an outage work planning under the weather uncertainties · 2016-12-13 · transient stability...

An outage work planningunder the weather uncertainties

(不確定条件下での作業系統計画について)

N. Yorino, *Y. Zoka (Electric Power & Energy System Lab.)(餘利野直⼈，*造賀芳⽂，電⼒・エネルギー⼯学研究室)

Hiroshima University(広島⼤学)

ECC Workshop 2016

Table of Contents

¤ Overview of the current research (Hiroshima Univ.)¤ Our research works

¤ Outage works¤ What is the outage work?

¤ Problems to be appeared under uncertainties¤ Difficulties against weather conditions

¤ A solution for supporting system (IEEJ paper)¤ Formulation¤ Simulation studies

¤ Summary

Oct. 19, 2016 2

Overview of the current research(Hiroshima Univ. Team)

Oct. 19, 2016 3

Outline of Research Themes (HU)

Fields Groups Themes/Keywords

Analysis, Evaluation

Reliability analysis, evaluation (Y, S) Robust stability & security, Feasible region, Uncertainty, PV + Battery

Transient stability analysis (Y, S) Critical Trajectory, Computation efficiency, unbalanced fault

Reactive power evaluation (Y, S) Deregulation, “Q” pricing, OPF

EV/PHEV (Z) Frequency control, Uncertainty, Economic value evaluation

Frequency control analysis (Y) Performance indices for LFC

Operation, Control

Load forecasting (Z)Renewable-energy forecasting (Y, S)

Min/Max Load forecast, continuous-updating regressionPV/WT output prediction, Uncertainty

Dynamic ELD / Stochastic OPF (Y, S) Dynamic Economic Load Dispatch including PV/WT, Uncertainty

System stabilizing control (Z) PSS, Robust control (Uncertainty)，Support Vector Machine

Microgrid operation, control (Y, Z, S) Demand-Supply Control, Plug-and-Play, Stabilization, Uncertainty

Autonomous distributed control (Y, Z) Voltage control for distribution systems, PV, Multi-agent system, Uncertainty

Power electronics application (Y, Z, S)Inverter Design (Y, S)FACTS Controller Design (Y)

Quasi-synchronizing power invertors, FACTS design

Planning, Optimization

OPF (O, Y)Optimal FACTS allocation (Y)

OPF for distribution system, Security-constrained OPF (SCOPF)Cost minimization, Load shedding, Voltage stability and Security

Outage-work planning (Z, O) PV, Support system, UncertaintyUnit Commitment with RE (Y, S)Operation Planning Including batteries, PV, forecast error, Uncertainty

(Y): Yorino, (Z): Zoka, (S): Sasaki, (O): OutsideOct. 19, 2016 4

Treatment of Uncertainty

future

Variance of EstimationError

Stochastic Power Flow

Robust Power System Security

Present t0

Power System Security Analysis, Planning, Operation and Control

s

Co-variance Matrix Confidence Intervals (CI)

Increasing variance w.r.t. time

Future t1

CI2 CI4

2s4s

t1 t2 t3 t4

Oct. 19, 2016 5

Subject (1)

1. Robust Power System Security (RPSS)

> Robust stability under N-1 contingencies inside CI

Security Analysis with Uncertainty

Voltage Stability

Over Loading

Voltage Limitation

Transient Stability

Freq. Dev.

System must be stable for all single

contingencies.

Conventional N-1 Security Criterion

Robust N-1 Security (More strict concept)where uncertain parameters are allowed to vary in

Confidence Intervals

Definition of RPSS

Oct. 19, 2016 6

Robust Static Security Region (RSS)

7

ü Power Flow Equations 𝐹(#) 𝑢, 𝑝 =0ü Constraints 𝐺(#) 𝑢, 𝑝 ≤ 0, 𝑛 = 0,1,⋯ , 𝑁

ü Control Variables 𝑢 ≤ 𝑢 ≤ 𝑢

Measure of RSS Region

ü Uncertainties in CI 𝑝 ≤ 𝑝 ≤ 𝑝

The Worst Case Max & Min inside RSS Region with Uncertainties

Upper bound 𝛼122 = min6,7

{max6

𝑐<𝑢}

Lower bound 𝛼122 = max6,7

{min6𝑐<𝑢}

Objective function

Constraints

Oct. 19, 2016

Definition of RRS

13

Robust Static Security region (RSS)

