A Direct Load Control Model for Virtual Power Plant Manageme

IEEE TRANSACTIONS ON POWER SYSTEMS,VOL.24,NO.2,MAY2009959

A Direct Load Control Model for Virtual

Power Plant Management

Nerea Ruiz,I?igo Cobelo,and JoséOyarzabal

Abstract—In the framework of liberalized electricity markets, distributed generation and controllable demand have the oppor-tunity to participate in the real-time operation of transmission and distribution networks.This may be done by using the virtual power plant(VPP)concept,which consists of aggregating the capacity of many distributed energy resources(DER)in order to make them more accessible and manageable across energy mar-kets.This paper provides an optimization algorithm to manage a VPP composed of a large number of customers with thermostati-cally controlled appliances.The algorithm,based on a direct load control(DLC),determines the optimal control schedules that an aggregator should apply to the controllable devices of the VPP in order to optimize load reduction over a speci?ed control period. The results de?ne the load reduction bid that the aggregator can present in the electricity market,thus helping to minimize network congestion and deviations between generation and demand.The proposed model,which is valid for both transmission and distri-bution networks,is tested on a real power system to demonstrate its applicability.

Index Terms—Air-conditioning,load management,optimal con-trol.

I.I NTRODUCTION

T HE progressive liberalization of the electricity sector and the subsequent transition to a decentralized system ben-e?ts the emergence of new ways of planning,managing and operating transmission and distribution networks.In this con-text,distributed generation and controllable demand can play an important role in participating in energy markets and providing ancillary services to system operators.However,the integration of these resources into power system operation is not simple: it requires a major change in the current network control struc-ture.This challenge can be met by using the virtual power plant (VPP)concept,which is based on the idea of aggregating the capacity of many distributed energy resources(DER)—genera-tion,storage,or demand—in order to create a single operating pro?le.In this way,inpidual DERs would gain visibility and manageability to system operators,optimizing their position and maximizing their revenue opportunities.Moreover,the system

Manuscript received July11,2008;revised September12,2008.First pub-lished March27,2009;current version published April22,2009.This work was carried out as part of the FENIX project,a collaborative research project supported by the European Commission,under the Sustainable Energy Systems Programme of the6th Framework Programme,contract No.518272.Paper no. TPWRS-00568-2008.

The authors are with LABEIN-Tecnalia,Parque Tecnológico de Bizkaia,48160Derio,Spain(e-mail:nruiz@labein.es;icobelo@labein.es; joseoyar@labein.es).

Digital Object Identi?er10.1109/TPWRS.2009.2016607would bene?t from an optimal use of the available resources and the ef?ciency of operation would be improved.A VPP is comparable to a conventional power plant with its own oper-ating characteristics such as schedule of generation,generation limits,and operating costs.

This paper provides a tool for managing a VPP composed of a large number of end-users with controllable appliances.The developed algorithm is based on direct load control(DLC),and allows calculation of optimal control strategies to be applied to controllable devices in order to obtain the maximum load re-duction over a speci?ed control period.In this way,an aggre-gator can determine its load reduction capability so as to de-?ne the corresponding load reduction bid to be presented in the electricity market.The developed model is valid for both trans-mission and distribution systems.To check the applicability of the proposed algorithm,an application to the Spanish devia-tion management market is considered.This market is called by the transmission system operator(TSO)when deviations be-tween generation and demand over300MW are expected be-tween two intraday markets.When this occurs,market agents have30min to send their offers.Currently,only generators and pumped storage power stations can participate in this market. This case study seeks to demonstrate that a portfolio of control-lable customers integrated into a VPP can contribute to real-time transmission system operation by participating in the deviation management market.In this paper,the terms“VPP”and“aggre-gator”are used indistinctly.

