Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Abstract: Generation shedding is one of the most effective and GR Generation Reduction
commonly used corrective actions to mitigate insecurity of power GS or
systems under large disturbances. This paper presents the Generation Shedding
Genshed
implementation of an optimization scheme for generation CTG Contingency
shedding allocation within an EMS system using integer RAS Remedial Action Scheme
programming. The approach allows addressing at the same time,
EMS Energy Management System
various security concerns including thermal limit violation,
transient instability, and voltage instability taking into account
distribution factors of plants in different locations, which II. INTRODUCTION
contribute differently to security enhancement. The optimization
scheme can receive security enhancement requirements from
various sources, such as online DSA (Dynamic Security
R EMDIAL action schemes (RAS) are widely used in
power systems to enhance power system security [1-3],
which involve various types of corrective actions, such as
Assessment), offline planning study and manual input.
generation shedding, line tripping and shunt switching.
Keywords—generation shedding, corrective actions, According to the security problems including transient
preventive actions, generation tripping, generation dumping, stability, voltage stability and thermal limit violation, the
linear programming, security enhancement, RAS, remedial corrective actions would be different combinations of those
action schemes, transient stability, voltage stability, NP-complete, types of actions. Among them, generation shedding is one of
knapsack problem, optimization the most commonly used and effective actions to enhance
power system security of various types after severe system
disturbances. Hence, this note will be focused on the
I. NOMENCLATURE
discussion of optimization for generation shedding allocation,
which consists in allocating the total generation shedding
Ppl max(j): MW rating of plant j; amount required for a particular contingency to units of
Ppl rdc(j): Total MW generation reduction at plant j; different plants that are associated with the contingency in the
C(i): Total MW armed for generation shedding (or RAS schemes. In particular, the optimization formulation of
generation reduction) for contingency i; as cost the generation shedding will be discussed in detail for both
function of the optimization scheme.
under shedding and over shedding scenarios. In general,
dfj: Distribution factor of plant j;
under-shedding is caused by certain conflicts between
σSER(i,k): kth Security Enhancement Requirement in MW
requirements for shedding amount and the number of units
for contingency i;
SERi(k): kth Security Enhancement Requirement in MW online for post-contingency systems. From system planning
for contingency i; (interchangeable with study point of view, under-shedding is not allowed and the
σSER(i,k)); conflicts shall be resolved as soon as possible to remove
Dush_max(i): Dush_max(i) is the maximum Under-SHedding under-shedding. However, under-shedding can happen in real-
amount in MW within all security enhancement time operations and implementation of the solution to remove
requirements for contingency i under-shedding may take sometimes. Therefore, the
Pun(j,m): MW output of unit m at plant j; generation shedding allocation scheme has to take into
Pmax(j,m): MW rating of unit m at plant j; account the scenarios of under-shedding when providing RAS
Dush(i,k): Dush(i,k) is the Under-SHedding amount in arming patterns. It is worth noting that “under-shedding”
MW for the kth security enhancement discussed in this paper is relative to the shedding requirements
requirement for contingency i; based on the system performance assessment from either off-
S(j,m): Shedding factor (0: no shed; 1: shed) of unit m line planning study or online DSA. The discussion on how to
at plant j; establish the shedding requirements is beyond the scope of
this paper.
Ziwen Yao is with BC Hydro, Vancouver, BC, Canada (e-mail: In real-time operations, one particular disturbance may
Michael.Yao@bchydro.com). cause multiple security problems, which requires one set of
Qing Zhu is with BC Hydro, Vancouver, BC, Canada (e-mail:
generation shedding to address all the security problems. At
Qing.Zhu@bchydro.com)).
the same time, undesired impacts of generation shedding on 1) What the security enhancement amount required for each
the post-contingency systems shall be minimized as much as security concern would be needed to mitigate the problem.
possible. Therefore, certain optimization technique shall be In the case of thermal limit violation, the amount of
used to set up the generation shedding arming pattern to overloading plus some margin will be the security
resolve potential conflicts between different requirements of enhancement amount required. If the contingency can
insecurity mitigation. cause multiple security problems; multiple security
In reality, since genshed is a discrete control action, the enhancement amounts are required.
