1

A Novel Approach to Disco Planning in Electricity Markets: Mathematical Model

Ayed A. S. Algarni, Student Member, IEEE, Kankar Bhattacharya, Senior Member, IEEE
FCk FCapk Im{Iij,t} N NF NL PD PLoss QLoss Ps, Pg, P Pk Pinj PMax PMin PsMax Pf Pf1 QD Qk Qinj QLoss QMax QMin Qs, Qg, Q QsMax QCSTQ Re{Iij,t} Sk SDG SDGCAP SFk ST STCAP Feeder investment cost, $/p.u. MVA Feeder capacity investment status at any year of the planning period, p.u.MVA Imaginary component of injected current, p.u. Total number of distribution system buses Total number of system feeders Total number of system load buses Active power demand at bus, p.u.MW System active power loss, p.u. MW System reactive power loss, p.u. MVar Active power supplied through substation bus, DG bus, or either of them respectively Active power flow on a feeder branch, p.u.MW Active power injected at a bus, p.u.MW DG maximum capacity limit for active power, p.u.MW DG minimum capacity limit for active power, p.u.MW Disco substation transformer active power capacity limit, p.u. MW Power factor of DG power generation (0 to 1) Power factor of grid power import via substation buses (between 0-1) Reactive power demand at bus, p.u.MVar Reactive power flow on feeder branch, p.u.MVar Reactive power injected at a bus, p.u.MVar System reactive power loss, p.u. MVar Maximum DG capacity limit for reactive power, p.u.MVar Minimum DG capacity limit for reactive power, p.u.MVar Reactive power supplied through substation bus, DG bus, or both of them, respectively Disco substation transformer reactive power capacity limit, p.u.MVar Payment for reactive power from DG, $/MVarh Active power component of injected current, p.u. Complex power flow in feeder k, p.u.MVA Investment in DG capacity at a bus, p.u.MVA Maximum limit of installed DG capacity at a bus, p.u.MVA Feeder capacity installed in current year, p.u.MVA Investment in transformer capacity at a bus, p.u.MVA Maximum limit of installed transformer capacity at a bus, p.u.MVA

Abstract A novel approach to address the disco planning issues in a competitive electricity market environment is presented in this paper. Since there exists a variety of electricity markets and associated products and services (i.e. energy markets, ancillary service markets, etc.) which work in different time-frames, the mathematical model presented here uses the ac power flow, and the planning problem is a dynamic model, formulated to tackle all these issues simultaneously. The effects of DG injections on the distribution network, both in the short-term operations (impact on losses) and in long-term planning (impact on feeder capacity) is combined together and mathematically formulated within a comprehensive objective function for planning. Novel goodness factors accounting for DG contribution to system loss reductions and deferment in feeder capacity upgrades are formulated as monetary values. Index Terms Distributed generation, disco planning, incremental loss index, goodness factors, feeder capacity deferral

I. NOMENCLATURE i, j k g s t a, b, c J1 r U1 U2 U3 CPFeeder CQFeeder CDG DGCap Gij, Bij Set of all system buses Index for feeder branch Set of buses with DG units, g i Set of disco substation buses, s i Index for a period of the planning problem, year Operating cost parameters of a DG unit Disco objective function, $/h Discount rate, % Binary variable for DG investment (1=yes, 0=No) Binary variable for feeder k investment (1=yes, 0=No) Binary variable for transformer investment at substation s (1=yes, 0=No) Active power component of feeder capacity cost, $/MW Reactive power component of feeder capacity cost, in $/MVAr DG investment cost, $/p.u.MVA Updated capacity investment in DG over the horizon year of planning, p.u.MVA Conductance, susceptance of feeder branch, p.u.

