31
Stochastic optimization of distributed energy resources in smart grids Joao Soares, Zita Vale GECAD / IPP – Polytechnic of Porto [email protected] 1 Paper No: 15PESGM2908

Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Embed Size (px)

Citation preview

Page 1: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Stochastic optimization of distributed

energy resources in smart grids

Joao Soares, Zita Vale GECAD / IPP – Polytechnic of Porto

[email protected]

1

Paper No: 15PESGM2908

Page 2: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Outline

• Introduction • Energy Resource Management • Uncertainty in ERM • ERM considering uncertainty • Case study

– Smart distribution network– scenario for year 2040

2

Page 3: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Introduction 3

Wind Power

Solar Power

Energy Storage

Electric Vehicles

Utility Grid

Demand Response

Loads

The network can be islanded from the Utility Grid

Market

VPP

Page 4: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

4

Energy Resource Management

Resource Management

(day-ahead, hour-ahead, real-time)

Resource Forecasting Island Mode Capability

Demand Response Self-healing Capability

Energy Resource Management • Market transactions • Contracts with suppliers • Generation • Demand Response • Storage • Electric vehicle • Power Flow

Real Time data acquisition Loads monitoring Generation monitoring Storage levels

Self-healing Capability

Direct Load Control

Resource Forecasting

Market Day-ahead scheduling

Hour-ahead scheduling

Real-time scheduling

24h

1h

……

Island Mode Capability

Demand Response

Page 5: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Uncertainty in ERM • Why considering uncertainty in ERM?

Forecast errors

– Wind (wind variation, weather conditions) – Solar PV (irradiation, clouds, etc.) – Load demand (user’s behavior, buildings, etc.) – EVs (user’s behavior, charging patterns, location)

5

Page 6: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Wind and Solar Scenarios • The forecasting error varies with the geographical location,

weather conditions and time horizon; • The wind scenarios are independent with solar scenarios:

– Assuming the errors follow a normal distribution – Wind and Solar: Monte Carlo simulations can generate a

set scenarios; – Scenario reduction techniques (e.g. clustering)

6

Pinson, Pierre, et al. "From probabilistic forecasts to statistical scenarios of short-term wind power production." Wind energy 12.1 (2009): 51-62. Su, Wencong, Jianhui Wang, and Jaehyung Roh. "Stochastic energy scheduling in microgrids with intermittent renewable energy resources." Smart Grid, IEEE Transactions on 5.4 (2014): 1876-1883. Botterud, Audun, et al. "Use of wind power forecasting in operational decisions." Argonne National Laboratory (2011).

Page 7: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Wind and Solar Scenarios 7

Botterud, Audun, et al. "Use of wind power forecasting in operational decisions." Argonne National Laboratory (2011).

Page 8: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Electric Vehicles uncertainty

• Why is uncertainty present?

8

– Location in the grid, e.g. network bus, network zone

– Energy demand

– Users’ behavior but diferent from regular load

– When, where and how much?

Page 9: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Electric Vehicles uncertainty

• Challenge: Reliable user’s behavior data

9

• Possible approach: cloud-based information

• Near future?

Page 10: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Cloud-based Information 10

Page 11: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Cloud-based Information 11

Data Parameters Description of data structure

Type of data Data class Size per record Vehicle’s ID Identification of EV int32 4 bytes

Location GPS location (latitude

and longitude and timestamp)

Double 3x8 bytes

Battery status Battery capacity; SOC level int32 2x4 bytes

Connected to outlet Connected (0/1); Outlet ID if available Binary and int32 1 + 4 bytes

Charged energy Charged energy; Start and end timestamp int32 and Double 4 + 2x8 bytes

Charge rate Charging power int32 4 bytes

Trip consumptions Energy consumption in the trip; Start and end

timestamp int32 and Double 4 + 2x8 bytes

• Table shows a very simple data scheme for the information that could be transmitted between the EVs and the cloud

• How much data will be stored?