actual total MW of units armed for shedding for a contingency 2) How these multiple security enhancement amounts can be
is seldom exactly the same as required and will be either allocated to units for shedding that are associated with the
contingency in the RAS scheme. According to the location
under-shedding or over-shedding according to shedding
of a unit, its contribution of the MW shed to a particular
requirements. On one hand, under-shedding cannot mitigate
security enhancement will be a portion of its total MW
the problems as required, thus it shall be avoided as much as
shed, which is modeled as distribution factor (greater than
possible; on the other hand, however, too much over-shedding 0.0 and less than 1.0) of the unit (or the plant assuming
may reduce the capability of meeting the load demand and/or that all units in the plant are electrically in the same
voltage control capability of regions where the number of location).
online units in post-contingency system is insufficient, thus
risking voltage limit violation and / or self-excitation of
remaining generating units after heavy generation shedding
that will reduce significantly loading on the remaining 4 16 6 8
transmission lines.
3
Therefore, the generation shedding arming patterns must be 9
calculated diligently using robust optimization processes, 2
which shall be able to resolve all insecurity issues caused by 5 7
contingencies, on one hand; and shall not create new
insecurity problems, such as unit self-excitation and voltage
limit violations, on the other hand. Also, it shall maximize 20 19 10
power supply to the post contingency systems.
British Columbia (BC) Hydro has installed massive RAS 1 ∞
12
schemes in BC grid to enhance its security against various ∞
credible contingencies. Many of these schemes address single
15 11
security concern, such as transient stability or thermal limit
violation on single equipment, which is modeled as single
constraint in the optimization formulation for genshed
allocation. And many of them are designed to mitigate 18 17
overloading on multiple devices, voltage instability and
13
transient instability. Therefore, multiple constraints shall be
taken into account in genshed allocation optimization
processes to address multiple security issues caused by each 14
contingency.
Figure 1: System studied
This note presents an implementation of integer
programming within an EMS system, which calculates the
Given the security enhancement amount required and
generation shedding arming patterns subject to multiple
distribution factors of units associated with the contingencies,
constraints in real-time for the bulk power systems of BC
among other things, the basic requirements for “optimal”
Hydro.
generation shedding allocations are as follows:
a) Meet all security enhancement requirements as much as
III. PROBLEM FORMULATION
possible;
In a typical bulk power system as illustrated in Figure 1, in b) Keep as much as possible power supply online after
occurrence of a contingency of losing L1_2, the remaining generation shedding to meet the load demand for the post-
devices such as L3_5, L5_7, and T7_9 may be overloaded and contingency systems;
units in plant 4 may become transient instable depending on c) Prevent as much as possible the post-contingency system
the pre-outage system conditions. In the case when the from new insecurity problems potentially caused by
contingency causes insecurity problem to the system, generation shedding. The potential insecurity problems
generation shedding is one of the most effective corrective caused by generation shedding would be i) unit self-
actions to mitigate the problem. However, we shall answer the excitation and ii) loss of voltage support. To prevent these
following two basic questions among other things when two problems, the requirement of “keep a minimum