_____________________________________________________________ The first author gratefully acknowledges the funding support received from King Abdulaziz City for Science and Technology (KACST), Riyadh, Saudi Arabia for supporting his Ph.D. research at the University of Waterloo. The authors are with the Department of Electrical and Computer Engineering, University of Waterloo, Ontario, N2L 3G1, Canada (e-mail: aalgarni@uwaterloo.ca)

978-1-4244-3811-2/09/$25.00 2009 IEEE

2 T TC TCap V VMin, VMax Yij Loss Loss Horizon year of planning Transformer investment cost, $/p.u. MVA Transformer capacity updated dynamically over the planning horizon, p.u.MVA Bus voltage, p.u. Minimum and maximum limit on bus voltage, p.u. Magnitude of complex Y-bus matrix elements Goodness factor of active power injection at a bus and its influence on feeder losses, $/p.u.MW Goodness factor of reactive power injection at a bus and its influence on feeder losses, $/p.u.MVar Goodness factor of active power injection at a bus and its impact on feeder loading/unloading, $/p.u.MW Goodness factor of reactive power injection at a bus and its impact on feeder loading/unloading, $/p.u.MVar Energy market price, $/MWh Payment for reactive power supply, $/MVarh Self active and reactive power Incremental Loss Indices (ILI) Mutual active and reactive power ILI Self active and reactive power Incremental Feeder Loading Indices (IFLI) Mutual active and reactive power BLI Angles of complex Y-bus matrix elements Voltage angle at a bus, rad II. INTRODUCTION and also captures with more details the electrical characteristics of each entity being available in the system, hence the operational efficiency can be maximized. In this paper a comprehensive optimization model aiming to assist the planner in solving the DG siting and sizing problem is proposed. The model also includes plan decisions for feeders and substation transformers over the horizon year of planning. The main features of the proposed planning approach are as follows: If DG technologies are not taken into consideration in explicit detail, except for using their average cost data, there is no difference between locations in the system and all locations have the same priority and features. However, when technology choices are incorporated in the model, such as specifying whether the DG is windturbine DG or a solar DG, etc., location aspects need to be considered for modeling. This paper does not consider the detailed features of the DG and hence the location aspects are ignored. Since technology choices are not considered, all DG units are considered to have the same cost data for their capacity and operational costs. This is done in order to first determine the most effective location of installation of DG units and their impact on system planning. A full ac-power flow model is employed and thus both active and reactive power flows are considered. This provides more indepth perspective on the system operation in the presence of new DG capacity. The planning model discussed in this paper considers dynamic addition of substation transformers and /or expansion decisions on feeder capacity as well as new DG capacity year by year. These are incorporated in the objective function using binary variables for each of these decisions. Furthermore, the operational costs of both active power generation and reactive power supply from the DG or purchased from the grid are also included in the model. A DG investment planning model is presented in [1] using various reliability indices in order to determine the optimal DG locations and sizes. It was concluded that although the DG addition may be the most expensive alternative, using the reliability techniques and the capital deferral credit obtained from disco, the DG option could become a cost-effective solution. A Benders decomposition solution is used in [2] to determine optimal DG sitings on network buses. The model considers stochastic nature of generator outputs, with power flows represented using linear models. A locational marginal pricing approach for the siting and sizing of DG units is proposed in [3]. A heuristic approach to DG investment planning that uses an optimization model is presented in [4]. In meeting demand requirements the disco has three options: purchase power from the grid, install DG capacity, or shed load. The planning method examines the benefit-cost ratio of each possible DG addition option and selects those who have a net benefit. In another work, binary variables are used for distribution system planning [5] and the problem attempts to

Feeder
Feeder P Q Pi, Qi Pi, Qi Pk,i, Qk,i Pk,i ,Qk,i

ij

he problems involved in siting and sizing of distributed generation (DG) sources, the related operational issues and the emerging retail competition effects have changed drastically the distributed electrical energy managements, short-term and long-term distribution system planning strategies. Moreover, it is obvious that with the introduction of deregulation and spot markets in the electric power sector, most of planning issues are no longer the same. For example, the injected power from DG sources at a bus has the ability to either reduce the net power flow in some feeders while increasing the power flow in the other feeders. Therefore, the long-term plans for feeder capacity addition are now not only be affected by the forecasted demand growth over the planning horizon, but also on the location and capacity of DG that may be commissioned over this period. This can play a vital role in precisely determining feeder capacity addition decisions. The benefits of installing DG units include deferment of distribution system upgrades. A DG unit investment should hence provide for the same level of reliability as the distribution system upgrade it is deferring. In addition to that, in the deregulated environment, the financial risks and the cost minimization approaches will require sophisticated and more complex calculations in order to solve the planning problem. The disco operator needs a modern planning tool suits the quick and increasing changes,