The study is focused on the aggregation of domestic and com-mercial customers with appliances that have thermal storage capabilities such as air-conditioning and electric space-heating systems.The possible control strategies are established by con-tract between the end-users and the aggregator and can consist of a temperature modi?cation of the thermostat settings or a dis-connection of the devices for a predetermined period.The algo-rithm developed is formulated using linear programming(LP). In the literature,several DLC algorithms have been devel-oped to determine the optimal load control schedules of groups of domestic devices[1]–[6].Most of them are based on linear programming[1],[3],[4],[6],or dynamic programming[2],[5], and try to minimize peak load[1]–[6]or electricity produc-tion cost[1],[2]over a certain time period.Lee and Wilkins [1]present a methodology for scheduling the control actions of water heaters.The persi?ed water heater load demand is modeled based on data obtained by monitoring different uncon-trolled residential water heaters.The payback or the shape of the controlled water heater demand for different control actions is also modeled.The method for the prediction of the persi?ed controlled water heater load shape is also based on empirical

0885-8950/$25.00?2009IEEE

960IEEE TRANSACTIONS ON POWER SYSTEMS,VOL.24,NO.2,MAY2009

data obtained from?eld tests.For a given number of total con-trollable water heaters,the number of water heaters that should be controlled following each control strategy is calculated. Cohen and Wang[2]present an optimal control scheduling method based on dynamic programming.The advantage of the scheme presented is that the candidate control schedules are not prespeci?ed.The algorithm calculates the ON and OFF cycles for a selected number of device groups.Each group can be con-trolled more than once during the control period.The size and composition of the groups is assumed to be known in advance. During the execution of the algorithm each group of devices is treated separately.The schedule for each group is calculated considering that the load consumption of the rest of the groups is?xed,and the obtained solution is suboptimal.This model can handle both peak reduction and production cost minimization. Kurucz et al.[3]present another optimization method based on LP.The approach is similar to that of Lee and Wilkins[1]but the list of predetermined candidate control schedules is more ?exible.Popovic[4]presents another optimization method based on LP.The main advantage of his approach compared to that of[3]is again the?exibility of the candidate control action list.The controllable devices are pided into two types: devices that can be turned off for a long period(water heaters) and devices that can be turned off for brief periods(air-con-ditioners).It assumes that water heaters can be controlled just once during the control period and that air-conditioners can undergo more than one control action during the control period. The objective is to reduce system peak load.

A similar approach to[2]is presented by Chu et al.[5].They target commercial buildings with air-conditioning and apply dy-namic programming techniques to schedule the OFF/ON cycles of groups of loads.The difference is that their objective is to re-duce peak load while minimizing the amount of load controlled. The reasons for minimizing the amount of load controlled are loss of income and customer satisfaction.

Hoe-Ng and Sheblé[6]present a LP optimization technique that looks not for peak reduction or energy production cost re-duction but rather for the maximization of utility pro?t over the control period.

The contribution of this work to the previous studies lies in the approach developed for modeling the behavior of loads with thermal inertia.For this purpose,the building energy simula-tion tool EnergyPlus[8]is employed.Through the de?nition of a model building for each customer type and its subsequent simulation with EnergyPlus software,the thermal consumption curves of air-conditioning and space-heating systems under the different possible control actions are accurately obtained.In this way,the dynamics of the house are included in the model. The developed methodology has the advantage of calculating the generated“payback,”which represents the extra amount of energy that must be supplied to the devices when the control actions are released in order to bring them back to the initial set-point.This procedure allows the de?nition of control strate-gies based not only on ON/OFF cycles but also on modi?cations of the temperature settings of the thermostats.The proposed al-gorithm is?exible on the selection of the start time and the dura-tion of the control actions as long as they fall within the speci?ed control period.

The paper is pided into?ve main sections.Section II de-scribes the methodology developed for simulating the load con-sumption curves of controllable devices.Section III presents the mathematical formulation of the DLC algorithm.Section IV ap-plies the model to a real power system.Section V describes the general requirements for implementing the proposed model.Fi-nally,Section VI summarizes the main conclusions drawn from the study.

II.C ONTROLLABLE L OAD M ODELING

The?rst step in the DLC algorithm developed is to perform a study of the loads in order to determine their nature and type and,consequently,establish their usual consumption patterns. Then the changes that the different control actions cause in these typical pro?les must be analyzed.In this way,the reduction in demand that can be achieved through the application of each control strategy can be determined.This analysis is performed by the aggregator every time it presents a load reduction bid in the electricity market.

Calculation of these consumption curves is in?uenced by many variables,such as building characteristics(dimensions, construction materials,etc.),local climate conditions(temper-ature,humidity,radiation,etc.),comfort settings,and existing HV AC equipment.For this reason,the building energy simula-tion tool EnergyPlus is employed.

