Review on Energy Resilience

Applied Energy 167 (2016) 397–406

Contents lists available at ScienceDirect

Applied Energy

journal homepage: www.elsevier.com/locate/apenergy

Reliability evaluation of integrated energy systems based on smart agent communication q

http://dx.doi.org/10.1016/j.apenergy.2015.11.033 0306-2619/� 2015 Elsevier Ltd. All rights reserved.

q This work was supported by the National Natural Science Foundation of China under Grant (51577147), the Doctoral Program of Higher Education for the Priority Development Areas, The Ministry of Education, China (20130201130001), the Fundamental Research Funds for the Central Universities, China (2014GJHZ05, XJJ2015034), the Independence research project of State Key Laboratory of Electrical Insulation and Power Equipment in Xi’an Jiaotong University (EIPE14106), and the China Postdoctoral Science Foundation (2014M562410). ⇑ Corresponding author. Tel.: +86 29 8266 8655; fax: +86 29 8266 5489.

E-mail address: [email protected] (Z. Bie).

Gengfeng Li, Zhaohong Bie ⇑, Yu Kou, Jiangfeng Jiang, Mattia Bettinelli State Key Laboratory of Electrical Insulation and Power Equipment, Xi’an Jiaotong University, Xi’an 710049, China

h i g h l i g h t s

� Established reliability evaluation models for Integrated Energy Systems (IESs). � Presented a two-hierarchy smart agent model to describe Smart Agent Communication (SAC). � Presented an IES reliability evaluation approach based on SAC. � Validated models and approaches on a multi-paradigm modeling and simulation platform.

a r t i c l e i n f o

Article history: Received 30 June 2015 Received in revised form 13 November 2015 Accepted 26 November 2015 Available online 15 December 2015

Keywords: Reliability evaluation Integrated energy system Smart agent communication K�1 algorithm

a b s t r a c t

Reliability evaluation of Integrated Energy Systems (IESs) based on Smart Agent Communication (SAC) is studied in this paper. The typical structure and reliability evaluation modeling for IESs is firstly intro- duced. A new reliability evaluation approach is then presented, in which SAC based system reconfigura- tion is innovatively integrated into the reliability evaluation process. Based on this combination, state evaluation (key procedure of the reliability evaluation) along with system reconfiguration can be con- ducted autonomously in reliability evaluation. Algorithm and procedures of the system reconfiguration based on a decentralized agent communication algorithm (K�1 algorithm) is described. The system reconfiguration does not depend on global information of whole system, which can effectively improve reliability evaluation and modeling efficiency. The presented models and approaches are conducted on the multi-paradigm modeling and simulation platform-AnyLogic, and validated by extensive cases studies.

� 2015 Elsevier Ltd. All rights reserved.

1. Introduction

In pressures of fossil energy shortage and global environmental deterioration, the concept of Integrated Energy System (IES) was presented [1–3]. IES is a new type of regional energy system including multiple sub-systems such as electricity, gas, cooling/ heating, and other energy supply systems. IES breaks up the existing mode of individually planning, designing and operating for these sub-systems. In the processes of planning, design, and operation for IESs, the production, transmission, distribution,

conversion, storage, and consumption of various types of energy are coordinated and optimized properly. These coordination and optimization will effectively improve energy efficiency, reduce pol- luting emissions, and lower the dependence of economic and social development on fossil fuels [4].

Generally, various energy supply terminals of electricity, cool- ing, heating, and natural gas are included in IESs, thus, IESs are directly and physically connected to corresponding energy con- sumers. Therefore, IESs become the most important part of energy supply, and their reliability performances have significant impacts on people’s daily life and industrial production. Along with rapid economic development, people’s requirements on the reliability of energy supply are increasing; therefore, ensuring the reliability of IESs has become a crucial task, and related research is in urgent need [5–7].