Robust Dynamically Reachable Security region (RRS)RRS : (RSS) + (Ramp Rate Constraints of u)

V

Time

t

t0

RSS

Reachable from u0

OperatingPoint

RSS

u0

Oct. 19, 2016

Robust Dynamic Analysis Example

9

Controllable parameterØ Generator outputs

Constraints for RRS : LinearØ Demand and supply balance Ø Generator output limits Ø Power flow equation (DC power

flow) Ø Security limits of line flowsØ Generator Ramp Rate ConstraintsØ Initial Operating Point at t=t0

ContingencyØ 1 of 2 Lines Trip at A

UncertaintiesØ CI for RE generations

Test System

40%

60%

G1 G2

G3Battery

① ②

③

④

⑤ ⑥

1F

2F

3F

4F 5F

6F7F

PVWT

PVWT

34z

36z56z

15z

12z 24z

25z

A

u=[G1, G2, G3, S]’

p=[RE1, RE2]’

Objective functionα = c@u : Total Generation

Oct. 19, 2016

Results of RRS Evaluations

Oct. 19, 2016

20002500300035004000450050005500

13:00 13:20 13:40 14:00

Total Thermal Generation

Operating Point

Time Time

20002500300035004000450050005500

13:00 13:20 13:40 14:00

Total Thermal Generation

Lower bound

Demand & Supply Mismatch

Upper bound of Gen

Case 2Case 1

RRS analysis at 13:00 for 1 hour ahead cT = [ 1 1 …1]cTu = Total Generation

Subject (2)

Transient Stability Analysis

¤ Critical Trajectory Method

Fast Stability Analysis

Oct. 19, 2016 11

Critical waveform by methods A & Bcompared with conventional simulation method

Rotor angle of generator 1 for fault at point G

-100

-50

0

50

100

150

200

250

300

350

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Time [s]

Roto

r A

ngle

[deg]

CCT by Method B = 0.2726 [s]

CT = 0.272 [s]

Conventional Simulation

CT = 0.273 [s]

CCT by Method A = 0.2732 [s]

Rot

or A

ngle

[deg

]

Time [s]

End Point of Method B Agrees with CUEP by Shadowing

End Point of Method A

Critical Trajectory by Method B

Critical Trajectory by Method A

Oct. 19, 2016 12

No. 17

Simulation Case for CT =0.272 [s] for fault at point G

-100

-50

0

50

100

150

200

250

300

350

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Time [s]

Roto

r A

ngl

e [

deg]

q0130 q1030

q1130

q0230

q2030

q2830

q2930

q1930

CT = 0.272 [s]

Time [s]

Rot

or A

ngle

[deg

]

CCT by Method B = 0.2726 [s]



CCT by Method A = 0.2732 [s]


UEP by ShadowingEnd Point of Method A

End Point of Method B


Oct. 19, 2016 13

No. 18

Simulation case for CT=0.273 [s] for fault at point G

-100

-50

0

50

100

150

200

250

300

350

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Time [s]

Rot

or A

ngle

[de

g]




UEP by Shadowing

CT = 0.273 [s]

End Point of Method A End Point of Method B

Time [s]

Rot

or A

ngle

[deg

]

CCT by Method B = 0.2726 [s]CCT by Method A = 0.2732 [s]

q0130

q1030

q1130 q0230

q2030

q2830 q2930

q1930


Oct. 19, 2016 14

Formulation of Critical Trajectory Method

Variables: CCT, Boundary conditions

>Initial point Condition for >End point Conditions for

Trapezoidal Conditions for numerical integrationNumber of points (m): specified. (Typically m=10)

0x

CP

xm

xk

ee

eEach point is connected by using Trapezoidal Method

x0 ~ xm+1: critical trajectoryx0

xm+1

x1

0 1, ,..., mx xe +

1mx +

Oct. 19, 2016 15

□ New End Condition, ■ Previous End Condition

0.45 0.28 0.12

4.242.39

3.92

0.930.00

1.09

5.68 4.29

21.26

-3.000

2.000

7.000

12.000

17.000

22.000

3machine 4machine 6machine 7machine 10machine 30machine

AverageerrorinCCT

computation[%

]