Page 12: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Cloud-based Information 12

• Data creation (estimation per year)

0

0

1

5

25

125

625

3.125

10k 100k 500k 10m

Tera

byte

s (T

B)

Number of EVs

Cloud data impact (annually)

5 min 1 min 15 s

1600 TB

80 TB

• Data creation (estimation per year)

Page 13: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Cloud-based Information 13

• How can this concept be useful?

• …for stochastic scenarios generation

• So to motivate further research…

Page 14: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Stochastic Scenario Generation 14

• The day-ahead scenarios can provide information about the EVs’ locations for the next day, considering the information contained in the cloud (historic data);

• The hour-ahead scenarios takes into account the actual EVs’ location, and uses the historical data with similar last-hours pattern to generate a set of possible scenarios for the next hour.

Page 15: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Monte Carlo Simulation • Values must be filtered (e.g. vehicles’ locations) and according

to rules: – Weekends/weekdays – Time horizon (day-ahead, hour-ahead) – Holidays – Special Events (football day, music festival, etc.) – Traffic conditions

• To reduce computational burden, the historical data may be shrinked: – For example: 4 periods, each period representing 6 hours; – The reduced data is an important aspect of the methodology efficient use; – A matrix is built with the historical connections to the grid (for instance

outlet/network bus) of a given EV.

15

Page 16: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Tree combination 16

1 - Scenarios can be generated using forecast errors 2- Combination (load, wind, EVs) to obtain the final probability 3- Scenario reduction (just a small set of representative scenarios)

Pinson, Pierre, et al. "From probabilistic forecasts to statistical scenarios of short-term wind power production." Wind energy 12.1 (2009): 51-62. Su, Wencong, Jianhui Wang, and Jaehyung Roh. "Stochastic energy scheduling in microgrids with intermittent renewable energy resources." Smart Grid, IEEE Transactions on 5.4 (2014): 1876-1883.

Page 17: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

( ) ( ) ( )

( ) ( ) ( ) ( )

( ) ( ) ( )

, , , , ,

, , , , , , ,

, , , , ,

ForecastLoad L t Load L t Load L t

ForecastDG DG t GCP DG t DG DG t DG DG t

ForecastTrip EV t Trip EV t Trip EV t

P P P

P P P P

E E E

ωω ω

ωω ω ω

ωω ω

= + ∆

+ = + ∆

= + ∆

Uncertainty modeling in ERM 17

Consumption Forecast errors: Normal distribution function

Only uncertainty in travel forecast considered in this

formulation Distributed generation

EVs trip

Generation curtailed power

Forecast Scenario

generation index ω representing each considered scenario

Page 18: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

( ) ( ) ( ) ( )

( ) ( ) ( ) ( )( )

( ) ( ) ( ) ( )

, , , , ,1 1 1 1

, , , , , ,1

, , , , , ,1 1

min

+SP DG

EV

DGL

N N NT

SP SP t SP SP t DG DG t DG DG tt SP DG

N

Dch EV t Dch EV t Ch EV t Ch EV tEV

NN

NSD L t NSD L t GCP DG t GCP DG tL DG

f

c P c P

c P c P

c P c P

ω

ω ωω

ω ω

ω ω

π= = = =

=

= =

=

× × ×

+ × − ×

+ × + ×

∑ ∑ ∑ ∑

∑ ∑

Energy Resources Management

• Objective function

18

Generation curtailed power

Vehicles charge

Distributed generation External suppliers

Vehicles discharge

Non-supplied demand

Minimize the operation cost

Probability of scenario ω

scenario

Uncertainty

Page 19: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Energy Resources Management • Constraints:

– Power balance – Voltage limits – Line thermal limits – Resource limits

• DG units • External suppliers • EVs • …

19

Page 20: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Energy Resources Management • Constraints for each scenario :

– Network power active power balance at bus i in period t

– Network power reactive power balance at bus i in period t

20

( ) ( ) ( ) ( ) ( )( ) ( ) ( )