setting up the generation shedding arming patterns: number of units online” can be specified as a constraint
3
Ppl max(4
Min [C(i) |min(Dush_max(i)] (1.1)
SER1(2) SER1(3)
Subject to:
n Feasible
(1.2)
∑ (df
j
j ⋅ Ppl _ rdc ( j )) ≥ σ SER (i, k ) Zone
k = 1, …, hi;
0 ≤ df j ≤ 1, SER1(1) C”(i)
C’(i)
0 ≤ Ppl _ rdc ( j ) ≤ Ppl _ max ( j ),
n
C (i) = ∑ ( Ppl _ rdc ( j ))
j
Ppl rdc(6)
⎧0, th
k constra of Eq.(1.2) Ppl_max(6
⎪⎪ is satisfied
⎨ n
[SERi(k): Eq. (1.2)]
⎪σ SER (i, k ) − ∑ (df j ⋅ Ppl _ rdc ( j )) > 0, otherwise
Figure 2: Over-Shedding is feasible
⎪⎩ j
4
Ppl_rdc(4)
SER1(5)
1) Minimize the total MW armed for shedding when over-
SER1(2)
shedding is feasible;
SER1(1) 2) Minimize under-shedding if over-shedding is infeasible
and maximize power supply for post-contingency systems;
Feasible
Ppl_max(4 Zone
Min [C(i) |min(Dush_max(i))] (2.1)
Subject to:
n nj
∑ (df ⋅ ∑ S ( j, m) ⋅ P
j =1
j
m =1
un ( j , m)) ≥ σ SER (i, k ) (2.2)
k = 1, …, hi;
C(i) 0 ≤ df j ≤ 1,
⎧0, if unit m of plant j is not armed
S ( j , m) = ⎨
SER1(4) ⎩1, if unit m of plant j is armed
SER1(3)
0 ≤ Pun ( j , m) ≤ Pun _ max ( j , m),
n nj
C (i ) = ∑∑ S ( j , m) ⋅ Pun ( j , m)
Ppl_max(6) Ppl_rdc(6)
match exactly the desired generation reduction. ⎪σ SER (i, k ) − ∑ ( df j ⋅ ∑ S ( j , m) ⋅ Pun ( j , m)), otherwise
⎪⎩ j =1 m =1
Hence, the general linear programming formulation will be
C(i)
is the total MW armed for generation
modified to an integer programming or a typical knapsack
shedding for contingency i;
problem presented in the next section.
min(Dush_max(i)) is the minimum Dush_max(i) within all
Ppl_rdc(4) possible combinations of units contributing
SER1(2)
C”(i)
to generation shedding.
C’(i)
hi is the number of potential security problems
B
Ppl_max(4) to be mitigated for contingency i;
dfj is the distribution factor of plant j;
Pun(j,m) is the MW output of unit m at plant j;
Pmax(j,m) is the MW rating of unit m at plant j;
SER1(1)
S(j,m) is the shedding factor (0: no shed; 1: shed)
Feasible
A Zone of unit m at plant j;
σSER(i,k) is kth Security Enhancement Requirement in
MW for contingency i;
Dush(i,k) is the Under-SHedding amount in MW for
the kth security enhancement requirement for
contingency i;
Dush_max(i) is the maximum Under-SHedding amount in
C(i)
MW within all security enhancement
requirements for contingency i;
Ppl_max(6) Ppl_rdc(6)
The objective function, Min [C(i) |min(Dush_max(i))] is to
[SERi(k): Eq. (1.2)] minimize the total MW to be shed within the minimum
Figure 3-b: Under-Shedding Case II
Dush_max(i) of all possible combinations of units arming for
B. Optimization for Generation Shedding generation shedding.
Unlike generation reduction, generation shedding is a
discrete control action that can only determine whether a unit IV. IMPLEMENTATION OF THE OPTIMIZATION ALGORITHM
is armed for genshed or not. The optimization algorithm is implemented in BC Hydro’s
The general objective of the optimization processes for EMS to set up generation shedding arming patterns in real-
generation shedding allocation is to reach the following goals, time, which are automatically updated every 4 minutes.
which are formulated as Eq.(2):
5
Figure 4 illustrates the functional diagram of the In the implementation, when preparing the list of candidate
optimization program of iteration for one contingency. The units, the pre-processing module eliminates the unsheddable
entire generation shedding allocation process will go through units that are in synchronous condenser mode or generating
all the contingencies within the installed RAS schemes. close to zero MW output, in ramping mode, or being disabled
manually by operators for any operation reasons.
Furthermore, a pre-defined tolerance (d0 – see Figure 4)
Begin of
contingency i can be used in the optimization process to increase the chance
of early stop without exhausting all possible combinations.