The followed methodology consists of de?ning a model building that represents the average behavior of each customer type and simulating their thermal behavior with the afore-mentioned software.The results of these simulations provide thermal demand curves of controllable devices for a base case, that is,when no control actions are applied.

Taking the model building for the base case as a reference and modifying the schedule of the controllable loads according to the control strategy considered,a new consumption curve that represents the in?uence of such control action is obtained.Con-trol strategies can consist of complete disconnection of the ap-pliance for a set time period or modi?cation of the thermostat settings.It is considered that only one control strategy can be applied to each device during the control interval.However,it can start at any time-step inside it as long as the duration of the control action does not extend beyond the end time of the control interval.This process is repeated for all possible control actions that the aggregator can apply to each customer type.The results represent input data for the algorithm formulated in the following section.

Next,an application of the described methodology to the modeling of a domestic air-conditioner is presented.The model building considered is

a?at sited in a block of apartments.It is represented in EnergyPlus as an orthogonal parallelepiped with

dimensions.It is west oriented and the construction materials ful?ll the actual Spanish regulation in relation to edi?cation[9].The30%of the fa?ade is glazed.Regarding the mode of operation of the air-condi-tioning system it is assumed to be switched on throughout the day,being the temperature setting of the thermostat

23.The EnergyPlus simulations are performed considering that the conditioning equipment is the“Purchased Air Component”

RUIZ et al.:A DIRECT LOAD CONTROL MODEL FOR VIRTUAL POWER PLANT MANAGEMENT

961

Fig.1.Thermal load consumption curve for a domestic air-conditioner

(?at).

Fig.2.Thermal load consumption curves of a residential air-conditioner in the base case and when it is switched off for 60min starting at 14:00.

which is used when the performance of the building with an ideal system is wanted to be studied [8].

Taking into consideration the described model and the fore-casted local climatic conditions relative to the building for the control period (temperature,relative humidity,atmospheric pressure,radiation,and wind speed),the aggregator runs the EnergyPlus simulations in order to estimate the thermal load demand or the domestic air-conditioner.Fig.1shows an ex-ample curve obtained for a particular summer weekday in the north of Spain area.

The previous curve represents the consumption of the air-conditioner for the base case,that is,when no control actions are applied.The next step consists of calculating the load demand curves under all possible control actions.For this purpose,the EnergyPlus simulations are performed again by modifying the scheduling of the controllable loads according to the considered control strategies.Fig.2shows the obtained load consumption for the previous domestic air-conditioner when it is switched off during 60min starting at 14:00.It can be observed the impact of such control action on the temperature inside the building that

reaches almost

27

and the demand peak occurred just after the control period.

III.A LGORITHM F ORMULATION

This section presents the mathematical formulation of the DLC algorithm.The objective of the optimization is to max-imize load reduction over a speci?ed control period through the selection of the optimal control strategies to be applied to a group of controllable devices.It is intended to be run by an aggregator managing a portfolio of controllable customers so as to determine its load reduction capability over a speci?ed con-trol interval.These customers are operated as a VPP taking part in the electricity market by offering load reduction bids to the system operator (SO).

Required inputs can be grouped into dynamic and static parameters.The dynamic parameters are the forecast load demand of the aggregator portfolio,existing customer types that have controllable devices and the number of controllable devices within each customer type;start and end time-steps of the control period and the length of each time-step.In addition,the forecast uncontrolled consumption pattern for each kind of controllable appliance and load consumption curves of controllable devices under the effect of all possible control actions are required.This information is obtained with the methodology presented in Section II.The static parameters include information on control strategies that can potentially be applied to the controllable devices of each customer type and the duration of each such strategy.

The results provide the optimal control strategies,their start times and the number of devices that must be controlled with each of them in order to achieve the maximum possible load reduction over the control period while constraints are ful?lled.

A.Mathematical Formulation

The input parameters and the decision variables involved in the optimization problem are listed below.

Input

parameters:

initial time-step of the control

period;?nal time-step of the control

period;time-step

duration;

forecast load demand of the aggregator at time-step

z;

customer type (e.g.,domestic customers,supermarkets and of?ces).This is an integer that varies from 1to the number of customer

types;

number of controllable devices within the group of customers of type k (e.g.,1900domestic

air-conditioners);

expected consumption at time-step z of the devices of the type k customer when no control actions are applied (base case).This information is obtained with the methodology presented in Section II;

962IEEE TRANSACTIONS ON POWER SYSTEMS,VOL.24,NO.2,MAY

2009

possible control actions.Each strategy has two components:action(e.g.,switch off the device)and duration of the action (e.g.,30min).The strategies can vary for the different customer

types;

expected consumption at time-step z of

the devices of the

type customer when

the

strategy is applied at

time-step.

