Introducing Electric Vehicles in the Microgrids Concept - IEEE Xplore

2 downloads 0 Views 1MB Size Report
Abstract- Electric Vehicles are anticipated to have a considerable percentage of the vehicle market sales in the forthcoming years. The additional energy ...
Introducing Electric Vehicles in the Microgrids Concept

Evangelos L. Karfopoulos, Student Member, Panagiotis Papadopoulos, Member, IEEE, Spyros Skarvelis-Kazakos, Member, IEEE,

IEEE, Ifiaki Grau, Liana M. Cipcigan, Member, IEEE, Nikos

Hatziargyriou, Fellow, IEEE and Nick Jenkins, Fellow, IEEE

Abstract-

Electric

Vehicles

considerable percentage of the forthcoming

years.

The

are

anticipated

vehicle

additional

to

have

a

market sales in the

energy

requirements

for

charging their batteries may affect the network operation, in terms of stability and reliability, especially when these are synchronized with the system peak demand. This issue will present power system operators with the challenge to efficiently integrate Electric Vehicles into power systems by exploiting their ability to behave as manageable loads. The Multi-Agent System technology

has

been

proven

a

promising

way

to

manage

distributed energy resources in electrical networks. A multi-agent system approach to manage the charging of the Electric Vehicle batteries is described. The types of agents required to satisfy the technical operation and the market participation of Microgrids with Electric Vehicles are provided.

Index

Terms-Electric

Vehicles,

Energy

Management,

Intelligent Agent, Mobile Agents, Multiagent systems, Smart Grids.

I.

C

INTRODUCTION

oncerns with global warming, oil shortages and increasing gas prices indicate the change in consumer preferences and automotive industry direction. Electric Vehicle (EV) presence in the transportation section is anticipated to increase in the forthcoming years. The private initiatives for developing the required charging infrastructure that will enable the integration of EV into various levels of distribution networks are considerable. The experience gained from various European and national projects aiming to evaluate different charging facilities through field tests can contribute to the development of a European standard for charging infrastructure that will enable the mass implementation of EV charging places. However, a mass deployment of EV into power systems raises a challenging issue for the network operators concerning the adequacy of the grid capacity. The energy requirements that fulfill EV needs are highly dependent on the traffic patterns of EV owners [1] . The charging demand of an EV fleet may be synchronized in certain cases as for example when EV owners plug their EV as soon as they return home from their last daily travel. This results in a high peak demand which can strike the network operation especially when this EV demand is synchronized with the system peak demand. In such cases, if the charging procedure of EV batteries is left uncontrolled, it is anticipated that grid capacity may not be sufficient [2] and distribution networks may suffer from severe feeder voltage excursions, and equipment overloads [3][4]. Thus, it is necessary to manage the synchronized EV

charging demand in a way that EVs can be treated as manageable loads which are able to provide elasticity to the network operator. To achieve this, some management arrangement should be implemented. The efficient management of EV charging demand can prevent the need of costly premature infrastructure updates. The aim of this paper is to present a management structure that enables the efficient management of the loading needs of an EV fleet ensuring that no grid technical violation will occur. Moreover, this management structure enables the participation of an EV fleet in the electricity market enabling EV owners to exploit the market opportunities. The presented EV management structure is based on the Multi-Agent System (MAS) technology. MAS technology is the evolution of distributed control and has been successfully implemented within the Microgrids and More-Microgrids projects [5]. This paper aims to extend the Microgrids and More-Microgrids concept, as presented in section II, to include and exploit EVs as mobile energy resources. The hierarchy of such a management concept is presented in section III, where new entities and their responsibilities are identified. Section IV presents the operational framework of the control concept which consists of two timelines: the Day­ ahead operation and the Real-time operation. II.