3 minimize the cost of supplying the demand by either purchasing electricity from the grid, inter-ties, or investing in DG. In [6], an approach to quantification of the distribution network capacity deferral value of DGs was presented. It was found that the most significant benefits from such deferral are obtained when DGs are installed at the end of long feeders and near load pockets. III. INCREMENTAL LOSS INDICES, INCREMENTAL FEEDER LOADING INDICES AND THEIR GOODNESS FACTORS A. Incremental Loss Indices (ILI) The active/reactive power self-ILIs (1) denoted by P and Q, are defined as the incremental change in system active/reactive power losses due to incremental active/reactive power injections at a bus i. The incremental active/reactive power injections can be represented as small changes in active/reactive power demand at a bus, generation remaining constant. The active/reactive mutual-IFLIs (6) denoted by P and Q can similarly be defined as the incremental change in feeder k active/reactive power flow due to reactive/active power injection at a bus i.
P k ,i =

Qk Pk Q k = , i Pinj ,i Qinj ,i

(6)

It should be noted that the indices and will take positive values when the active/reactive power injected from a given DG unit loads (i.e. stresses) the feeder k while it will take negative values when the active/reactive power injected by the DG unloads (i.e. relieves) the feeder k. C. Goodness Factors The contribution of active/reactive power injection by a DG unit to overall system loss reduction or feeder loading/unloading can be better understood if the ILIs and IFLIs are represented in corresponding economic terms. A money value is attached to these indices to arrive at a goodness factor for a DG at a given bus. The goodness factors therefore indicate the relative importance and contribution of 1 unit of DG power (active or reactive) at a bus, as compared to a DG located at another bus in the distribution system. The goodness factors can also be interpreted as the savings in operational costs (from ILIs) because of reduction in system loss and in investment costs (from IFLIs) because of deferral of feeder and transformer capacity investment costs, due to DG power injection at a bus. The proposed goodness factors are given below in (7)-(10).

iP =

PLoss Pinj ,i

Q i =

QLoss Qinj ,i

(1)

These indices can be obtained from the Lagrange multipliers of an active/reactive power loss minimizing optimal power flow (OPF) model. The corresponding active and reactive power loss functions, PLoss and QLoss respectively, are given below:
PLoss = 1 N N Gi, j Vi2 + V j2 2ViV j cos j i 2 i =1 j =1 1 N N Bi, j Vi2 + V j2 2ViV j cos j i 2 i =1 j =1

)]

(2)

QLoss =

)]

(3)

The active/reactive power mutual-ILIs (4) denoted by P and Q respectively, are defined as the incremental change in system active/reactive power losses due to an incremental reactive/active power injection at a bus i. P Q (4) iP = Loss iQ = Loss Qinj ,i Pinj ,i In any distribution system configuration (radial or networked) power injection at different buses would result in differing effects on the system. It is important to note that in a radial system, a feeder section with multiple possible candidate buses for DG injections will have correlated indices because of the inherent effects, the injection of power, will have on the other buses. On the other hand, in a networked configuration the DG injections will result in arbitrary counter-flows in the distribution system and hence the ILIs will not be correlated. B. Incremental Feeder Loading Indices (IFLI) The active/reactive power self-IFLIs (5) denoted by P and are defined as the incremental change in feeder k active/reactive power flow due to active/reactive power injection at a bus i.
Q

iLoss = P iP + Q iQ
P P iLoss = Q Q i + i
Feeder iFeeder = C P NF k =1 Feeder NF P iFeeder = C P k ,i k =1 Q Feeder kP,i + CQ k ,i NF k =1 Feeder NF Q + CQ k ,i k =1