This information is obtained with the

methodology presented in Section II.

Decision

variables:

number of devices of the

type customer which

are controlled after the optimization with

strategy

starting at time-step

;

number of devices of the

type customer not

controlled after the optimization.The total number

of

variables coincides with the number of

customer types,that is,there

are variables of

this type.

The aim of the optimization problem is to maximize load re-

duction over the control interval or to minimize?nal demand

over that period.The objective function,based on integer LP,

can be written as

follows:

(1)

where is the?nal demand obtained at time-step z after

applying the control strategies calculated with the optimization

algorithm.

The values of the?nal load at each time-step are obtained

with the following

formula:

(2)

where is an input data representing the expected

demand at time-step z

and is the variation in load that

occurs at that time-step when the control actions are applied.

This term can be pided into two parts:1)load variation at time-

step z due to control actions starting

at and2)load variation

at time-step z due to control actions starting

before

.

can be formulated as

follows:

(3)

where

load consumption variation in relation to the base

case of the devices of the

type customer at

time-step when

strategy is applied at time-step.

This is obtained with the following

formula:

(4)

Consequently,represents the variation in

the load consumption of the devices of a

type

customer at

time-step when

strategy is applied

at

time-step

;

number

of variables in which the control action

starts at

time-step.That is,the total number

of

in the problem.

Consequently,is the number

of variables in which the control action starts at

time-step.

The?nal formulation of the optimization problem is obtained

by substituting(2)and(3)in

(1)

(5)

subject to the following constraints:

1)The number of devices controlled cannot be

negative:

(6)

2)The total number of devices connected during the control

period for each customer type is the sum of the number

of devices controlled and the number of devices not con-

trolled:

(7)

where de?nes the total number of control variables

in the problem,that is,the number

of

:

(8)

3)The values obtained

for

and must be integers.

4)Demand after the control period must be below a speci?ed

maximum limit.This limit is optional and is considered as

set by the aggregator in order to ensure that the generated

payback is within acceptable

limits:

(9)

RUIZ et al.:A DIRECT LOAD CONTROL MODEL FOR VIRTUAL POWER PLANT MANAGEMENT

963

Fig.3.System’s and aggregator’s total load demand.

B.Solution

The solution to the algorithm provides the optimal values of

the

variables

that represent the optimal combination of con-trol strategies and the number of devices that should be con-trolled with each of them in order to maximize system load re-duction during the control interval.

The ?nal step consists of calculating the resulting daily load demand curve via (2).

IV .C ASE S TUDY

The DLC algorithm developed was tested on an actual power system in northern Spain.The analysis considers a particular power system area characterized as having a signi?cant number of domestic and commercial buildings connected.It is assumed that an aggregator operates in the region,offering DLC contracts to end-users with controllable devices in order to obtain a signif-icant load reduction capacity that can be offered in the market.The aim of this case study is to test the applicability of the pro-posed DLC algorithm for managing the participation of a VPP in the deviation management market.The study is performed in a summer scenario where air-conditioning system control is considered.The same methodology could be employed for the control of space-heaters in a winter scenario.A.Input Data

It is considered that the TSO calls the deviation management market for a period running from 14:00to 16:00.From the time of noti?cation,market agents have thirty minutes to submit their bids.Consequently,the aggregator runs the algorithm to gen-erate the load reduction bid corresponding for that control in-terval.In the calculation,the time-steps are set as 30min long.The daily load curve of the aggregator is shown in Fig.3.These data correspond to the registered consumption on a par-ticular summer weekday in the considered area resulting from both controllable and noncontrollable loads.

In this graph,the total demand of the transmission system for the considered day is also represented.It can be observed that peak demands of both curves do not occur at the same time.It can also be seen that the control period,which runs from 14:00to 16:00,is posterior to the peak demand of the system.This is because the objective of the developed control system is not the reduction of system peak but to provide a mechanism allowing

TABLE I

N UMBER OF C ONTROLLABLE D

EVICES

TABLE II

B REAKDOWN OF

C ONTROL A

CTIONS

the participation of demand side management into the ancillary services markets.