Powersystemmodel

%Errors in CCT

16Oct. 19, 2016

0.03 0.04 0.05 0.070.15

1.80

0.04 0.07 0.05 0.060.11

1.00

0

0.5

1

1.5

2

2.5

3machine 4machine 6machine 7machine 10machine 30machine

AverageCPU[s]

Powersystemmodel

CPU Time

17

□ New End Condition, ■ Previous End ConditionOct. 19, 2016

Subject (3)

Demand & Supply Management (Micro-EMS Controller)¤ PV Generation & Load Forecasts¤ Operation Planning (Unit Commitment) using BT¤ Computation of Dynamic Feasible Region ¤ Real Time Fast Economic Load Dispatch¤ Stochastic Power Flow¤ Frequency Control using BT, etc.

G

Thermal Gen.

GThermal Gen.

Residential

Industrial

WT

Battery (BT)

PV

Demand & Supply Manager

System Operation

under Uncertainty

Oct. 19, 2016 18

Micro-EMS Controller

�Power Demand, Weather Info.�Generator, Battery(BT), Electric

Vehicle(EV) Info., Network Info., etc.

�Real-time Demand, Weather Info.,�Gen./BT/EV Operating Info.,�Network Operating Info., etc.

Day-ahead Forecast�Power Demand�Photovoltaic Gen. (PV)�Wind-turbine Gen. (WT)�Confidence Interval(CI)

Gen. Scheduling�Unit Commitment (UC) �Battery(BT) Operation�Robust Security�Reserve Evaluation

Real-time Forecast�Power Demand�PV, WT�CI

Gen. dispatching�UC�Economic Load

Dispatching (ELD,TDF, PLF)

�Robust Security�BT re-scheduling

Normal Control�Load Frequency

Control (LFC)�AR EvaluationEmergency Control�Gen. Shedding�Load Control(Shedding,

BT, Electric Vehicle(EV))

Re-dispatching�Gen./BT re-dispatching�S&D Mismatch(SDM)

�GOV

-

PV, WT

Customer

Information Board

>

<=

SDM

Red: under VerificationBlue: under DevelopmentBlack: under Consideration

Day-Hour (UC) orderDay-ahead Planning

Manager

Hour-Min. (ELD) orderReal-time Operating

Manager

Min.–Sec.(LFC,Gov)orderReal-time Operating

Manager

Off-line Database On-line Database

�GOV

-

�GOV

-

Customer

Customer

BT, EV

Oct. 19, 2016 19

Stochastic Line Flow Model

Nde

Line Flow Limit

Line Flow [pu]

Pr(X>Xlimit)

Pr

Expected Line Flow

Node: Cov(P) à Line Flow: Cov(F) à Line Flow Limits Assumption: Normal distribution of RE Prediction Error

≦0.26％

[ ] [ ] [ ]TijCov Cov s= × Þ = × × =F S P F S P S

Linear Line Flow Model

Line Flow Control :Pr (Line Flow Violation) < a

Oct. 19, 2016

QP Problem to be solved every 5 minutes to update 1 hour Generation Schedule (GS).minimize:

subject to:

Formulation for Robust Dynamic ELD

21

( 1) ( )k Gk Gk kt tP Pd d-- £ - £

1 1( ) ( ) ,( )

n nN N

Gk Dkk k

t tP E P= =

=å å D-S Balance

( )kt Gk kttPa a£ £ FOR

Ramp rate

(1)

(2)

(3)

(4)

1( ) ( ) ( )

nN

l ll l lj Gj l ll lj

F D t S P t F D tbs bs=

- + + £ £ - +å (5)

1

1

2

1

( )

( )2

,

( ) ( )n

T t

t tN

kk k k k

k

f f t

af t P b P ct t

+

=

=

=

= + +æ öç ÷è ø

å

å

Network

Oct. 19, 2016

t = t1,!, t1 +T

24-hours scheduling resultsPG1 PG2 PG3 BTout PD Pt PPV PWT BTSOC

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0:00 5:00 10:00 15:00 20:00Time [hour]

