virtual power plants – general review: structure ......warsaw university of technology, institute...

Open Access Journal

Journal of Power Technologies 92 (3) (2012) 135–149journal homepage:papers.itc.pw.edu.pl

Virtual Power Plants – general review: structure, application andoptimizationI

Łukasz Nikonowicz∗ , Jarosław MilewskiWarsaw University of Technology, Institute of Heat Engineering)

21/25 Nowowiejska Street, 00-665 Warsaw, Poland

Abstract

The article presents information about Virtual Power Plants (VPP). Numerous papers have drawn attention tothe new concept of generation and management of energy. The VPP concept underpins the growing numberof installed Renewable Energy Sources (RES). The concept of smart controlled Distributed Energy Resources(DER) merits consideration. This article is a review of some VPP ideas and gives insight into and a generaloverview of VPP. Some VPP structure and control methods are described, test fields of VPP are presentedand it ends with a short conclusion about VPP.

Keywords: Virtual Power Plant, VPP

1. Introduction

The Virtual Power Plant is quite a new concept.The VPP idea was born a few years ago and has acouple of advantages working in its favor. The mainconcept is based on a centralized control structurewhich connects, controls and visualizes a work ofdistributed generators. Combined heat and powergenerators (CHP), fuel cells (FC), photo voltaics(PV), heat pumps (HP), solar collectors and anyother sources of power and heat might be aggre-gated and cooperate together in the local area. Thisis a good solution for harnessing Renewable EnergySources (RES). At present RES have problems hook-ing up to power networks. This happens because of

IPaper presented at the 10th International Conference onResearch & Development in Power Engineering 2011, Warsaw,Poland

∗Corresponding authorEmail addresses: [email protected]

(Łukasz Nikonowicz∗),[email protected] (Jarosław Milewski)

a lack of transmission capacity in the power networkfor RES. Further complicating matters is the irregu-lar schedule that some RES work to: wind turbinesfor instance are obviously wind-dependent. Thiscauses serious problems to the Transmission SystemOperator (TSO). RES might be suitable for installa-tion on household sites. Capacity of thousands ormillions of watts in such RES might eventually rivalthe capacity which is now installed in wind powerplants etc.

VPP provides an opportunity to lower the load inthe power network. More power is generated lo-cally and is shared by participants without needing totransmit it over long distances at high tension. There-fore one energy loss factor is either minimized oreliminated. VPP causes a sea-change in energy re-lations. The participant is no longer merely a passiveuser. Being a part of VPP means everyone involvedcan influence the power system in an active way, al-though naturally only to a certain extent: it does notmean that participants are responsible for switching

Journal of Power Technologies 92 (3) (2012) 135–149

devices on and off.Heading the VPP is a computer system controlled

by the Distribution System Operator (DSO). Thiscould be organized on the basis of an artificial neu-ral network. In fact, a VPP could be supported byany household which has as little as 1 kW capac-ity in a generator such a PV, FC, CHP etc. In VPPirrespective of how much generation capacity is in-stalled in a single building, the most important fea-ture is connecting all the sources together and run-ning them so as to obtain a state of self-balance inthe most effective way. VPP places more attention onlocal generation, meaning that central generation canoperate in more stable conditions. All peaks of heatand power demands can be more easily optimized byDSOs. Storage of heat or electricity should be usedas well. This will help to achieve appropriate condi-tions of VPP work.

2. Main concept

The term Distributed Energy Resource (DER)comprises Distributed Generation (DG), EnergyStorage and even Electric Vehicles.

If DER can cope with electricity peaks, it is possi-ble to use power capacity to generate additional en-ergy off-peak. This energy can then be sold on theenergy market. DERs can be grouped and managedby a central unit, thereby becoming visible on the en-ergy market. And it is open to any type of generationtechnology.

The main focus points within VPP research havebeen:

• feasibility of DER market participation;

• VPP control and coordination optimization;

• design of VPP and power system.

2.1. Structures of VPPThree different approaches to VPP can be used:

• CCVPP (Centralized Controlled Virtual PowerPlant)—Fig. 1—in this design all control logiclies with the VPP and all knowledge about themarket and the planning of production is sepa-rated from the DER. The advantage of this de-sign is that the VPP is given a simple way ofutilizing the DERs to meet market demand.

Figure 1: The CCVPP Design, after [1]

Figure 2: The DCVPP Design, after [1]

• DCVPP (Distributed Controlled Virtual PowerPlant)—Fig. 2—DCVPP introduces a hierarchi-cal model by defining VPPs on different levels.A local VPP supervises and coordinates a lim-ited number of DERs while delegating certaindecisions upwards to a higher level VPP. Thisdesign can help simplify the responsibilities andcommunication of the individual VPPs.

• FDCVPP (Fully Distributed Controlled VirtualPower Plant)—each DER acts as an indepen-dent and intelligent agent which participates inand reacts to the state of the power systemand market. This design holds much promiseas regards supplying a dynamic and optimizedpower system.

2.2. Examples of VPP usageThe most significant European projects that in

some way use the concept of VPP and integratedDER:

— 136 —


Figure 3: The FDCVPP Design, after [1]

• SmartGrid;

• FenixProject—the goal of the project is to moveaway from traditional management of smallunits in a power system. Thus, the ’fit and for-get’ principle must be rejected. Through theFenixProject all sources will be integrated in anactive way with the system. The new approachshould be used for every kind of DER unit. TheFenixProject tests two types of VPP: commer-cial and technical.