Three different customer types are taken into consideration in the study:domestic customers,supermarkets and of?ces.Table I shows the number of air-conditioners that the aggregator can control within each group.

The control strategies that the aggregator can potentially apply to the controllable devices,which have been agreed in advance with the end-users,are the same for all customer types and consist of:1)disconnecting the air-conditioner for a maximum of 1h;2)increasing the temperature setting by

a maximum of

2

for a maximum of 2h;3)increasing the temperature setting by a maximum of

3

for a maximum of 1.5h;and 4)increasing the temperature setting by a maximum

of

4

for a maximum of 1h.This information needs to be processed so as to break it down into all control possibilities.For example,the ?rst option would lead to the seven different options that are presented in Table II.

As a result,44different control possibilities for each cus-tomer type are obtained.

Daily consumption curves for controllable devices are provided in Fig. 4.These curves represent the expected uncontrolled consumption of a domestic air-conditioner,a supermarket air-conditioner,and an of?ce air-conditioner,re-spectively (base case).They are obtained via the methodology explained in Section II,i.e.,building modeling and simulation.To convert the thermal demand curves resulting from the En-ergyPlus simulations into electrical ones,it is assumed that the air-conditioning systems considered in the study comply with the U.S.regulation that establishes a seasonal energy ef?ciency ratio (SEER)of 10or higher.

Domestic air-conditioners are assumed to be connected throughout the day,while supermarket and of?ce conditioners are only switched on at the times when the buildings are occupied.

It can be observed that the domestic air-conditioner is discon-nected in the morning between 6:30and 8:30,because at that time the temperature inside the building is below the tempera-ture setting.Peak demand in commercial appliances occurs just

964IEEE TRANSACTIONS ON POWER SYSTEMS,VOL.24,NO.2,MAY

2009

Fig.4.Uncontrolled electrical load consumption curves.(a)Domestic air-conditioner.(b)Supermarket air-conditioner.(c)Of?ce

air-conditioner.

Fig.5.Total and controllable load demand curves.

when the air-conditioning systems are turned on because that is when the biggest input of energy is required to reach the tem-perature setting,which is

23for all customer types.

It is assumed that all controllable devices belonging to the same customer type have the same load consumption patterns.Aggregating the above curves for all customers in the consid-ered region,the controllable fraction of the demand in relation to the total can be obtained (see Fig.5).The resulting hourly average controllable load is 3.7%.

The consumption curves of the controllable devices under the effect of the control actions de?ned above are also inputs to the algorithm.An example curve was provided in Fig.2.This curve represents the load consumption obtained for a do-mestic customer when the air-conditioning system is switched off for 60min starting at 14:00.The load consumption curve corresponding to the base case when no control strategies are applied is also shown,as well as the temperature variation in-side the building.It can be observed that the set

temperature

is maintained throughout the day excepting the period

when the control action is applied.During that interval,the air-conditioning system is disconnected and the temperature in-side the building rises to almost

27.The same calculation is performed by the aggregator for all possible control actions and customer types.

Finally,it is assumed that the aggregator constrains ?nal de-mand to below 96MW,which represents the forecast peak con-sumption for the day.In this way,it limits the payback generated after the control period in order to ensure that it is within accept-able limits.

TABLE III

O PTIMAL C ONTROL A

CTIONS

TABLE IV

L OAD R EDUCTION C APACITY OF THE

VPP

B.Results

Table III shows the results of the implementation of the DLC algorithm.It includes the optimal control actions and their du-rations,the initial time and the number of devices that must be controlled with each action so as to obtain the maximum load reductions over the control interval,which runs from 14:00to 16:00.

By applying the above control actions,the aggregator can at-tain the load reductions presented in Table IV.This table also in-cludes the percentage of forecast demand represented by those variations.

The load reductions achieved are practically constant for all time-steps in the control period,and average 2MW.This rep-resents a drop of 2.2%in relation to the expected demand and a potential energy reduction of 4MWh for the whole control pe-riod.

Fig.6shows a graphic representation of the load reduction achieved.

Fig.7is an enlargement of the part of the graph that runs from 14:00to 18:00.It can be observed that consumption drops for the two hours of the control period,but rises in the subsequent time-steps.This is the payback effect that represents the extra amount of energy that the air-conditioning systems demand in order to restore their temperature settings.The values of the pay-back generated are provided in Table V.