Out

put p

ower

[MW

]

0

0.4

0.8

1.2

1.6

2

2.4

SOC

[MW

h]

G1 G2 G3 Bd Pd Pt Ppv Pwt BsPG1 PG2 PG3 BTout PD Pt PPV PWT BTSOC

(a) Weekday (b) Weekend

12:15 12:25 12:35 12:45 12:55 13:05 13:150

1000

2000

3000

Time

PG

1[k

W]

TDF and Output Power

Upper limit of TDFLower limit of TDFOutput Power

12:15 12:25 12:35 12:45 12:55 13:05 13:150

1000

2000

3000

Time

PG

2[k

W]

12:15 12:25 12:35 12:45 12:55 13:05 13:150

1000

2000

3000

Time

PG

3[k

W]

10:00 10:10 10:20 10:30 10:40 10:50 11:000

0.5

1

1.5

2

2.5

3

Time

TD

F P

1[M

W]

TDF

0% Upper limit of TDF10% Upper limit of TDF0% Lower limit of TDF10% Lower limit of TDF

The feasible region and dispatch value on weekday TDF at 10:00 on weekday.Oct. 19, 2016 22

High PV penetration case

0

400

800

1200

1600

2000

2400

0

1000

2000

3000

4000

5000

6000

0:00 2:00 4:00 6:00 8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00

SO

C[k

Wh]

Pow

er D

eman

d[kW

]

Time[hour]

G1 G2 G3 Bd Pd Pt Ppv Pwt Bs

The feasible region and dispatch valueOct. 19, 2016 23

Daily Operation SimulationTotal Demand Total Demand - PV,WT OutputsGen. Output

Oct. 19, 2016 24

An outage work planningunder the weather uncertainties

IEEJ Transactions on Power and Energy, (to be published)

Oct. 19, 2016 25

Outage Work?

¤ Stable power supply¤ Important mission of power systems¤ Inspection, repair, reinforcement, …¤ à Outage works are necessary

¤ Outage work planning is to determine…¤ System configuration, work combination, work

schedule, etc.

¤ In this study,¤ Regarded as the problem of system configuration

Oct. 19, 2016 26

At planning stage, a maximum load is assumed for work days.

However, if the weather is different from the assumed condition…

Outage work planning must be done one-year aheadbased on annual plan.

PV output

Load forecastOutage work plan

Power flow changes depending on PV outputs

PV impacts?

27

PV output

Load forecastOutage work plan

Oct. 19, 2016

IEEJ Technical Report (Analysis Tech.)

¤ IEEJ Technical Report¤ Hiroshima Univ.¤ Chugoku

Electric Power Co.

¤ Universities¤ Manufacturers¤ Institutes¤ Gas companies¤ Generation Co.¤ 10 Utilities

28Oct. 19, 2016

Typical actual workflow of utilities

¤ Utilities

¤ Outage works

Oct. 19, 2016 29

Typical actual workflow of utilities

¤ Utilities

¤ Outage works

Oct. 19, 2016 30

Power BalancePower Grid related

Dec

ades

Seve

ral y

ears

Year

ly

Mon

tyly

Wee

kly

Day

lyRe

al ti

me

Generators Networks

Flowchart (general)

Oct. 19, 2016 31

Normal system

LoadG SS

SS

SS

SS

Power supply = Stable

: SubstationSS

Oct. 19, 2016 32

Normal system

LoadG SS

SS

SS

SS

Target line

No route!

Supply failure to Loads

Outage work = infeasible!

Oct. 19, 2016 33

LoadG SS

SS

SS

SS

Target line

Outage work system

Power supply = stable

Outage work = feasible

Oct. 19, 2016 34

Flow of the planning

35

Make ranking of the set

Select the outage work system

Calculate the index

Build a feasible candidate set

Check the constraints

Enumerate candidates

Define of the search space

*) Hamming Distance constraint: the number of switching from the normal system.**) N-2 supply failure power: the failed amount of power supply when simultaneous 2-lines outage occur.