(7) (8) (9) (10)

P k ,i =

Pk Pinj ,i

Q k ,i =

Qk Qinj ,i

(5)

In (7), iLoss is the goodness factor denoting the impact on system losses, in dollar terms, because of active power injection from a DG at bus i. Similarly, in (8), iLoss is the goodness factor denoting the impact on system losses, in dollar terms, because of reactive power injection from a DG at bus i. The parameter P is the market price of energy while Q is the payment made by the disco for reactive power supply. In (9), iFeeder is the goodness factor denoting the dollar value equivalent of the net feeder unloading because of active power injection by a DG at bus i. Similarly, in (10), iFeeder is the goodness factor denoting the dollar value equivalent of net feeder power unloading because of reactive power injection by a DG at bus i. The parameters CPFeeder (in $/MW) and CQFeeder (in $/MVAr) are the active and reactive components

4 respectively of the feeder capacity cost (in $/MVA) that is deferred or added because of the DG. IV. DISCO OPTIMAL PLANNING FRAMEWORK The discos objective function given in (11) represents the long-term discounted cost of investments in new DG capacity, new feeder capacity and new sub-station transformers. It also considers the long-term discounted cost of operation of DG units, purchase cost of power from grid, and cost associated with reactive power generation from DG units and that purchased from the grid. One novel feature of this objective function is the inclusion of the last two terms in (11) that accounts for a cost credit to the disco (denoted by the negative sign). This cost, which is also discounted over the planning horizon, represents the monetary savings to the dicso accrued from reduced feeder losses because of DG power and equivalent monetary savings accrued from feeder unloading because of DG power injected at a bus. Objective Function

Qs,t + Qg ,t QDi ,t = Vi ,t V j ,t Yij sin ( ij + j ,t i ,t )

j

(13)

Grid Purchase Constraints and Dynamic Capacity Updates These constraints ensure that the active and reactive power purchased by the disco from the external grid / market is within the transfer limits imposed by disco transformer capacity, from year to year taking into account the existing capacity of year t-1 and the new capacity added in year t.

TCapi, t = TCapi, t 1 + STi U 3i, t t > 1, i = s

STi STiCAP t = 1, i = s

(14)

(15) (16)

Pi, t TCapi , t pf 1

i = s

J=
S i s i

( FCk SFk ,t U 2k ,t t t =1(1 + r ) k

g i

NF

Qi, t TCapi, t 1 pf 12

i = s

(17)

+ TCi STi ,t U 3i ,t + CDGi SDGi ,t U 1i,t + tP Pi,t + tQ Qi ,t + (ai Pi2 i ,t + ci ) (11) ,t + bi P

i i s g

+ Qi ,t QCSTiQ
i g

i =1

( iLoss + iFeeder ) Pi,t

Bus voltage limits These limits ensure acceptable voltages at all buses and that the bus voltages do not drop below certain specified values, which is particularly an issue in distribution systems. The substation bus voltage is held at a constant value, similar to the slack bus in classical load flow programs.

i =1

( iLoss + iFeeder ) Qi,t )

Vi , t = Constant ViMin Vi, t ViMax

i = s i NL
(18)

The discos objective will be slightly modified when the DG units are investor-owned instead of utility-owned. In such a case, the DG units will inject a pre-determined amount of power into the disco system, and will not be included in the discos dispatch program. Further analysis and required mathematical modifications will be considered in another research work. Network Equations The active and reactive power flow equations are modified to include grid purchased active and reactive power through disco transformers as well as DG generated active and reactive power. It is to be noted that the index of time, t appears with all the variables here and the model will simultaneously determine the optimal dispatch for all years of the planning horizon.

Generation Limit on DG Units, Dynamic Capacity Updates These constraints ensure that the power dispatched from a DG unit is within its maximum and minimum limits of generation in a specific year (19)-(20). It also ensures that the DG capacity at a bus is aggregated over the planning horizon considering the existing capacity at year t-1 and the new capacity added in a year t, in a dynamic manner (21)-(23).