( ) ( ) ( ) ( )( ) ( )

( ) ( ) ( )( ) ( )

2, , , , , , ,

, , , , , , , ,1 1 1

, , , , , , ,1 1

cos sini

i i iSP DG EV

iiEVL

ii ij iji t i t j t ij t ij t Gi t Di tj L

N N Ni i i i

Gi t SP SP t DG DG t GCP DG t Dch EV tSP DG EV

NNi i i

Di t Load L t NSD L t Ch EV tL EV

G V V V G B P P

P P P P P

P P P P

ω ω ω ω ω ω ω

ω ω ω ω

ω ω ω ω

θ θ∈

= = =

= =

+ + = −

= + − +

= − +

∑ ∑ ∑

∑ ∑

( ) ( ) ( ) ( )( ) ( ) ( ) ( )

( ) ( ) ( )

( ) ( ) ( )( )( ) ( ) ( ) { } { } { }

2, , , , , , ,

, , , ,1 1

, , , , ,1

, , ,

sin cos

; 1, , ; 1, , ; 1, ,

i

i iSP DG

iL

ij ij iii t j t ij t ij t i t Gi t Di tj L

N Ni i

Gi t SP SP t DG DG tSP DG

Ni i

Di t Load L t NSD L tL

Bij t i t j t

V V G B B V Q Q

Q Q Q

Q Q Q

N t T i N

ω ω ω ω ω ω ω

ω ω

ω ω ω

ωω ω ω

θ θ

θ θ θ ω

= =

=

− − = −

= +

= −

= − ∀ ∈ ∀ ∈ ∀ ∈

∑ ∑

𝜔𝜔

Page 21: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Energy Resources Management • Constraints:

– Voltage magnitude and angle at bus i in period t

– Line thermal limits at line k in period t

21

( ) { } { } { }, 1, , ; 1, , ; 1, ,i iMin Max Bi tV V V N t T i Nωω ω≤ ≤ ∀ ∈ ∀ ∈ ∀ ∈

( ) { } { }, 1,..., ; 1,...,i iMin Max Bi t t T i Nωθ θ θ≤ ≤ ∀ ∈ ∀ ∈

( ) ( ) ( )( ) ( )

( ) ( ) ( )( ) ( )

{ } { } { }

_, , , ,

_, , , ,

1, , ; 1, , ; 1, ,

Maxij sh i Lki t i t j t i t

Maxij sh i Lkj t j t i t j t

B

U y U U y U S

U y U U y U S

N t T i N

ω ω ω ω

ω ω ω ω

ωω

× × − + × ≤

× × − + × ≤ ∀ ∈ ∀ ∈ ∀ ∈

Page 22: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Energy Resources Management • Constraints:

– Active and reactive generation limit for the DG unit in period t

– Active and reactive generation limit for the external supplier SP in period t

– Reactive demand power for the load L in period t

22

( ) ( ) ( ) ( ) ( ) { } { }, , , , , , , , , , 1,..., ; 1,..., DGMin DG t DG DG t DG DG t Max DG t DG DG tP X P P X t T DG Nω ω ω ω ω× ≤ ≤ × ∀ ∈ ∀ ∈

( ) ( ) ( ) ( ) ( ) { } { }, , , , , , , , , , 1,..., ; 1,..., DGMin DG t DG DG t DG DG t Max DG t DG DG tQ X Q Q X t T DG Nω ω ω ω ω× ≤ ≤ × ∀ ∈ ∀ ∈

{ } { }( , ) ( , ) ; 1,..., ; 1,...,SP SP t SPMax SP t SPP P t T SP N≤ ∀ ∈ ∀ ∈

{ } { }( , ) ( , ) ; 1,..., ; 1,...,SP SP t SPMax SP t SPQ Q t T SP N≤ ∀ ∈ ∀ ∈

( ) { } { }( , ) ( , ) ( , ) ( , ) ( , ) tan 1,..., ; 1,...,Load L t Load L t Red L t Cut L t NSD L t LQ P P P P t T L Nϕ= − − − × ∀ ∈ ∀ ∈