Actually, the structure, operation mode and operation optimiza- tion of IESs have previously been studied. Refs. [8,9] explore the structure and operation mode of IESs, where an IES structure based on compact transformer was proposed. Ref. [10] presented an idea

398 G. Li et al. / Applied Energy 167 (2016) 397–406

to integrate communication systems into IESs, and basic steps for the combination were introduced as a preliminary study. Ref. [11] proposed a robust operation optimization model for IESs. In the model, wind power uncertainty, operating constraints of natu- ral gas system, coal supply and electricity infrastructure were con- sidered. In Ref. [12], a general and optimal energy conversion path was presented by considering hydrogen as a part of IESs. Besides, researches on energy utilization efficiency of IESs have also drawn widespread attentions. Ref. [13] studied cascade utilization of chemical energy in Combined Cooling Heating and Power (CCHP) systems, and explored how to improve IES energy utilization effi- ciency. Ref. [14] established an IES multi-objective optimization model based on linear programming methods, where the economic operation of IESs was analyzed. In Ref. [15], the energy utilization efficiency of urban integrated energy systems was evaluated using an incremental evaluation method according to cost-benefit analy- sis theory. Ref. [16] proposed a new type of community-level IESs, and an operation control approach based on economic analysis was introduced.

Although quite a few works have been published, researches on reliability evaluation of IESs are still in early stages. In Ref. [17], a gas system was combined with an electric distribution system, reliability evaluation models for the combined system were established, and the largest electricity output of the system was quantitatively evaluated. Ref. [18] analyzed the positive effects of micro-turbine based CCHP systems on the reliability performance of power systems. In Ref. [19], the Markov model was used to quantitative analyze reliability performance of small CCHP system in buildings, the results indicate that combining the supply of heat- ing, cooling and electricity can effectively improve the reliability of energy supply. It is worth noting that existing reliability researches aimed at CCHP systems, which focus on the aspect of energy generation and conversion. Reliability evaluation of IESs consider- ing energy distribution network and interaction effects of electric- ity, gas, cooling and heating systems have not been reported.

Main contributions of this paper include: (1) Reliability evalua- tion models for IESs are established. The Two-state model for most repairable components is introduced, and the power output model for wind turbines is established, in which the uncertainty of wind energy is involved. Besides, a two-hierarchy smart agent model including component agent models and zone agent models is pre- sented, which is the foundation of smart agent communication. (2) An original reliability evaluation approach for IESs is presented, in which Smart Agent Communication (SAC) [20] based system reconfiguration is innovatively integrated into reliability evalua- tion process. By the SAC, system state evaluation (key procedure of reliability evaluation) after failures can be conducted autono- mously along with the reconfiguration process, which effectively improves the reliability evaluation efficiency for IESs. (3) The pre- sented models and approaches are conducted on a multi-paradigm modeling and simulation platform-AnyLogic [21], and validated by extensive cases studies.

This paper is organized as follows: Section 2 introduces the structure and reliability evaluation modeling of IESs. Section 3 describes IES reliability evaluation algorithms. Extensive test results are presented in Section 4. Finally, Section 5 concludes the paper.

2. IES structure and reliability evaluation modeling

IESs contain multiple sub-systems and have significant multi- disciplinary features. This section describes basic structure and reliability evaluation models of IESs. In these models, operating characteristics for different sub-systems and interaction effects between them are taken into account.

2.1. Basic structure of integrated energy systems

Generally, an IES includes electricity distribution network, dis- tributed renewable energy system, gas system, cooling, and heat- ing systems, and a typical structure of IESs is illustrated in Fig. 1.

As shown in Fig. 1, the electricity distribution network mainly consists of electric lines, transformers, and electrical loads. Mean- while, the electricity network is interconnected with the heating system via steam turbine as well as distributed generation system via wind turbine and electricity storage devices. The gas system includes gas wells, gas pipelines and gas boilers. Through gas boiler, chemical energy in natural gas is transformed into heat energy in high-temperature steam, and thus, the gas system is interconnected with the heating system. The high-temperature steam in the heating system can supply heat loads, steam turbines, and heat-driven cooling equipment.

In conclusion, IESs consist of various sub-systems, which have different operating characteristics and interaction with each other. Thus, establishing reliability evaluation models to properly describe these features is of significance to IES reliability evalua- tion. In sub-sections below, IES reliability evaluation models will be introduced.