• Ecogrid Project—introduces the concept of theDistributed Energy Market (DEM). The mainpurpose of DEM is to put the end-user at thecenter of the power market and provide the sys-tem operator with the most cost effective solu-tions for system management. An example ofthis is functioning on the island of Bornholm.

2.3. Challenges for VPP implementationMany different approaches have been taken to im-

plementation and VPP control. Each has had to copewith some challenges:

• loose coupling—the production unit ownershould be able to freely choose the VPP withwhich he would like to be grouped. The VPPgroup of members is not static.

• generic adoption—communication between theDER units and VPP operator must be standard-

ized. Only one standard can be allowed and aninformation package must be determined.

• interchangeable strategies—the behavior exhib-ited by a production unit should change depend-ing on the choice made by the unit’s owner orVPP operator.

• security and robustness—the system must beprotected from external dangers and must have aprocedure for operating in the event of lost com-munication.

Proposal solutions:

• loose coupling solution—the ’match maker’module will be introduced. All informationabout the DER unit will be entered in thedatabase of that module. It will be responsi-ble for searching for connections with possibleVPP operators. In the next step, after choosinga VPP, the match maker sets up a connectionbetween a VPP and the unit.

• generic adoption solution—all exchange of datashould be made according to one standard.XML is a proposal for this task.

• interchangeable strategies solution—the systemhas to react in a quick and efficient way so asto dynamically change behavior to fit the situa-tion. It must be possible for the VPP operator,for instance, to change the strategy or logic of aproduction unit.

• security solution—the security standards andspecifications for web services must be defined.In the case of a lost connection between theDER unit and VPP operator, the DER unit con-nects with ’Match Maker’ to gain informationabout new, achievable connection with the op-erator of another VPP. Then the connection isestablished in a dynamic way.

3. VPP review

3.1. European Union 5th Framework Programme

In [2] a project is presented that was realized un-der the EU’s 5th Framework Programme. The aim

— 137 —


was to develop the VPP concept, to implement andtest it and show the results. Under considerationwas whether fuel cells as DER for VPP could be in-stalled at household locations. 31 stand-alone res-idential fuel cell systems were installed. Each unithad 4.6 kWel and 9 kWth. An Energy Manager wasset up to control the whole system. This module wasresponsible for benefits for end-users and grid pur-suer. A Central Control System (CCS) was created tomanage all the fuel cell systems. CCS communicatedwith the on-site Energy Manager and allowed theutilities to control the micro CHPs in terms of peakdemand and defined load profiles. Wireless trans-mission standards, GSM and radio ripple control re-ceivers were used for communication purposes. Theproject was successful. The whole system was sta-ble. There were no emergency cases involving unitsbeing turned off. Fuel efficiencies of up to 90% wereachieved was (with 30% electrical efficiencies).

The low temperature PEM fuel cell system workedfor 138 000 hours. In that time ca. 400 MWh of elec-tricity were generated. More than 50 million mea-surements of data were taken and analyzed. The sys-tem was tested to check how VPP delivers electricitysupplies. The results demonstrated that there was nolatency time in delivery. There are some problemswhich have to be solved before developing this typeof system for the mass market:

• costs must be reduced significantly to increasethe technology’s economic viability;

• the system must be simplified to improve relia-bility;

• the temperature of the heat output must beincreased to become compatible with existingheating systems, and to give opportunities fortri-generation.

The total cost of the project was EUR 8.3 million.The fuel cell system comprised a fuel cell battery,

peak heat boiler, hot water tank and control mod-ule. An Energy Manager controlled the whole sys-tem. The primary aim of the Energy Manager was tosupply heat energy to meet heat demand in buildings.To do this, it communicated with the CHP system inhousehold sites and controlled the fuel cell, boiler

and hydraulic system. The MicroCHP was fed bynatural gas. Heat was consumed on site and the elec-tricity first went to the inverter where DC current wastransformed into AC current and then the electricitywas supplied to the building’s main network.

3.2. EDISON Project

The EDISON Project [3] aims to integrate an elec-tric vehicle fleet with the power system. The inte-gration is realized by VPP. RES cause problems withbalancing in the power system. To this end a compre-hensive solution must be developed. Electric vehi-cles can be treated as energy storage units. The firstissue is to draw up a schedule for charging electricvehicles. This must factor in all boundary constraintsof the system and aim to minimize costs. To solvethis problem a special platform must be developed.It will be coordinated with the power system and en-ergy market to gain all necessary information. RESwill be taken into consideration as well. The com-prehensive system must supply energy to all electricvehicles immediately. All boundary constraints ofthe electric distribution network must be taken intoconsideration. The platform described above willbe located on the island of Bornholm. Every elec-tric vehicle in every location on the island can belinked with the power system via VPP. A simulationwas performed to gauge the influence of a fleet ofelectric vehicles on the local electric distribution net-work. One aspect which differentiates the EDISONProject VPP from other VPP is the common usage ofelectric vehicles as active energy storage units. MostVPPs concentrate only on intelligent management ofgeneration units. Two possible ways of implemen-tation are considered. In the first one, VPP will beintegrated as part of the power system. In the sec-ond one, VPP will be a new system that cooperateswith the existing power system. In the second ap-proach VPP will be a new subject on market. If VPPis introduced with the power system as is being con-sidered, then it will stay part of the power company.VPP will provide balancing tasks as Balancing Re-sponsible Party (BPR). VPP might be a perfect toolto smooth the boundary between demand and supply.Standalone VPP architecture is an alternative to theabove. There VPP is BRP too. But it is indepen-dent and works as every other member of the market.