Page 23: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Energy Resources Management • Constraints:

– Energy stored at the end of period t for each EV

– Minimum and maximum stored energy in the electric vehicle V in period t

23

( ) ( ) ( ) ( ) ( )( )

( )

( ) ( )

{ } { } { }

, , , 1, , , , , , ,

, 1,

1

1 ; 1

1, , ; 1, , ; 1, ,

Stored EV t Stored EV t Trip EV t c EV Ch EV t Dch EV td EV

Stored EV t Initial EV

EV

E E E t P P

t E E t

N t T EV N

ω ω ω ω ω

ω

ω

ηη

ω

= − + ∆ × × − ×

= → = ∆ =

∀ ∈ ∀ ∈ ∀ ∈

( ) ( ) ( )

{ } { } { }, , , ,

1, , ; 1, , ; 1, ,BatMin EV t Stored EV t BatMax EV t

EV

E E E

N t T EV Nω

ωω

≤ ≤

∀ ∈ ∀ ∈ ∀ ∈

Page 24: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Energy Resources Management • Constraints:

– Charge and discharge processes are not simultaneous in the

electric vehicle V in period t

– Charge and discharge limits of the electric vehicle V in period t

24

( ) ( )

{ } { } { }, , , , 1

1, , ; 1, , ; 1, ,Ch EV t Dch EV t

EV

X X

N t T EV Nω ω

ωω

+ ≤

∀ ∈ ∀ ∈ ∀ ∈

{ } { }( , , ) ( , ) ( , , ) 1,..., ; 1,...,Ch V t ChLimit V t Ch V t VP P X t T V Nω ω≤ × ∀ ∈ ∀ ∈

{ } { }( , , ) ( , ) ( , , ) 1,..., ; 1,...,Dch V t DchLimit V t Dch V t VP P X t T V Nω ω≤ × ∀ ∈ ∀ ∈

Page 25: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Case study

• Goal: operation cost minimization • 33-bus distribution network – scenario for year 2040 • High DER penetration

– 66 DG units – 218 consumers – 1000 electric vehicles

• 10 external suppliers

25

Page 26: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Case study - Forecast 26

Production and consumption

Energy trips of EVs

Several scenarios were based on the deviations on these forecast resources:

• 5% consumers demand • 15% renewable DGs

• 10% EVs trip

Page 27: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Case study - Results

• Energy resource scheduling results

• Scheduled power for the external suppliers

27

Deterministic approach Stochastic approach Cost (m.u.) 6979.31 7069.82 +1.3%

External suppliers represent around 49% of the necessary energy

consumption in the grid

Page 28: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Case study - Results • Energy resource scheduling results – stochastic approach

28

Production

Consumption

Page 29: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Computational burden

29

Deterministic (1 scenario) Stochastic approach 33-bus 1000 EVs 6 hours 1 week

180-bus 6000 EVs 2 weeks ???

• The execution time refers to full AC model using GAMS (MINLP)

• The DC model is much faster because it is linear but not considers losses

• Centralized scheme is hard to solve

• EVs treated individually are the problem in ERM (it may be aggregated)

• Decentralized scheme is an alternative

• Metaheuristics can reduce the 1 scenario problem to seconds to minutes

• The need for metaheuristics in stochastic optimization is important

Page 30: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

Stochastic optimization of distributed

energy resources in smart grids

Joao Soares, Zita Vale GECAD / IPP – Polytechnic of Porto

[email protected]

30

Paper No: 15PESGM2908

Page 31: Stochastic optimization of distributed energy resources …sites.ieee.org/psace-mho/files/2015/07/01_15PESGM2908.pdf · Stochastic optimization of distributed energy resources in

• Following the same cloud-based approach

• Cloud data can be used to obtain users’ charging/driving behavior

• It can be assumed to follow a normal distribution with mean and standard deviation for every period.

31

Energy demand