Database of
Database of Initialization:
Candidate
In addition, the following constraints are also taken into
Security Dmin, Cmin, d0…
Enhancement Units account in EMS implementation:
Requirements Participating in • Alignment of armed units within a group of contingencies:
Related Data Generation
Input Data: SENHi(k), Figure 5 shows an arming pattern that respects this
dfi, candidate units
requirement. In this example, units 5 and 6 of plant 16
armed for different contingencies. This is to avoid
Calculate Dush(i, k) for all k and Dush_max(i), C(i) for
one combination of units armed that respects the
unnecessary heavy over-shedding in case of potential
requirement of keeping minimum number of units multiple contingencies within a group even though the
online after generation shedding. probability of such multiple contingencies is extremely
low.
• Preserve the previous arming pattern as much as possible
Y
|Dush_max(i)| < to reduce frequent changes of arming patterns: this is to
Next Combination
N
avoid massive arming data being transferred back and forth
between control center and substation RTUs, thus to
> =
Dush_max(i) < Dmin improve the communication performance.
?
N
< These two requirements are modeled as weighting factors
Early Stop
C(i)
Store optimal combination and set assigned to unit generation in the objective function C(i).
Y
Dmin = Dush max(i), Cmin = C(i) However, a detailed discussion on those topics is beyond the
scope of the paper.
An example of the summary of generation shedding Theoretically, with generation reduction (GR), the thermal
amount is illustrated in Table 4, which is extracted from figure overloading on L2-19 is relieved exactly just below its rating;
5. On column “Shed Total”, “[L2-20]” is the indication of the whereas generation shedding (GS) with over-shedding
dominant security concern that is thermal overloading the line relieves the overloading slightly more than GR.
L2-20 in this case.
Table 2: Unit MW of pre/post-contingency systems
Table 1: Power flow of the system studied Basecase GR GS
Basecase Post- Post- Post- Plant-4 MW MVAr MW MW MVAr
Branch CTG CTG CTG MVA G1 52.00 10.29 35.00 52.00 10.29
G2 52.00 8.66 35.00 52.00 8.66
(FRM- MW no RAS w/GR w/GS
G3 52.00 9.77 34.00 52.00 9.77
TO) MVAr MW MW MW G4 52.00 11.36 34.00 52.00 11.36
MVAr MVAr MVAr Plant-16
L1-2 -176.67 0 0 0 G5 76.50 4.09 59.00 0.00 0.00
(CTG) 75.30 0 0 0 G6 76.50 0.00 59.00 0.00 0.00
L2-20 0 0 0 0 G7 76.50 9.11 59.00 76.50 9.11
(Outage) 0 0 0 0 G8 76.50 7.44 59.00 76.50 7.44
L2-19 200.46 287.89 229.13 221.18 230.0 Plant-6
-2.39 -8.97 -6.74 -10.59 G1 65.88 -5.60 65.88 65.88 -5.60
L15-20 37.27 168.28 37.15 37.14 G2 63.78 2.91 63.78 63.78 2.91
11.83 -12.83 11.71 11.70 G3 17.94 0.68 17.94 17.94 0.68
L19-15 222.45 307.12 250.29 242.57 Plant-8
-10.95 -37.43 -19.26 -20.30 G1 60.22 -18.51 60.22 60.22 -18.51
L15-18 142.05 36.95 155.69 151.85
-17.65 11.53 -21.46 -21.13
As the input to the optimization process, the original
L15-17 134.73 152.06 145.59 142.52
-2.15 86.18 -4.86 -4.71 “Amount required for security enhancement” from off-line
L18-14 104.94 145.72 118.25 114.53 planning study is in a formula format involving the power
-28.87 -51.80 -35.05 -34.08 flow variables of related transmission lines and unit
L17-14 44.71 76.28 55.01 52.18 generation, which is defined in a system operating order. In
-42.60 -58.34 -46.77 -46.14
L4-3 199.16 198.94 133.99 199.10 EMS, the actual amount required for the security enhancement
-16.20 3.71 12.25 -10.18 presented in Table 3 is calculated in real-time every 4 minutes
L16-3 304.70 304.67 235.27 152.34 using the formula and state estimation solution.