1st=sys7, 2nd=sys2, 3rd=sys10

Outage work system = sys7

Index** = N-2 supply failure power

{sys2, sys7, sys10}

Power balance, voltage, line power flow, …

{sys1, sys2, …, sysN} | N: total candidates

Search space based on Hamming Distance*

Oct. 19, 2016

Background

Oct. 19, 2016 36

*) Outage work: temporal, partial stop for inspection, repair, etc. (not blackout)

Outage work planning* = determine work schedules, orders, combinations, system configurations

Outage work system candidates

All system configurations

Usual systems

Outage work system candidates

Highly reliable outage work systems

Determinethe work-date

Rearrange of work-date

Built up a chart日 1 2 3 4 5 6

作業設備月火水木金土作業1作業2作業3作業4作業5作業6作業7

LoadG SS

SS

SS

SS

Research Target

In case of huge, rapid change of PV outputs

Due to the over loads, feasible outage work systems are different depending on PV output.

New outage work planning taking into account PV output uncertainties is necessary.

An impact to outage work by PVs

37

N1

N3

N4N2

N6N5

L1L2L3

L4

L7

L5

L6

PV

PV

PV

Over load!

Outage work system

Target line

Oct. 19, 2016

Intersection of the candidates

Oct. 19, 2016 38

PV

PV

PV

PV

PV

PV

��

PV

PV

PV

{sys1, sys2, sys3}

{sys2, sys3, sys4}

{sys2, sys4}

PV installed n nodesà 2n patterns

If prepared as sys2, the plan will be feasible even if any patterns of PV output occur.

A typical example

¤ Data¤ Total load: 3,000MW¤ 7 generators

¤ 109 nodes, 138 branches¤ Upper system: loop-based¤ Lower system: radial-based

39

System model

Oct. 19, 2016

Assumptions (example)

¤ PV assumption¤ Divided into 3 areas (A, B, C)

¤ Same PV output states within the same area

¤ PV Install conditions¤ PV installation types

¤ Mega-solar : Roof-top PV= 1 : 1

¤ Extrapolate for larger cases

¤ Amount of installation¤ Mega-solar = actual data¤ Roof-top = based on

household statistics

¤ Outage works¤ Target = L57 (1 cct)

40Oct. 19, 2016

Lost of feasible plan due to PVs

Oct. 19, 2016 41

Finally, no outage work plan obtained because of no feasible solution.

The more PV penetration, the fewer feasible outage work system candidates due to overload line.

N1

N3

N4N2

N6N5

L1L2L3

L4

L7

L5

L6

PV

PV

PV

Over load!

Outage work system

Target line

0

20

40

60

80

100

120

140

0 2 4 6 8 10

実行可能な作業系統候補数[個]

PV導入量 [%]PV penetration rate (%)

Feas

ible

sys

tem

can

dida

tes

Lost of feasible plan due to PVs

42

Feasible system candidates depending on PV output patterns and penetration rates

0

20

40

60

80

100

120

140

0 2 4 6 8 10

� ��[�]

PV�� [%]PV penetration rate (%)

Feas

ible

sys

tem

can

did

ates

*) PV output conditionsØ 1 = 100%Ø 0 = 0%

PV output conditions* PV penetration rate assumptions

Area A Area B Area C

Common feasible systems

Rapid reduction in the cases 2, 5, 6, and 8.

In area C, large number of roof-top PVs in residential area and many mega-solar systems in industry area.

Large power flow change occursdepending on PV output patterns.

Oct. 19, 2016

Problem to be solved

¤ Problem¤ No feasible outage work system obtained by PVs

¤ Mainly due to overload problem¤ à An additional function to avoid overload

conditions.

Oct. 19, 2016 43

¤ All weather feasible

Oct. 19, 2016 44

PV

PV

PV

PV

PV

PV

・・・PV

PV

PV

sys1 sys2 sys4

sys3 sys2 sys4

sys2 sys4

PV installed n nodesà 2n patterns*

*) PV output is assumed 0% or 100% because the most severe cases must be considered.

If prepared as sys2, the plan will be feasible even if any patterns of PV output occur.