MULTI AGENT SYSTEM

In Multi-Agent Systems (MAS), two or more physical or virtual (software) entities, referred to as agents, have the ability to interact in order to reduce the complexity of a problem. This is achieved by dividing it into smaller sub­ problems. MAS adopt the object-oriented paradigm by keeping private the information needed to solve each sub­ problem. The benefits of agent-based control approaches include [5]: • Flexibility • Extensibility • Fault tolerance In the Microgrids and More-Microgrids projects [6], a multi-agent control concept was developed, as presented in Fig. 1. In a MAS environment, each distributed energy resource is represented by an agent, which is characterized by a certain level of autonomy in taking decisions. Its decision depends on its resources, e.g. the available capacity in case of a storage unit (Le. battery). Moreover, agents have certain behavior and tend to satisfy certain objectives using its resources, skills and services. Agents have only partial or none at all representation of the environment. Each agent knows the

2

Multiple Microgrids can form a Multi-Microgrid system as presented in Fig. 1, managed by a Multi-Microgrid aggregator. Each Microgrid participates in the larger MAS as one entity and is represented by an agent. This agent must fulfill certain objectives, considering the aggregated resources of the controlled Microgrid, through communication with other agents of the larger MAS.

MAS Load Agent Load Agent DGAgent Case 1 Single Microgrid MAS



-

Agents of the larger MAS

III.

DGAgcnt

Case 2 Multiple Microgrids

Fig. l. Microgrids and Multi-Microgrids MAS (source: [7] )

state of the unit it controls, e.g. the State Of Charge (SOC) of a battery, the allowable depth of discharge that prolongs battery life etc. However, agents can be informed via communication with other agents about the status of the neighboring system. In the Microgrid concept, there is a central unit called the MicroGrid Central Controller (MGCC), which is responsible for the optimal participation of the Microgrid in the market. The MGCC is responsible for the necessary negotiations with the Market operator in order to achieve an energy agreement at the best energy prices. The allotment of an energy schedule is optimally allocated to the Microgrid entities through a negotiation process.

CONTROL SYSTEM HIERARCHY

A control system hierarchy that enables the integration of EVs into power systems is analyzed. The core of this control scheme is based on the MAS presented in the previous section. The new Microgrid and Multi-Microgrid entities required for the management of an EV fleet are identified and their responsibilities are analyzed. The operation of power networks with Electric Vehicles may require a management structure which ensures that the grid technical constraints are not violated while the market opportunities are best exploited. Distribution System Operators (DSOs) are expected to be responsible for the technical operation of future power networks and Microgrids. EVs cannot participate in the energy market as standalone entities due to their low power capabilities which are in the order of tens of kW. Their visibility to the system operation and participation in the electricity markets will require a new entity that enables the coordination of the aggregated capacity of an EV fleet.

MV/LV Substation

HV/MV

Level IamiiimilzmC· ·· ..· ..· ..· ..· ..· ..

Substation Level

..................·Iamiilli:m:i:mq· MV Level

··

LV Level

...-:;.... ; .... . ...

To Medium Voltage Network

To Low Voltage Networks

Key - Power Network

I�I

Communications

- Control Hierarchy

Charging Point

Electric Vehicle

RES



Smart Meter

[!] Transformer

OMS EVS/A CAMC RAU CVC MGAU MGCC VC:

Fig. 2. Hierarchical Management Structure of EVS/A in a Distribution Network [II]

Distribution Management System EV Supplier/Aggregator Central Autonomous Management Controller Regional Aggregation Unit Clusters of Vehicles Controller MG Aggregation Unit Microgrid Central Controller Vehicle Controller