RUIZ et al.:A DIRECT LOAD CONTROL MODEL FOR VIRTUAL POWER PLANT MANAGEMENT

965

Fig.6.Final load demand

curve.Fig.7.Detail of the load variation obtained.

TABLE V P AYBACK V

ALUES

It can be observed from Fig.7that maximum demand oc-curs from 16:00to 16:30,when the value is 96MW.This co-incides exactly with the maximum admissible limit established by the aggregator,so the solution provided by the optimiza-tion algorithm ful?lls the imposed constraint.If the calculation had been performed without taking that constraint into account,the load demand curve provided in Fig.8would have been ob-tained.The difference in the energy reduction is 1.2MWh for the whole control period,with an average power reduction for each time-step of 2.6MW.

Finally,Table VI shows the load reduction bid that the ag-gregator would send to the TSO (constrained scenario).For this purpose,the operation procedures for the Spanish transmission system have been taken as in [12].

The planning periods considered are 1h long,coinciding with the current characteristics of the Spanish market.The bid for-mulated is upward because consumption reduction is similar to generation increase.The bid includes the energy offered for each planning period and the corresponding price.These

prices

Fig.8.Final load demand curve—no constraint.

TABLE VI

L OAD R EDUCTION B

ID

should be set by the aggregator taking into consideration ?nan-cial compensation for customers and its own costs and pro?t.In addition,the complex condition that establishes the inpisi-bility of the bid is de?ned.

V .P RACTICAL I MPLEMENTATION

Implementation of the proposed DLC model would require two different communication stages:1)the communication be-tween the SO and the aggregator,and 2)the communication be-tween the aggregator and the controllable appliances.

The SO performs communication with the aggregator in the same conditions as it does with the remaining agents involved in the electricity market (generators,suppliers).In this way,the number of agents the SO interacts with is minimized.

The aggregator is responsible of the physical communica-tion with the controllable loads in order to ensure that the con-tracted load reduction service is delivered.For this purpose,it makes use of a central controller whose main functions are the following ones:1)maintaining a database that stores informa-tion on the contracted control possibilities of each customer;2)running the optimization algorithm;3)submitting the resulting load reduction bids on the market;and 4)sending the calculated control signals to the controllable appliances.These appliances would require an intelligent thermostat that receives the corre-sponding temperature setting.Such devices are fairly common on today’s market.

To check whether the target load reduction has been achieved smart meters with communication capabilities should prefer-ably be installed at all controllable customers.However,the cost of the metering infrastructure could be reduced.The aggregate load consumption could be accurately estimated using a statis-tically valid number of metered customers [13],[14].

The most suitable communication technology should be se-lected according to the preexisting infrastructure and the costs

966IEEE TRANSACTIONS ON POWER SYSTEMS,VOL.24,NO.2,MAY2009

of installation and operation.Radio signals,Internet,or power line communication systems could be used.

VI.C ONCLUSION AND F UTURE W ORK

In this paper,a DLC algorithm based on LP is developed with the aim of operating a VPP composed of a large number of customers with load reduction capabilities.It is intended as a tool to enable aggregators managing portfolios of controllable customers to provide load reduction bids to the TSO/DSO and consequently participate in electricity markets.The algorithm is able to obtain the maximum load reduction over a considered control period by determining the optimal control strategies to be applied to a group of controllable customers.These control actions act on domestic and commercial appliances which offer some thermal inertia such as air-conditioning and space-heating systems,and can consist of disconnecting the device for a set time or modifying the temperature settings of the thermostats. One advantage of the proposed model lies in the method used to model controllable loads,which are calculated by setting up a building model for each customer type and then simulating it with the building energy simulation tool EnergyPlus.In this way the thermal behavior of air-conditioners and space-heaters under the effect of the different control actions can be accurately obtained and introduced into the algorithm.

The applicability of the optimization algorithm is tested on an actual power system in northern Spain,where the control of air-conditioning systems is considered.Results show that the developed model provides an effective approach for generating load reduction bids of a VPP,thus allowing the participation of a group of controllable customers in transmission system oper-ation markets.