Pi, t DGCapi , t pf Qi, t DGCapi, t 1 pf 2

i = g i = g

(19) (20)

DGCapi, t = DGCapi , t 1 + SDGi, t U 1i , t t > 1, i = g DGCapi, t SDGi , t U 1i, t t = 1, i = g

(21)

Ps,t + Pg ,t PDi ,t = Vi,t V j ,t Yij cos ( ij + j ,t i,t )

j

(12)

(22)

SDGi, t SDGiCAP

i = g

(23)

Feeder Limits and Dynamic Capacity Updates The power carrying capability of distribution feeders are limited by the feeder current limits, which are consequently represented by their MVA limits (Sij). These limits are determined by the total capacity in a given year, which is dynamically related to the previous years capacity and the new feeder capacity added in the current year (25).

used appropriately in the planning objective function. This model clearly demonstrates the effectiveness and contribution of DG units in the distribution network both in the short-term and long-term framework.

FCapk , t = SFk

t =1

(24)

FCapk , t = FCapk , t 1 + SFk U 2k , t t >1 S k , t FCapk , t t

(25)

(26)

The complex power on the feeders can be determined from the power flow equations and the final relations are given as under: (27) Re I k ,t = f (Vi ,t , i ,t )

{ }

Im I k , t = f (Vi ,t , i, t )

{ }

(28) (29)
Fig.1. Schematic diagram of the 18-bus distribution system (highlighted buses are the DG buses, set g)

Sij , t = f Re I k , t , Im I k , t

( { } { })

VII. REFERENCES The disco operation model presented above is a mixedinteger nonlinear programming problem, which is solved using the GAMS/DICOPT solver [7]. V. SYSTEMS ANALYSIS AND CASE STUDY A. System under study A 18-bus distribution system (Fig.1) will be considered for the proposed planning study. The distribution system has three substation transformers connected it with the external grid / market at buses-1, 11 and 16. This system has been extracted from the well-known IEEE-30 bus system by considering only the 33 kV network. The network parameters and the load at each bus are known and used in this work without any modification. VI. CONCLUDING REMARKS This paper presents a detailed mathematical model of a novel planning framework for discos operating in competitive electricity markets. The planning framework includes investment in DG capacity, distribution system feeder addition / expansion and substation transformers capacity addition. The effects of power injected from DG on system network either on loss reduction or feeder capacity deferment are formulated in the model by using goodness factors. A monetary value is attached to these factors so that they can be
[7] [1] A. Chowdhury, S. K. Agarwal, and D. O. Koval, Reliability modeling of distributed generation in conventional distribution systems planning and analysis, IEEE Ind. Appl. Mag., vol. 39, no. 5, pp. 14931498, Sep. 2003. S. McCusker and B. F. Hobbs, A nested decomposition approach to locating distributed generation in a multiarea power system, Networks and Spatial Economics, vol. 3, no. 2, pp.197223, Jun. 2003. A. Agalgaonkar, S. Kulkarni, S. Khaparde, and S. Soman, Placement and penetration of distributed generation under standard market design, International Journal of Emerging Electric Power Systems, vol. 1, no. 1, pp. 116, 2004. W. El-Khattam, K. Bhattacharya, Y. Hegazy, and M. Salama, Optimal investment planning for distributed generation in a competitive electricity market, IEEE Trans. Power Syst., vol. 19, no. 3, pp. 16741684, Aug. 2004. W. El-Khattam, Y. Hegazy, and M. Salama, An integrated distributed generation optimization model for distribution system planning, IEEE Trans. Power Syst., vol. 20, no. 2, pp. 11581165, May 2005. H. A. Gil and G. Joos, On the quantification of the network capacity deferral value of distributed generation, IEEE Transactions on Power Systems, vol. 21, pp. 1592-1599, Nov. 2006. GAMS release 2.25, in A Users Guide, GAMS Development Corporation, 1998.

[2] [3]

[4]

[5] [6]