— 138 —


It buys and sells energy based on the collected dataand state of each generation unit. The most impor-tant task is to create balancing schedules. VPP ofEDISON Project contains 3 main modules:

• control module for each DER;

• data collecting module;

• connection, cooperation and communicationmodule.

Each of the above modules contains other modules.For standalone VPP architecture the whole system ismore complicated than if VPP is integrated in the ex-isting power system structure as part of a power com-pany. Bornholm was chosen as test field for VPP.52 DER units are located around the island and 35of them are wind turbines. It is good place to testsuch a system operating in island mode. 27 000 con-sumers of electricity are on the island. Total capac-ity is 135 MW and maximum load is 55 MW. TheEDISON Project checks how electric vehicles cancope with wind farm generation. The potential ex-ists to have active management of electric cars with-out any disruption to car owners. A simulation wasperformed based on a model of the power networkof the island. This might be done using commercialsoftware of Matlab/Simulin/Powersin or DigSilent’sPowerFactory, which simulate and analyze transmis-sion or distribution networks, power sources andconsumers. The model can be used for data manage-ment, prediction and optimization of the operation ofthe whole system. With the present model it is pos-sible to simulate energy flow in a power network de-pendent on electric car movement. All calculationsare made with 15-minute intervals. The calculationsare used to make an energy flow map. This can beused to determine where the network and transform-ers are overloaded. Generated energy and consump-tion are balanced by regulating wind turbines andpower plants. Further work will be dedicated to twoproblems:

• prediction—electricity demands must be pre-dicted in order to create a schedule of powersource generation. Charging electric cars is an-other issue. Wind conditions must be predictedusing weather forecasts and historic data. All

these tasks require solutions and better modelsfor improved results.

• optimization—various objective functions canbe optimized, i.e. costs, power balancing inthe case of intermittent operation of RES unitsand power supply for electric vehicles. Opti-mization must be developed to achieve a betterglobal optimum based on local optimums.

3.3. Konwers 2010

Basic requirements and experience with regards toVPP are set out in paper [4]. Increasing reliance onRES causes more problems in terms of balancing.RES depend strong on weather conditions. CHP op-erations too are driven based on heat demands whichare determined largely by weather conditions. Liq-uidating the reserve power of central power sourcesis justified from an economic point of view. This re-serve is used to compensate for a lack of power inthe power network resulting from unpredictable RESgeneration. It is more sensible to transfer a balanc-ing task to a different level of structure. This struc-ture should contain different types of DERs, energystorage units and demand control facilities. It can allbe clustered into VPP structure which can performas system power plant. The operations of each unitcan be scheduled in advance. A distributed energymanagement system (DEMS) supervises the wholesystem, taking into consideration all boundary con-ditions.

All these tasks can be performed because of in-novative data transfer methods, communication andremote control which together can monitor a largenumber of distributed energy sources. VPP con-trols all energy flows in the system and factors in theweather forecast.

Modern power systems are based on the central-ized generation of electricity and/or heat energy.However, global trends are heading toward increas-ing numbers of distributed generators. This meansthe management process has to adapt to the presenceof distributed generators and their unique method ofoperation. The power system has to cope with un-predictable conditions of operating DG. Therefore,an innovative approach to management of the powersystem is demanded. The rise in and penetration of

— 139 —


RES is a challenge. The new approach to manage-ment must be cost effective, economic and providea stable operating system. The balancing process insome areas can be taken over by VPP. VPP will op-erate based on schedules made in offline mode. VPPwill supervise the schedule realization of each DERin online mode. VPP can be integrated vertically orhorizontally. One VPP system can be a part of an-other, bigger VPP system. It is possible to connectmany VPPs to the existing power system. As canbe seen, VPP architecture is a very flexible structure.This is one of VPP’s biggest advantages. The basicfunctionality of a VPP is provided by DEMS. TheDEMS system performs generation, storage and loadmanagement. The main goal of the DEMS system isto achieve a win-win situation in the power system,meaning that it will benefit both the power systemand the customers.

3.3.1. VPP—operating descriptionRenewable generation and electrical and thermal

demand within the supply area is forecasted foreach 15-minute billing period by offline modules ofDEMS. Based on this, the operating schedule foreach DER unit is made for each 15-minute period.All schedules are made 1–3 days in advance. Onlyunits with a certain share on the maximum powerof the VPP are considered. Small units and non-controllable units are only forecasted. It should beborne in mind that VPP can be optimized in severalways. The control of scheduled operation is made inonline mode. Unplanned power fluctuations and de-viations from the schedule require rapid adjustmentof the real power flow within the individual periodby dispatching controllable generation, storage unitsand demand in a one-minute time interval.

