

0030

0630

1230

1830

In Day Forecast

Forecast Demand

Wed

Thu

Fri Sat

Sun

Mon

Tue

Current Energy Forecasting projects

Sheffield Solar in partnership with the University of Sheffield Better view of the live out-put of PV generation in GB

Met Office Improving the methodology of forecasting solar radiation

University of Reading Improve the conversion of solar RA into generated MW Provide historical time series of renewables load factor Investigate use of satellite and radar data to modify the solar RA forecast

Modelling Embedded Non-Weather Variable Generation Improved modelling of PV at substation level Improved Embedded Wind modelling Improved PV Power Curves Live display of estimated PV output in Control Room Short term PV forecast based on current observations New methods to estimate current PV capacity

Demand Forecasting - University of Readingblogs.reading.ac.uk/tsbe/files/2017/04/1.1_J_Caplin... ·...

Documents

Transcript of Demand Forecasting - University of Readingblogs.reading.ac.uk/tsbe/files/2017/04/1.1_J_Caplin... ·...