Formulation (minimization of N-2)

Oct. 19, 2016 45

ååå¹= = =

•=MC

jii

MC

j

MB

bkbbijk pLyxpyxA

1 1 1)(),(),,( w

),,(min 1pyxA

Minimize the index: N-2 supply failure power(decision variables: x = facilities connection)Subject to: feasible for any PV output patterns

Objective function

Subject to

N-2 supply failure power

!n

kF kXxx

2

1

)(,=

Î"

*1

*2

kp : Load parameters for PV output pattern k

*１

*2 )(kXF : Feasible solution set for PV output pattern k

Find the system with minimum N-2 supply failure power when the weather is fine and feasible for any PV output patterns.

¤ Overload cases

Oct. 19, 2016 46

PV

PV

PV

PV

PV

PV

・・・PV

PV

PV

sys1 sys2 sys4

sys3 sys2 sys4

sys5 sys2

PV installed n nodesà 2n patterns*

*) PV output is assumed 0% or 100% because the most severe cases must be considered.

If not obtained feasible solution, generator output adjustment (control) takes place and adopt it as additional candidates.

Formulation (G adjustment)

Oct. 19, 2016 47

Minimize: amount of G adjustment(decision variables: G = output adjustment)Subject to: operation restrictions

Determine the amount of G adjustment and its locationfor N-1 overload banishing.

)(min1

-

=

+ +å m

MP

mm GG

maxmaxjjj FFF ££-

0)(1

=+ -

=

+å m

MP

mm GG

maxminmmm PPP ££

Objective function

Constraints

: Amount of G adjustment

: Limits of line flows

: Power balance

: Limits of G outputs

Flowchart of the algorithm

Oct. 19, 2016 48

Numerical simulations

¤ Data¤ Total load: 3,000MW¤ 109 nodes, 138 branches

¤ (Same as before)

¤ PV assumption¤ Divided into 3 areas (A, B, C)¤ PV Install conditions

¤ (Same as before)

¤ Outage works¤ Target = L57 (1 cct)¤ Target = L30 (1 cct) ß

overload case

Oct. 19, 2016 49

Case Area A Area B Area C1 1 1 12 1 1 03 1 0 14 0 1 15 1 0 06 0 1 07 0 0 18 0 0 0

Area A

Area C

Area B

L30

×

×L57

Simulation results

¤ All weather feasible¤ Obtained a solution feasible for all weather

conditions¤ Shaded part of the table below¤ Covering all cases

Oct. 19, 2016 50

Case off→on on→off Connection change off→on on→off Connection change off→on on→off Connection change off→on on→off Connection change

1 L107 L29 L129(N13→N14) L107 L29 L107 L29 L127(N13→N14) L107 L29 L128(N14→N13)

2 L107 L29 L107 L29 L127(N13→N14) L107 L29 L128(N14→N13) L107 L29 L131(N16→N17)





7 L107 L29 L127(N13→N14) L107 L29 L30(N14→N13) L110 L29 L107 L29 L30(N14→N13)

8 L107 L29 L127(N13→N14) L107 L29 L30(N14→N13) L110 L29 L110 L29 L131(N16→N17)

4th

Ranking

1st 2nd 3rd

Simulation results

¤ Overload case¤ In Case 7 & 8, no feasible system obtained.¤ à Additional G adjustment works well.

¤ Shaded part of the table below

Oct. 19, 2016 51

Case off→on on→off Connection change off→on on→off Connection change off→on on→off Connection change off→on on→off Connection change

1 L107 L30 L129(N13→N14) L107 L30 L107 L30 L128(N14→N13) L107 L29 L131(N16→N17)






7 L107 L30

8 L107 L30

Ranking

1st 2nd 3rd 4th

Conclusions

¤ Summary¤ PV penetration affects outage work planning

¤ It has been found out that the number of feasible outage work systems decreases depending on PV penetration.

¤ A new method has been proposed¤ To avoid overload cases.

¤ Future works¤ PV penetration assumption should be brushed up.¤ PV installation patterns should be analyzed more in detail.¤ PV output classification (area) should be studied based

on actual data.

Oct. 19, 2016 52

an outage work planning under the weather uncertainties · 2016-12-13 · transient stability...

Documents