Additional reserve strategies must be provided tocope with unavoidable prediction errors. It will coverthe reserve power locally with all technical con-straints. The system must stay very simple, compat-ible and complete. An emergency situation shouldnot eliminate the operation of the whole VPP. In pa-per [4] is described Konwers 2010 project. VPP in-cludes CHP fired with biomass, wind turbines, so-lar plant and conventional power plants. All haveto supply households, industry, hotels and offices.The main part of the electricity and heat demand

is covered by DERs of VPP. If electricity produc-tion is insufficient, then extra electricity will be de-livered to the VPP from external sources. Connec-tion to the external power network is necessary. TheKalman filter was used to supply predictions of elec-tricity and heat demand. The algorithm uses histor-ical data—solar exposure, temperature and calendardata as well. Mean absolute deviation varies between6–14% depending on the type of load. It must benoted that this filter is very sensitive to changes insystem structure. New units cause increasing error.Prediction of RES generation is performed on the ba-sis of weather forecasts, but its accuracy is very lowwith a mean absolute deviation of 40%.

3.4. FENIX Project European project FENIX—Northern and Southern Scenarios

In Europe VPP is considered as a new ap-proach to meeting power demand. Two concepts ofVPP—Technical VPP (TVPP) and Commercial VPP(CVPP) are developed in the project. TVPP is a con-cept of aggregated generators which are located inthe same geographic area. DSO is given the real-time local demands of capacity. Those demands canbe covered by DER. Moreover, the cost and operat-ing characteristics of each generator are given too.In other words, TVPP is a local power managementsystem which gives detailed information about all as-pects of the local system. CVPP has functions whichcontain information about the costs and character-istics of distributed power sources. CVPP does notdeal with the technical delivery of loads. It is a sys-tem which enables trading in the energy market andthe balancing of trading. TVPP and CVPP do nothave to be the same system. TVPP can contain morethan one CVPP in FENIX. VPP were implemented intwo networks. The first of them was the real powernetwork of Iberdrola in Spain (Southern Scenario)and the second was the EDF Energy network in theUK (Northern Scenario).

3.4.1. Northern ScenarioThe Northern Scenario concentrates on the usage

of CVPP. It is dedicated to small scale generation: inhouseholds and municipal facilities (i.e. civic cen-ters, conference centers etc.). The main parts of thedevices are CHP and PV, connected to a low voltage

— 140 —


network. But medium scale devices take part in thetest too. The aim of the VPP tests was to prove thatit is possible to use the VPP concept in the present-day network. The idea was to check if loads can becontrolled for short periods of time. The VPP systemcontains a couple of different structures to fulfill dif-ferent tasks. The first are generators. Without themit is impossible to deliver the VPP concept. Nextare control boxes. They provide the data-gathering.All generators in the structure tree are visible in thecontrol box system. The system gathers informationabout the actual generation level, demands and con-trol flexibility. The data are sent to DEMS hostedby DSO and then to CVPP. The solution is preparedin response to incoming information and is then sentto generators as returned data. The answer containsinformation about the best generation level from theaggregated devices in the context of the actual situ-ation on the energy market. Consideration has beengiven to involving the DSO as an active participant ofthe network in the future development of VPP. TheDSO will have more rights to control and to act in allVPP processes. The main advantage of the system athand is the near real-time visibility of generation anddemand in the area where VPP is installed. It shouldnot be forgotten that another important value is thevisibility of flexibility as regards all elements whichare aggregated in the main layer of VPP—from gen-erators to the distributed network.

3.4.2. Southern ScenarioIn contrast to the Northern Scenario, the Southern

Scenario focuses on generators which are connectedto a medium voltage network. They might serve asan ancillary service to DSOs and TSOs. In the net-work at hand DER 12 capacity of about 170 MVAis installed, which determines about 35% of all ca-pacity linked to this medium voltage network. TheVPP works as a parallel control system and is op-erated by DSO. There is no interference in real net-work movement. All information about the networkis downloaded in real-time from a SCADA system.In this system no difference exists between DEMSand CVPP. DEMS is the same as CVPP. Data are ex-changed between CVPP module and control boxesby the GPRS network. VPP was used to show itsusefulness in: Voltage Control—support for main-

taining a determined voltage level by providing reac-tive power to the network; Network Contingencies—generators available to work just in case; TertiaryReserve—power reserves that can be put into the net-work within 15 minutes and help to cope with imbal-ances; Participation in the Day Ahead Energy Mar-ket. As previously, distributed generators can be usedas part of the energy market. Each unit reports theirstate of availability and current level of work andcreates a bid. All information is collected and sentto CVPP. CVPP processes the received data, makesone common bid and submits it to the energy market.After the trading session the returned information isdelivered to CVPP, which divides it into single bidsand informs the DSOs. The DSOs receive informa-tion about the assigned capacity output and a workschedule for each generator. Then every DSO hasto make a decision about accepting or rejecting eachbid in terms of technical feasibility. The results ofthis validation from all DSOs are sent back to CVPPsystem which hands it over to TSO. The TSO has fullinformation about the work schedule of each gener-ator. As a result, aside from the traditional balanc-ing procedure, a new participant of the transmissionnetwork has been created. This manner of cooper-ation between DSOs, TSO and other participants ofenergy markets is well known in energy systems inmany countries. The difference is the possibility toearn money from the hundreds and thousand of unitswhich operate in the VPP structure. It benefits fromthe rule that the bigger you are, the more you cando. The Southern Scenario delivers experience inthis field. It shows that the energy market is a placewhere VPP can be used in a commercial way.