2.2. Integrated energy system reliability evaluation models

Aim at different operating characteristics of various sub- systems in an IES, different reliability models are established in this paper.

A. Two-state model

The two-state model (see Fig. 2) is a widely used component model in reliability evaluation [22].

As shown in Fig. 2, based on the two-state model, components have two states: Normal state (N) and Repair state (R). In the two-state model, failure rate and repair time of components are assumed to be exponentially distributed. The time to failure (TTF) and time to repair (TTR) of components are generated by:

TTF ¼ � 1 k ln b1 ð1Þ

TTR ¼ �r ln b2; ð2Þ where k and l are the average failure rate and repair rate of a component, respectively. r ¼ 1=l is the average repair time of a component. b1 and b2 are uniform random number between [0, 1].

Based on the two-state model, the state durations (Fig. 3) of a component can be determined, and thus, system state durations can be determined by combining those of all components in the system.

In this paper, most components in IESs are modeled using the two-state model. However, wind is a kind of energy resources that is highly intermittent and random, thus, distributed electricity generators (such as wind turbines) need some more specific mod- els to describe the uncertain electricity output of them, which could have significant effects on reliability.

B. Distributed generation system model

In this section, the reliability evaluation model for wind tur- bines will be introduced. In this model, the uncertain power output is an essential factor. Power output of wind turbines depends on the wind speed, which is commonly modeled using Weibull distri- butions or Autoregressive-Moving Average models.

In this paper, historical wind speed data is fitted by a series of two parameter Weibull distributions, and in every six hours, wind

Power Grid

Steam Turbine

G

Wind Turbine

Cooling Load

Thermal Load

Heat Transfer Refrigeration Equipment

Electrical Load

Energy Storage Device

Power Distribution System

Heating System

Distributed Generation System

Cooling System

Power Line

Heat Pipeline

Natural Gas Well

Gas Pipeline

Gas Boiler

Gas System

Heat Exchanger

G

Transformer

Fig. 1. Schematic diagram of an integrated energy system.

Fig. 2. Two-state model of components.

Fig. 3. Up–down–up cycle curve of a two-state component.

Fig. 4. Power output curve of a wind turbine.

G. Li et al. / Applied Energy 167 (2016) 397–406 399

speed is described using different distributions [23]. Then, the power output of wind turbine Pw(t) can be obtained from hourly sampled wind speeds applying wind turbine output curve (see Fig. 4).

As shown in Fig. 4, the power output of a wind turbine can be determined by:

PwðtÞ ¼ 0 vðtÞ < Vci or vðtÞ P Vco Pr � ½A þ B � vðtÞ þ C � v2ðtÞ� Vci 6 vðtÞ < Vr Pr Vr 6 vðtÞ < Vco

8>< >:

ð3Þ

where Pr is the rated power output of the wind turbine, Vci, Vr, and Vco are the cut-in wind speed, rated wind speed, and cut-out wind speed, respectively. The parameters A, B, and C in Eq. (3) can be evaluated using equations as follows:

A ¼ 1ðVci�VrÞ2 VciðVci þ VrÞ � 4ðVci � VrÞ VciþVr 2Vr

h i3� �

B ¼ 1ðVci�VrÞ2 4ðVci þ VrÞ VciþVr 2Vr

h i3 � ð3Vci þ VrÞ

� �

C ¼ 1ðVci�VrÞ2 2 � 4 VciþVr 2Vr

h i3� � ð4Þ

In reliability evaluation simulation, the power outputs of wind turbines at time t can be sampled based on the models above, and thus, the impacts of wind energy on reliability performance of IESs are involved.

C. Two-hierarchy smart agent modeling

As mentioned above, a smart agent communication based reli- ability approach is presented in this paper. Smart agent modeling is the foundation of agent communication. In this section, a two- hierarchy smart agent model will be introduced. The first (low) hierarchy is the component smart agent, in which components of the same type are modeled using the same smart agent. For instance, line-agent, transformer-agent, and electricity load-agent are established to represent the power lines, transformers, and electricity loads in a power distribution networks. The second (high) hierarchy is the zone smart agent, in which multiple compo- nent smart agents are included, and the agents could be of different types. Generally, zone smart agents are formed based on system topology.