3

This new entity will facilitate the participation of the EVs in the electricity markets and/or provide services to the Distribution System Operators (DSO). Such a new entity can be considered similar to the Multi-Microgrid aggregator but with enhanced responsibilities. This new entity is named Electric Vehicle Supplier Aggregator (EVS/A) [8]-[9]. The use of the term "Supplier" indicates the enhanced capabilities of the EVS/A, compared to the Multi-Microgrid aggregator. The aggregator is the entity which is responsible to group the charging demand of a number of EVs and offer demand side management services to the market, managing this aggregated group of EVs. The supplier is the entity that purchases electricity from the power market and sells it to the users. The EVS/A is the responsible entity for both selling electricity to the users and aggregating their load demand. An EVS/A can act either as a standalone business entity or can be part of an Energy supplier, DSO or Energy Service Company (ESCO). In this paper, EVS/A is assumed to be a standalone entity. The EVS/A is hierarchically layered to take advantage of the Multi-Microgrids structure [10]-[11] and enable each Microgrid controller to follow its own policy. Fig. 2 presents the hierarchical management structure of EVS/A in a Distribution Network. The Electric Vehicle Supplier Aggregator (EVs/A) agent manages dispersed EVs in large geographical areas. The EVS/A agent is a physical entity that acts as the intermediary between the EVs, the Transmission System Operator (TSO) and the electricity markets. The EVS/A agent communicates with the downstream network only through the Regional Aggregation Unit (RAU) agents. It is important to note that several EVS/A agents may operate in a single distribution network, thus different MGAU and CVC agents may belong to different companies. The Regional Aggregation Unit (RA U) agent is located at primary substations (HV!MY). The RAU agent manages a number of Microgrid Aggregation Unit (MGAU) agents and Cluster of Vehicles Controller (CVC) agents. The RAU agent aggregates the demand profiles of the controlled entities and sends the compound demand to the EVS/A agent. The compound demand profile should be in compliance with the technical grid constraints. The Microgrid Aggregation Unit (MGAU) agent is located at secondary substations (MVILV). The MGAU agent manages EVs that are dispersed in a LV area. The MGAU agent reduces the amount of information transferred to the upstream network (RAU agent) by providing a single EV load demand for its Microgrid. The Cluster of Vehicles Controller (CVC) agent is located at secondary substations (MYILV). The CVC agent manages EVs that are clustered in large public or private access parking areas. The CVC agent is responsible for managing the demand of the controlled EV fleet in order to adhere to a specific energy schedule. The Vehicle Controller (VC) agent represents the EV owner to the EVS/A agent. The VC agent has the ability to communicate with the respective MGAU in order to express the EV owner's preferences according to their travel profile

and/or with neighboring VC agents in order to be informed about the status of its environment and adjust its state. The direction of the communication depends on the way decisions are taken, centralized or distributed, as it is explained in Section V. VC agents communicate with other parts of the network through the communication channels provided by the smart metering system installed at every charging point. A VC agent will be able to provide either simple or complex functionalities for controlling the power flow between EV and the power grid. In the simplest case, VC agent acts as an ON/OFF switch allowing full power flow at a level depending on the power line's capacity or nothing. In that case, the agent would be located in the charging point and requires the minimum resources, in terms of memory and processing power. It communicates the driver's charging preferences to the MGAU, which decides about the charging schedule (Le. charge or not) for each EV. In the more complex case, the VC agent has the ability to retrieve specific information from the EV concerning the Battery Management System, the EV battery, charger/inverter or any other features. This agent would be located in the EV and the required resources increase as the complexity of the tasks being accomplished increases as well. Apart from local information, the VC agent can find out the current state of its external environment through communication, either in a centralized or a decentralized way. The combination of local and global information enables VC agents to act autonomously and take decisions that fulfill local goals, without violating the constraints of their environment. Examples of local goals are to minimize the battery degradation by adjusting the power flow via the EV charger, or to minimize the charging cost. Constraints of the environment will be considered, e.g. (i) the power exchange between EVs and the grid must be within a specific value range depending on the line capacity, or (ii) the total EV charging demand should not exceed the power limits of MVILV transformers [3]-[4],[12]-[13]. IV.