3.5. ’Smart’ Heat Pumps

The increasing capacity of wind farms causesproblems with balancing generation and demand.There is self-evidently little ability to affect gener-ation. On the other hand, demand can be used asa factor to cope with problems of surplus capacityin the network. Demand might be shaped in an ap-propriate way to meet available generation. This isthe main idea behind using heat pumps in the en-ergy system. ’Smart’ heat pumps can help to balancegeneration and demand and to manage network con-gestion. The role of heat pumps in the distribution

— 141 —


network might be enhanced through active partici-pation in the energy market. The smart heat pumpsidea is being considered in a few countries, i.e. Ger-many, Switzerland and Denmark. A description ofthe smart program is set out below.

Germany. Vattenfall Europe launched VPP inBerlin. VPP controls the operation of CHPs and heatpumps which are aggregated in VPP. In Germany thewhole concept is based on energy prices in the en-ergy market. Currently, the whole system containsonly 30 heat pumps and CHP in total—20 HP (withheat capacity of <25 kWth each) and 10 CHP. Thecapacity is 30 MW and the goal is to reach 500 MWby the end of 2011. The system is a response to in-creasing unpredictability due to more and more windgeneration capacity. This causes sudden energy pricefluctuations on the energy market. This can be re-duced by using heat pumps. When the energy priceis low, the heat pumps are turned on. When the priceis high then CHPs are turned on and generate surpluspower. The whole process is controlled by a Vatten-fall control center. The VPP structure in Germanyis simple. There are units which generate power orheat. Their operating schedule is mainly dependenton the profile of heat and power demands of cus-tomers. These units have a communication modulewhich sends information about demand. All unitsare aggregated by VPP. VPP gathers all informationflows from units and processes them. Everything ispassed to the control center to ensure quality ser-vice and control. After correlation with energy pricestaken from the day-ahead market, the control centergenerates a dispatch schedule for each HP and CHPfor the following day. The report is then sent to theunits. There is two-way communication between theunits and control center. The aim is to achieve an op-timized system which meets all demand and is ableto generate extra income while saving energy.

The Netherlands. The usage of household mi-croCHPs was described in [5]. Those microCHPscan be used as sources of electricity and heat for lo-cal demand. 10 units were clustered in VPP and theperformance test was conducted. ECN and Gasuniecoordinated this test field. It was found that the op-portunity exists to reduce substation peak load by30–50%. Stirling engines were used as microCHP

units, each unit having 1 kW electric. During testsnot all of the units were connected to the same low-voltage network, but they all had a common sub-station. ECN developed the PowerMatcher module,which is responsible for coordinating supply and de-mand in the electricity network. A special softwareand communication module was installed in each mi-croCHP unit. The software controlled a work of uniton the basis of a number of important economic fac-tors. In such a case there was no need to implement acentral optimization algorithm. Bids took the form ofinformation exchanged between units and the Power-Matcher operator—bids as a signal for the electronicmarket. The bids consisted of information about theprice which is acceptable for the unit’s owner in thesense of buying or selling electricity (and capacity).In reply from the market, price signals were deliv-ered. An autonomous decision about operating ornot operating and waiting for the next round of bid-ding was taken on the basis of those signals. A Pow-erMatcher module can be connected to DER or otherPowerMatcher modules. This is possible due to stan-dardization of the transmitted information package.Various units can be clustered onto different levelsand there is no problem with connecting units toPowerMatcher modules. In other words, one Pow-erMatcher can coordinate several units and can beconnected to another PowerMatcher module—on ahigher level—which in turn coordinates the opera-tion of other units etc. The unit’s owner could de-termine the main goal of operating the unit. Bidsare then sent to the PowerMatcher module. The bidscontain the aim preferred by the owner.

Field tests were made in 3 cases:

1. the load was assumed to be the standard pro-file of energy usage in households in theNetherlands—no micro CHP;

2. the same energy consumption as in 1, mi-croCHP units were installed and operated onlyin a heat-demand driven manner;

3. as in 2, but VPP was launched and microCHPfunctioned in a new environment of collabora-tion between VPP and the market.

The field tests were conducted in May 2007. Therewas only heat demand for hot water. All house-holds had a 120 liter tank for hot water. The paper

— 142 —


Figure 4: Field test result of clustered control of 10 microCHPsafter [5]

showed results for one day during which only 5 mi-croCHP units were participating without any prob-lems and disruption. Figure 4 shows total electric-ity demand during one day. As can be seen, therewere 4 peaks. PowerMatcher reacted to those peaks,shifting additional power production. The field testshowed that 5 microCHPs mitigated the maximumvalue of peaks. But the boundary factor that re-sulted in a poorer quality outcome was the lack ofsufficient heat storage space: units could not producemore electricity without producing additional wasteheat. The hot water tanks were designed to cope withthe waste heat problem, but they had limited capac-ity. And in a situation where peak loads appearedin very short intervals there was insufficient free ca-pacity to take over additional heat production. Hotwater tanks were filled during the first peak. The pe-riod of time between the first and second peak wasvery short, so there was no time to use the heat storedin tanks. As a consequence, the engines could notoperate. All tests collected information about VPPpotential. They clearly demonstrate that the DER in-telligent management system produces benefits. Thecurrent way of power system management—fit andforget—delivers more system load. The suitable op-erating of DER units can result in lowering and mit-igating the demand curve in the system.