As shown in Fig. 5(a), in the power distribution network, main feeders 1, 2, and 3 are connected with disconnectors, while sub- feeders 4, 5, 6, and 7 are connected with fuses. The distribution network can be partitioned into several zones according to the location of disconnectors and fuses. The zones are then modeled as high hierarchy smart agents. For instance, agent 4 is a high

Fig. 5. Schematic diagram of the two hierarchy smart agent modeling.

400 G. Li et al. / Applied Energy 167 (2016) 397–406

hierarchy smart agent, where three component agents are included: line-agent 4, transformer-agent 1, and load-agent 1.

Based on the AnyLogic modeling platform, in each agent, reliability parameters (such as failure rate, repair time), state simulation functions, and communication interface are defined. It is worth noting that we can build a complex system using several kinds of low hierarchy smart agent models by changing the connecting relationship and parameters of them. Thus, the pro- gramming efficiency can be improved effectively.

3. IES reliability evaluation approach and indices

In this paper, the concept of Smart Agent Communication (SAC) is adopted in the reliability evaluation of IESs. Based on a SAC algo- rithm (K�1 algorithm), system reconfiguration can be conducted autonomously, and system state determination for the reliability evaluation will be implemented along with the reconfiguration process, and thus, the reliability evaluation efficiency can be effec- tively improved. Besides, reliability index system is established to quantify the reliability performance of IESs.

3.1. System reconfiguration based on smart agent communication

As shown in Fig. 5, there are no switches between low hierarchy smart agents, their connection relationships are fixed during relia- bility evaluation, and thus, the communalization between them is simple. However, for the high hierarchy smart agents, they may change the system topology by opening or closing switches accord- ing to the communalization with each other, leading to a complex communication between high hierarchy smart agents.

In this paper, a smart agent communication algorithm named K�1 algorithm is adopted to conduct the communication between high hierarchy smart agents. Moreover, based on the agent com- munication, system reconfiguration process after a failure can be implemented autonomously, which will be a part of the reliability evaluation approach. Generally, a radial topology structure should be guaranteed for IESs, which can prevent large short circuit cur- rent and make the fault identification and isolation easier. It is worth to noting that the K�1 algorithm is suitable for a radial structure, and can guarantee the system reconfiguration preserve the radial structure.

The definition of the K�1 algorithm can be described as follow [20]:

‘‘If agent Xi has K neighbors, Xi will send out a message to its Kth neighbor only after receiving (K�1) message from the other (K�1) neighbors”. In other words, each agent waits until it receives K�1 messages from its K neighbors before sending its message to its Kth neighbor.

In K�1 algorithm, high hierarchy smart agents can be divided into three classes: leaf agents, middle agents, and root agents.

(a) Leaf agents

When a component failure occurs in IESs, according to the K�1 algorithm, agents with only one connected neighbor will send out messages without waiting. These agents are defined as Leaf agents, and they will send out they demand list and initiate the whole reconfiguration process.

(b) Middle agents

If an agent has sent out its message before received all the K messages, it can be defined as a middle agent. Once a middle agent received a message sent by a neighbor, the demand energy list (the amount of energy demand of neighbors) and supply energy list (the amount of energy offered by neighbors) of the agent will be updated. For instance, a middle agent has three neighbors. The agent received two demand lists from two neighbors as follows: DemandList#1 = [L1 L2] and DemandList#2 = [L3], while its own load = [L4]. Then, the updated demand energy list of this agent will be [L1 + L3 + L4, L2 + L3 + L4, L3 + L4, L1 + L4, L2 + L4, L4]. Similarly, the agent will form its supply list according to received supply lists.