MOBILE AGENTS FOR MOBILE EV RESOURCES

SO far, the MAS concept of Microgrids and More­ Microgrids refers to a distributed system consisting of several computational units that are connected to each other by a dedicated network. Such networks are more or less static, which means that their topology rarely changes and each computational unit is registered and remains connected to the same network. EVs are characterized by their mobility. They behave as mobile resources which can be connected to different parts of a power network. Thus, the static consideration of the previous networks should be expanded in order to efficiently deploy EVs. Two major requirements should be met in such a Microgrid-EV concept: i) nomadic computing and ii) pervasive computing [14]. Nomadic computing means that EVs are moving from place to place, they are connecting to different charging points (with private or public access) but they still experience the same quality of services and functionalities. Pervasive computing means that everything

4

might become a node in a distributed system and accomplish elementary tasks. These nodes are characterized by limited resources, especially with regard to memory and processing power. The mobility of EVs in the real world should also be considered in the development of MAS for EV deployment. Software agents can be classified in terms of a space by three dimensions as presented in Fig.3: Intelligence, Agency and Mobility [15]. The third dimension named mobility produces a new class of agents called mobile agents. Mobile agents' main functions are their transmission capabilities between different nodes on the same network or different networks, in addition to the inherited capabilities of stationary agents (i.e Agency and intelligence). They are not bound to the system where they begin execution. Contrary to the remote evaluation concept which allows only "code mobility" the concept of a mobile agent supports ''process mobility". In this concept not only the code but also the state information of the agent can be transferred to the destination. The code includes the attributes and methods necessary for the agent to execute. The state refers to the values of the agent's attributes that help it to determine what to do when it resumes execution at its destination. The code and process mobility supported by mobile agents enables the MAS developer to simulate the mobility of electric vehicles from the real world to a MAS environment. This approach can be beneficial when VC's behavior does not follow a uniform pattern (i.e. maximization or minimization of a mathematical expression) but it results from an intelligent computational algorithm (i.e. learning algorithms). In such a case, VC holds the appropriate knowledge and intelligence for exchanging power with the grid adhering to preferable charging strategies but it has limited resources (i.e. memory and processing power) to execute complex tasks (e.g. storing data from their environment during charging and processing them for learning purposes). The transfer of the appropriate VC knowledge and its state to a host where adequate resources exist would enable the execution of these tasks. A possible destination host for the VC mobile agent would be that of MGAU agent (Fig. 4). This requires the existence of a sufficient computational system located at each MVILV substation. SeIVice lnteractivity

Data interacth-;ty Representationofuser Asynchronous Message Passing Remote procedure call Remote execution Weak migration Strongmigration

Mobility

Fig. 3: Space of Software Agents (source: (15) )

MV/lV

MV/lV

SUbstation

SUbstation

.................................... Smanl\itur





Electric Vehicle

@

Vehicle Controller Mobile

��;�

Migratire Mobile

Fig. 4: Mobile Agent concept for EV resources

Any time an EV is plugged at a charging point and acquires a valid authentication access for charging, VC captures the entire agent state, clones and transfer itself to the destination host, where it automatically continues its execution locally at the point they stopped before migration. It may be expected that the transfer of code and execution state may cause additional load to the communication network. However, in an application where distributed intelligence requires multiple interactions to accomplish a task, mobile agent technology can overcome the network latency. V.

OPERATIONAL FRAMEWORK

The operation of the system consists of two timelines: the day-ahead operation and the real-time operation. A. Day Ahead Operation

It is anticipated that the EVSIA will buy a significant part of the required energy, in the forwards electricity market. These transactions will be based on the EV battery charging demand forecasts. The day before energy delivery, the EVS/A will fme-tune its EV load demand forecasts and may prepare bids/offers according to the predicted market behavior of the following day. This demand will consist of schedules in delivery period time steps. These schedules will be sent to the CAMC agent for validation, which ensures conformance to the technical constraints of the distribution system. Post validation, the EVSIA may proceed with market negotiations in the wholesale market. The validated schedules are sent to each RA U agent, which disaggregates and distributes them to each MGAU agent. B. Real-Time Operation

During real time each MGAU/CVC agent communicates with the VC agents. During normal grid operation, the responsibilities of the MGAU/CVC agents depend on the control concept. Two control concepts are identified: the centralized (Fig. 5a) and the distributed (Fig. 5b) control.

5

(a)

-- .

; ��::::

--------

: : :�

'I

t