The authors claim there is no problem with group-ing together different types of power sources. Thecase presented above featured Stirling engines only,

but implementation with other, different units doesnot pose a problem and should not cause any errorsin the VPP system.

4. Methods of VPP control

4.1. Ohmic resistance control method

The Virtual Power Plant can be considered as acluster of distributed electricity generators (DEG).They are attached to the telecommunication net-work. In the paper [6] the possibility of VPP operat-ing without additional costs of transmission are de-scribed. MicroCHPs which generate electricity andheat are considered. They have a high total effi-ciency. The typical heating system for a householdcomprises a CHP integrated in the central heatingsystem and an additional burner which is used dur-ing winter peak time, plus a hot water tank. Theelectricity generated can be used on site or trans-ferred by a low voltage network to the power system.When there is insufficient generation of electricityand unbalancing occurs, then external electricity issupported by a low voltage network. Two modes ofoperating are defined - heat driven and power driven.Hot water storage gives extra capacity for storingheat, which can be used to improve the operations ofthe whole system, and for switching between oper-ating modes in light of the heat storage status. Eachkind of system has specific parameters. The ohmicresistance of a low voltage network is higher thanthat present in a mid or high voltage network. Thisfact can be used to define areas where the power loadis the greatest. Such areas are characterized by highvoltage drops. Thus, the signal can determine theload of each CHP unit in the VPP system. This re-lates to units which operate in power driven mode.The idea is to increase the generation of CHP unitsin cases where the voltage in the feed-in nodes dropsbelow certain levels and vice versa.

4.2. Marginal costs

Supplying power demand to balance the powersystem in real time is one of the key advantages ofVPP. The reserve power is at TSO’s and DSO’s dis-posal. VPP manages all units in the most cost ef-fective way and minimizes total balancing costs. To

— 143 —


operate correctly some input factor must be deliv-ered to VPP. The impulse can take the form of themarginal electricity cost of the individual DER units.Marginal electricity costs are highly dependent onthe local context and change over time. For exam-ple, CHP generates electricity depending on heat de-mands. The more heat is needed, the more electricityis generated and vice versa. VPP is a structure whichcontains a number of different types of DER. For thetypical situation the power of each unit is not veryhigh and so the DER units are small. The productionof energy is a dynamic process, so the marginal en-ergy costs are dynamic too. In article [6], a methodis presented to determine DER marginal costs andbenefits achieved through using bid strategies. DERunits have to send information about marginal coststo VPP. This information is sent in the form of bid-ding formulas or demand curves which determinethe needs of VPP for electricity at established prices.Negative values in the curve diagram means the DERunit is able to produce extra power at set price level.Bid offers are made by an agent using special soft-ware which is able to create a complex bid schedulefor a particular moment. These offers are made basedon:

• current operating state of the DER unit;

• economic parameters such as marginal operat-ing costs;

• market environment with all market mecha-nisms;

The author of [7] sets out 3 different strategies whichmight be used in the bidding process:

• fully marginal-cost based strategy—this canbe used in a situation where the VPP sys-tem contains only distributed electricity gener-ators. It considers fuel prices, unit efficiency,running-history dependent maintenance costsand startup costs.

• fully price history based strategy—this can beused when storage devices are used, e.g. ac-cumulators. It considers minimum and maxi-mum prices from previous time periods and thelevel of available storage space. Storage devices

Figure 5: Structure of the prognosis based optimization algo-rithm for cogeneration system, after [8]

are controlled by price signals—if the energyprice is low then the storage device runs andvice versa.

• mixed strategy—this can be used for CHP unitswith additional heat sources and heat storageunits. It is hybrid of the above two methods.In paper [6] a bidding strategy is presented fordifferent types of DER.

4.3. Optimization of CHPThe algorithm for CHP optimization is described

in [8]. The aim of the algorithm is to set all devicesin such a way as to maximize the benefits from sys-tem operation. Everything is based on the variationof electricity prices over time. The operation sched-ule of units must be known one day before physicalsupplies. This schedule has to be delivered to EEX.Underpinning the algorithm are the predicted elec-tricity prices on the spot market and the heat demandforecast. Heat can be produced by boilers, CHP ora heat storage tank. The heat storage tank gives anopportunity to separate the production of electric-ity from heat. A CHP can operate even if there isno demand for heat from the user. The main needis to maintain heat supply to a customer at an ade-quate level. The CHP can operate in such a way asto generate surplus electricity which can be sold onthe market. The algorithm maximizes benefits fromsales of surplus electricity and handles heat supplies.At the same time there are boundary conditions onthe algorithm objective function. The boundary con-ditions come from the technical restriction of oper-ating CHP units. To achieve the best fitted result ofoptimization, prediction data must be delivered. The