Before sending messages, a Middle agent will conduct a demand–supply comparison. If the maximum demand cannot be satisfied, it will send out a demand list, otherwise, a supply offer with a value equals the agent’s maximum supply less the maxi- mum demand as well as a demand list will be sent out.

(c) Root agents

If an agent received messages from all of its neighbors, it is a root agent. Similar to middle agents, the Root agent will conduct a demand–supply comparison based on received messages, and then send out its decision. The decision will be one of the follow- ings: (i) ‘‘DISCONNECT”: this decision will be taken if the Root agent decided not to supply this particular agent. (ii) ‘‘SUPPLY”: when the Root agent decided to use the extra energy that was sent to it from this particular agent. (iii) ‘‘DEMAND ENERGY”: if the Root agent decided to supply this particular agent.

According to the decision of the root agent, all middle agents will take their decisions properly, and implement it via switch actions. Ultimately, leaf agents will receive the decision of their upstream agents, decide their final decision, execute it and the algorithm will terminate.

To illustrate the K�1 algorithm, Fig. 6 shows an example system introduced in Ref. [20]. As shown in Fig. 6, the example system

G. Li et al. / Applied Energy 167 (2016) 397–406 401

includes 7 agents, their demand energy are listed in the figure using symbol ‘‘(�value)”. For instance, the demand of Agent 1 is 7 units. The detailed topology of Feeder 1, Feeder 4, Feeder 3, and Feeder 5 are omitted, and the possible supply energy values of them are 4, 30, 4, 5 units, respectively.

It is assumed that a fault occurred on Feeder 2, and the left dis- connecting switch of Agent 1 will be opened to isolate the fault. In this situation, Agents 1, 6 and 7 are connected only one neighbor, thus, they will send out messages without waiting, in other words, they are Leaf agents and will initiate the whole reconfiguration process.

Based on the execution of the K�1 algorithm, one of Agents 2, 3, 4 or 5 will identify itself as the Root agent. It is worth to noting that the determination of the root agent is done dynamically in real time; it depends solely on the time taken by each Agent to execute its algorithm and sends its messages in real time. In other words, prior to the execution of the restoration algorithm, it is not possible to know which agent will be the Root agent. Here, we assumed that the Agent 3 is the Root agent to illustrate the execution process of the K�1 algorithm. Under this assumption, Agent 3 received mes- sages from Agents 2 and 4 before it sends out its massage, thus, it is the Root agent. For Agents 2, 4, and 5, they have sent out their own messages before received messages from all neighbors, thus, they are Middle agents.

According to the K�1 algorithm, the supply of Agent 1 cannot satisfy its demand, thus, it will send out its demand list [L1 = 7]. Once received the massage from Agent 1, Agent 2 will send out its demand list DemandList#2 = [L2 = 2, L1 + L2 = 9].

Meanwhile, the demand of Agent 6 can be satisfied by its supply (Feeder 5), thus, it will send out its supply list [P6 = 1] as well as demand list [L6 = 4]. Similar to Agent 1, Agent 7 will also send out its demand list [L7 = 7]. For Agent 5, its demand (L7) cannot be satisfied by the supply of Agent 6, thus, it will send out its demand list [L6 = 4, L7 = 7, L6 + L7 = 11]. Therefore, Agent 4 will send out its demand list DemandList#4 = [L4 = 10, L4 + L6 = 14, L4 + L7 = 17, L4 + L6 + L7 = 21].

For Agent 3, it is the Root agent. The demand list of Agent 3 can be obtained by combining demand list [L3], DemandList#2 and DemandList#4, which is [L3 = 7, L3 + L2 + L4 = 19 L3 + L2 + L4 + L6 = 23, L3 + L2 + L4 + L7 = 26, L3 + L2 + L4 + L6 + L7 = 30, L3 + L1 + L2 + L4 = 26, L3 + L1 + L2 + L4 + L6 = 30, L3 + L1 + L2 + L4 + L7 = 33, L3 + L1 + L2 + L4 + L6 + L7 = 37]. While, the possible supply energy of Agent 3 is 30, thus, the alternative reconfiguration schemes include ‘‘L3 + L2 + L4 + L6 + L7 = 30” and ‘‘L3 + L1 + L2 + L4 + L6 = 30”. It is worth noting that the final scheme can be determined by supply priority of agents. If the priority of Agent 1 is higher than that of Agent 7, the final scheme will be ‘‘L3 + L1 + L2 + L4 + L6 = 30” (see Fig. 7), and vice versa.