-53.17 -33.07 -35.95 -23.33
L3-5 139.44 225.99 153.95 144.56
-13.31 -16.68 -12.45 -17.15 Table 3: Security enhancement requirements
L6-5 145.80 145.77 145.80 145.79 No. Security Amount df df df df
-18.41 -7.42 -16.91 -15.98 Concern required for plant plant plant plant
L5-7 284.34 369.37 298.66 289.39 security 4 16 6 8
-28.53 -35.83 -28.06 -30.78 enhancement
L8-7 25.88 25.91 25.88 25.89 1 L2-19 O/L 59.9 MW 0.43 0.43 0.28 0
-41.45 -32.28 -40.03 -39.76 2 Transient 80.0 MW 1.0 1.0 0 0
L9-10 307.82 391.11 321.88 312.77
-59.23 -84.49 -61.27 -61.63 The result of the generation shedding allocation is
L10-11 153.45 236.94 167.63 158.65
-126.03 -159.70 -129.24 -128.78
presented in a matrix format shown in figure 5, which assigns
L13-11 -235.89 -253.57 -235.24 -232.72 the units to be tripped for the corresponding contingency.
21.55 31.54 21.14 20.87
T3-2 364.46 277.58 215.28 206.92
-54.07 -10.96 -9.61 -14.48
T7-9 308.21 391.76 322.31 313.17
-42.78 -57.00 -43.21 -44.52
T12-11 -266.64 -202.42 -251.43 -257.51
-171.44 -200.08 -175.08 -174.13
Also, the major difference between GS and GR is the [3] Moshref, A.; Henville, C., Curtis, R., Morison, K., Albassam, L.,
Owayedh, M., El Said, O., Ashiq, M., “Design of a special protection
number of units kept online for post-contingency system, system to maintain system security at high import”, Power Engineering
which is most important for the post-contingency system to Society General Meeting, 2003, IEEE Vol. 1, 13-17 July 2003, p.319.
have enough voltage support. [4] Van Cutsem, T., Moors, C., Lefebvre, D., “Design of load shedding
schemes against voltage instability using combinatorial optimization”,
Power Engineering Society Winter Meeting, 2002. IEEE, Vol. 2, 27-31
Table 4: Arming pattern extracted from Figure 5 Jan. 2002, pp. 848 – 853.
Plant 4 Plant 16 Plant 6 Plant 8
CTG Shed SENH Actual df / df / df / df /
[5] A. J. Orman and H. P. Williams, “A Survey of Different Integer
Total Require SENH TTL TTL TTL TTL Programming Formulations of the Travelling Salesman Problem,” in
-ment Amount 1 2 3 4 5 6 7 8 1 2 3 1 Optimisation, Econometric and Financial Analysis, C. Gatu and E.
153.0 Kontoghiorghes, Eds. New York: Springer, 2006, vol. 9.
L1-2 59.9 65.8 0.43 / 0.43 / 0.28 / 0.0 /
[L2-20] 0.0 153.0 0.0 0.0 [6] D. L. Applegate, R. E. Bixby, V. Chvatal, and W. J. Cook, The Traveling
: indicating unit being armed; : indicating unit being disarmed; TTL: Salesman Problem: A Computational Study. Princeton, NJ: Princeton
Total MW to be shed at the plant; SENH: Security Enhancement; [L2-20]: University Press, 2006.
dominant security concern which is overloading L2-20 in the case study.
VII. REFERENCES
[1] Bertsch, J., Carnal, C., Karlson, D., McDaniel, J. Vu, K., “Wide-Area
Protection and Power System Utilization”, Proceedings of the IEEE, May
2005, Vol. 93, No. 5, pp. 997-1003
[2] Van Cutsem, T., “Voltage instability: phenomena, countermeasures, and
analysis methods”, Proceedings of the IEEE, Vol. 88, No. 2, Feb. 2000,
pp. 208 – 227.