— 144 —


authors suggest using statistical methods with em-pirical functions to fit loads (prognosis with multiplelinear regression). Physical model based functionscan be used. In the opinion of the authors, neuralnetworks are not well-suited to make forecasts. Inthe method employed the forecast is done using adatabase with a time horizon of at least 1 year. Thedata used are: weather conditions, calendar data andthe commercial weather forecast. As a result the heatdemand curve is obtained for a specific building. Theresults correlate reasonably well with real data andvalues. The idea of optimization is to find the max-imum or minimum value of some objective functionin some calculation area. The authors used the MILPmethod to optimize the objective function (Mixed In-teger Linear Programming). This method can findthe optimum in a model which is described by linearfunction as well as integer variables. Optimizationcalculations for a local heat system are formulatedin the solver-independent language AMPL and aresolved by CPLEX optimization software. A coupleof formulas which are the mathematical models ofdifferent elements of the energy system are presentedin the paper. There are formulas for the heat system,CHP, boiler and heat storage. The objective functionis determined as maximum benefit from total heatand electricity sales considering the costs of energyproduction in each CHP and boiler unit. The max-imization of benefits can be contrary to other goals,e.g. to minimization of primary fuel energy. Themodel based on MILP was used to optimize the op-eration of an existing heat system which is suppliedby [3] the same CHP units. The forecast was doneusing data from the EXX spot market of 2004. Thesimulation was performed and compared with realvalues. 2 power diagrams were obtained—one withthe simulation and other with real operation. The to-tal analysis for the whole year showed that there wasa cost savings potential of 10%. Moreover, the ap-propriate usage of heat storage reduced the energylosses to a small extent. As is shown, great potentialfor optimal management still exists.

4.4. Reduction of generation costs

The emergence of new sorts of power sources suchas RES causes new problems. Due to their connec-tion to the low or mid voltage network, those prob-

lems are located exactly on the low or mid voltagenetwork level. They manifest themselves in: changein energy flow direction, overload of network, prob-lems with frequency and balancing. The control sys-tem of VPP and optimal structure of VPP are de-scribed in paper [9]. The authors highlight need for acentral Energy Management System (EMS) module.Each VPP has to be connected directly or indirectlywith EMS so as to enable the exchange of importantdata. The EMS controls the entire exchange proce-dure in real time. The data contain information aboutthe current situation and state of each market partic-ipant. The appropriate communication must be pro-vided to transfer data. The VPP system should belocal in nature, react quickly with minimal latencytime and have the ability to connect new units. Thenetwork hierarchy structure should adjust accordingto the number of users. Too many points connectingwith the EMS at any one time may result in the wholesystem slowing down and thus overloading. The au-thors point out problems with energy flow. In theconsidered VPP more attention must be put on con-trol of the energy flow in the network. Thus, the mea-surement of energy flow has to change. There is atpresent only measurement on the power sources side.The new concept assumes measurement will takeplace in real time on the demand side too. Measure-ment devices will have to be installed and a specialprotocol for data transmission implemented too. Theauthors suggest using a measurement system coupledwith GPS technology. Transmission will be possi-ble by satellites. The operator demands full possibleknowledge about the network to determine what hap-pens in individual lines of the network. Several ser-vice tasks must be performed to avoid network faults;usually the TSO will be responsible for this. Throughthis activity the system can operate without distur-bance. The role of the DSO is very limited. This sit-uation will change due to VPP presence. The DSOwill cooperate with VPP and more service tasks willbe transferred to the lower level—that of the DSO.The power system operation schedule will be doneby DSOs. They will deliver information about thereserve power of the DER for the following day.

— 145 —


Figure 6: Structure of the VPP, after [9]

4.4.1. Optimization approach

The objective function of EMS is determined asminimum: energy generation costs, total costs andtransmission losses. The boundary for the objectivefunction is network overload. Measurement technol-ogy must be improved, as mentioned above. Theconsidered case involved a VPP system with 3 en-ergy suppliers and 4 different consumers. A desalina-tion plant and electrolyzer for hydrogen productionwere implemented as consumers. It clearly demon-strates that VPP also has great potential for operatingwith different kind of units. Nowhere else in the liter-ature was a solution of this type described. EMS liesat the head of the whole system and directs all energyflows to fulfill the needs of the consumers and to op-timize use of the generated energy. The entire modelwas tested using data from the German energy sys-tem. A simulation of the VPP system was performed.It transpired that there is a specific capacity of windfarm for which total operation costs of VPP are min-imal. For the data used, this capacity was 125 MWlevel. More information about structure and test re-sults are described in [9].

Figure 7: Flow chart of the optimization control, after [9]

5. Decision-support software for VPP manage-ment

The decision-support tool for Virtual Power Pro-ducers is presented in article [10]. A tool usage anal-ysis is made. The authors highlight that long dis-tance energy transmission is unjustified in terms ofDER production. All consumption of the producedenergy takes place at or near to the site of produc-tion. The DER units are connected with the distribu-tion network. Due to the increasing number of RESthere will be a marked increase in energy productionfrom such decentralized sources in the near future.Thus, a solution needs to be designed that will beable to make effective use of that renewable energy.To achieve that aim, broad-based knowledge must begathered on all kinds of DER units. The main ideais to develop a system which provides an opportu-nity to coordinate all units in one common systemwithout sharing DER units. All units must be able tocooperate together to achieve the common aim. Therequired knowledge concerns the particular operat-ing features of units such as: technology ripeness,profitability, availability, reliability, production and

— 146 —


capture of greenhouse gases, relation with externalfactors and lifetime.