Future work will improve the?exibility of the VPP by removing the inpisibility condition from the load reduction bid.To that end a new optimization algorithm capable of determining the control actions to be applied to the control-lable devices of the VPP in order to obtain the load reduction schedule?nally accepted will be developed.

R EFERENCES

[1]H.Lee and C.L.Wilkins,“A practical approach to appliance load con-

trol analysis:A water heater case study,”IEEE Trans.Power App.Syst.,

vol.PAS-102,pp.1007–1013,Apr.1983.

[2]A.Cohen and C.Wang,“An optimization method for load management

scheduling,”IEEE Trans.Power Syst.,vol.3,no.2,pp.612–618,May

1988.

[3]C.N.Kurucz,D.Brandt,and S.Sim,“A linear programming model for

reducing system peak through customer load control programs,”IEEE

Trans.Power Syst.,vol.11,no.4,pp.1817–1824,Nov.1996.

[4]Z.Popovic,“Determination of optimal direct load control strategy

using linear programming,”in Proc.CIRED,Nice,France,1999.

[5]W.Chu,B.Chen,and C.Fu,“Scheduling of direct load control to min-

imize load reduction for a utility suffering from generation shortage,”

IEEE Trans.Power Syst.,vol.8,no.4,pp.1525–1530,Nov.1993.

[6]K.Ng and G.Sheblé,“Direct load control-A pro?t based load manage-

ment using linear programming,”IEEE Trans.Power Syst.,vol.13,no.

2,pp.688–694,May1998.

[7]I.Cobelo,“Active control of distribution networks,”Ph.D.dissertation,

Dept.Elect.and Electron.Eng.,Manchester Univ.,Manchester,U.K.,

2005.

[8]U.S.Department of Energy,EnergyPlus Energy Simulation Software.

[Online].Available:49ab4228e2bd960590c67749/buildings/energy-

plus.

[9]Building Technical Code,Spain.[Online].Available:http://www.

49ab4228e2bd960590c67749.

[10]J.F.Kreider,Handbook of Heating,Ventilating and Air Conditioning,

J.F.Kreider,Ed.Boca Raton,FL:CRC,2001.

[11]Building Thermal Installation Regulation,Spain.[Online].Available:

http://www.idae.es/revision-rite/documentos.asp.

[12]Operation Procedure O.P.3.3:Resolution of Deviations Between

Generation and Demand.Resolution of24.Of?cial Spanish Gazette

(B.O.E.),May2006.[Online].Available:http://193.148.96.3/opera-

cion/pdf/po/RES_PO_3.6_Tratamiento_indisponibilidades.pdf.

[13]ISO New England,Load Response Manual,2003.

[14]R.Formby et al.,Task XI Final Report,IEA/DSM Initiative,2007. Nerea Ruiz was born in Ordu?a,Spain,in1977.She received the B.Sc.degree in industrial engineering and the M.Sc.degree in electrical engineering from the School of Engineering,University of the Basque Country,Bilbao,Spain,in 1999and2002,respectively.

Since2003,she has worked at the Energy Unit of LABEIN-Tecnalia,Derio, Spain,on European research projects regarding demand response,microgrids, and distributed generation integration.

I?igo Cobelo was born in Bilbao,Spain,in1977.He received the M.Sc.degree in industrial engineering from the University of the Basque Country,Bilbao, in2001and the Ph.D.degree in electrical engineering from the University of Manchester,Manchester,U.K.,in2006.

He joined the Energy Unit of LABEIN-Tecnalia,Derio,Spain,in2002where he is currently a Project Manager.He works in the?elds of demand response, the impact of distributed generation,and active distribution networks.

JoséOyarzabal was born in Bilbao,Spain,in1967.He received the B.Sc.de-gree in industrial engineering and the M.Sc.degree in electric engineering from the School of Engineering,University of the Basque Country,Bilbao,in1989 and1992,respectively.

Since1994,he has been working at LABEIN-Tecnalia,Derio,Spain,on sev-eral applications related to power transmission and distribution control systems as well as protective relays.His specialist?eld includes the application of infor-mation and communication technology standards to energy management sys-tems.Since2006,he has been a Lecturer in the Faculty of Engineering of the University of Deusto,Bilbao.

Mr.Oyarzabal is a member of CIGRE SC6and the Association of Industrial Engineers of Bilbao.

本文来源：https://www.bwwdw.com/article/whgl.html