The schematic diagram of VPP is presented in thepaper. It consists of 5 layers which are closely inter-connected. They are: meteorology, power producer,VPP, Market and System operator. Fig. 8 shows thechallenges facing VPP and the operations requiredto achieve results. The schematic gives a generaloverview on VPP foundations.

The name of the tool is ViProd. It is responsiblefor VPP simulation and its interaction with market.This tool uses the characteristics listed above andhelps with the decision-making process. It has beensplit into 2 parts. The first part calculates the en-ergy production for the one day ahead and the secondone simulates generations. It is possible to simulateenergy production from wind turbine, small hydro-electric and photovoltaic plant. Real external data isneeded to make correct calculations, e.g. wind direc-tion and speed, water flow rate, temperature or solarradiation coefficient.

The tool simulates the energy production possibil-ities for each type of DER unit, taking into consider-ation the characteristics of local demand, peaks andtime variation of factors which can influence produc-tion. Prediction about factors gives approximate realconditions and reduces forecasting error. The arti-cle presents only the part of the tool which forecastswind condition based on various input data and thepart of the tool which simulates the operation of windturbines. The simulator uses results which are de-rived by the first part of the tool, and informationabout possible energy generation over a defined pe-riod of time is delivered as a result. The applicationis responsible for running all clustered units in onespecific VPP structure. Power is distributed properlyamong a number of DER units in function of sold en-ergy, production cost and available power. A check-ing procedure is performed regarding the predictedamount of energy and power. Next a report is createdabout the reserve power of each unit which is at theoperator’s disposal. A software test was conducted.Ten units were implemented as the VPP system. Thetest was performed in order to determine whether thereserve power is sufficient to cope with unbalancingin the power network. Unbalancing might be causedby, for instance, turning the wind farm off. Two tests

were conducted. The first one showed that VPP wasable to cope with the problem of a sudden turningoff of the wind farm. The second test simulated thesituation where the balancing from the VPP side wasnot at an appropriate level, so some unbalancing oc-curred in the distribution network.

Matlab/Simuling was used to make ViProd.The following conclusions were presented by the

authors. The appropriate designed VPP structure caneliminate:

• uncertainty of energy production forecasts;

• fines for unbalancing;

• the lack of small energy producers on themarket—there is space for small players;

• problems with the CO2 market;

• high management costs.

The following step will be to connect ViProd to theenergy market simulator—MASCEM.

6. Conclusions

The number of small DER units will increase overtime. Currently, the world is powered by large powerplants which supply power systems. But this situa-tion will change due to RES among other things. Thecentral power sources will be turned into decentral-ized sources. This trend is observable in Denmark,where the number of small CHP units has increasedsignificantly in recent years. To adapt to change, asuitable management system will be needed. VPPis a concept which is able to cope with managing alarge number of different types of DER units. Thebiggest advantage of VPP is its modular structure. Itcan be connected to power systems and comprise anumber of DER units. Depending on requirements,extra modules can be added, in order to optimizethe system, secure transmission and/or report results.VPP has the flexibility of building blocks.

As is reiterated in the paper, every VPP system dif-fers in the detail. There are many approaches to theVPP concept. Every scientist and engineer involvedhas her/his own vision of VPP. But the core idea staysthe same—gather together and manage as many DERunits as possible to achieve better, more cost effectiveand environment friendly results.

— 147 —


Figure 8: Schematic representation of the functioning of VPP, after [10]

— 148 —


References

[1] M. Braun, Virtual power plants in real applications.[2] K. Kok, Short-term economics of virtual power plants, in:

20th International Conference on Elecricity Distribution,2009.

[3] P. B. Andersen, B. Poulsen, M. Decker, C. Traeholt,J. Oestergaard, Evaluation of a generic virtual powerplant framework using service oriented architecture, in:2nd IEEE International Conference on Power and Energy,2008.

[4] G. Kaestle, Virtual power plants as real chp-clusters: anew approach to coordinate the feeding in the low voltagegrid., in: 2nd International Conference on Integration ofRenewable and Distributed Energy Resources, 2006.

[5] A. Dauensteiner, European virtual fuel cell power plant -management summary report.

[6] H. Morais, M. Cardoso, L. Castanheira, Z. Vale, I. Praca,A decision-support simulation tool for virtual power pro-ducers, in: International Conference on Future PowerSystems, 2005.

[7] B. Roossien, M. Hommelberg, C. Warmer, K. Kok, J.-W.Turkstra, Virtual power plant field experiment using 10micro-chp units at consumer premises, in: CIRED Semi-nar 2008: SmartGrids for Distribution, 2008.

[8] C. Binding, D. Gantenbein, B. Jansen, O. Sunstroem, P. B.Andersen, F. Marra, B. Poulsen, C. Traeholt, Electric ve-hicle fleet integration in the danish edison project - a vir-tual power plant on the island of bornholm (2010).

[9] L. Sugden, ’smart’ heat pumps: Enablers of low carbonfuture, A Delta Whitepaper (2011).

[10] P. Lombardi, M. Powalko, K. Rudion, Optimal operationof a virtual power plant, in: Power and Energy SocietyGeneral Meeting, 2009.

— 149 —

virtual power plants – general review: structure ......warsaw university of technology, institute...

Documents