After the execution of the K�1 algorithm, Agent 7 will receive massage ‘‘DISCONNECT”, other agents will receive message ‘‘DEMAND ENERGY”, and Feeder 4 will receive massage ‘‘SUPPLY”.

Feeder 2

Feeder 1

Feeder 4

Agent 1 Agent 2

Agent 3

(+4)

(+30)

(-7) (-2) (-7)

Fault

Fig. 6. An example system to ill

In conclusion, system reconfiguration can be autonomously car- ried out based on the K�1 communication algorithm. The flow chart of the algorithm is shown in Fig. 8.

It is worth noting that the determination of the root agent is done dynamically during the reliability evaluation process; it depends on the system topology and fault location. In other words, prior to the execution of the system reconfiguration algorithm, it is not possible to know which agent will be the root agent.

3.2. Reliability evaluation approach of integrated energy system

In reliability evaluation, random faults are simulated based on component reliability evaluation models. Once a fault occurs, its consequence needs to be evaluated accurately and quickly. Based on the reconfiguration algorithm mentioned above, the fault con- sequence can be obtained and recorded via the agent communica- tion process, thus the reconfiguration algorithm is very suitable for IES reliability evaluation.

In this paper, a new IES reliability evaluation approach is pre- sented, in which the K�1 communication algorithm is integrated into a sequential Monte Carlo simulation, and the simulation is conducted on AnyLogic platform. The flow chart for the IES reliabil- ity evaluation approach is shown in Fig. 9.

A. Initialization

Set system time T = 0, generate Time to Failure (TTF) for each component agent, and determine power output of the wind turbine.

B. State detection for high hierarchy agents

In every time interval t (could be one minute or one hour), the states of high hierarchy agents are detected, if there is a state change, go to step C, and a reconfiguration process will be initiated. Otherwise, go to step D, to update low hierarchy agent states.

C. System reconfiguration

As mentioned in Section 3.1, a system reconfiguration process is carried out based on K�1 algorithm. It is worth noting that, after the reconfiguration, the energy supply status of loads are deter- mined, and are recorded to calculate reliability indices. Then, go to step E.

D. State update for low hierarchy agents

In the sequential Monte Carlo simulation, states of low hierar- chy agents are updated in each time interval. If an agent in an up-state, its state can be updated by:

TTFTþt ¼ TTFT � t; ð5Þ

Feeder 5

Feeder 3

Agent 4 Agent 5 Agent 6

(+5)

(+3)

(-10)

(-7)

(-4)

(-0)

Agent 7

ustrate the K�1 algorithm.

Feeder 2

Feeder 1

Feeder 4 Feeder 5

Feeder 3

Agent 1 Agent 2

Agent 3 Agent 4 Agent 5 Agent 6

(+4)

(+30) (+5)

(+3)

(-7) (-2) (-7)

(-10)

(-7)

(-4)

Fault (-0)

Agent 7

Fig. 7. The example system after topology reconfiguration.

Leaf agents send out messages, N=1

N = N+1

The N-th agent has received K massages ?

Yes

No

The N-th agent has received K-1 massages ? No

Identify the Root Agent, and form its supply-demand list

Yes

Demand-supply comparison, and form the Final Decision

Send the Final Decision to adjacent agents

All of the Leaf Agents have received the Final Decision ?

Implement the Final Decision, and accomplish the system reconfiguration

End

Start

Yes

No

Form supply-demand list, and send to K-th neighbor

Fig. 8. Flow chart of the K�1 algorithm.

Meet convergence criterion?

Yes

end

No

Start

Each agent detects its own state

State change? No

System reconfiguration and state evaluation based on K-1 algorithm

Yes

Update reliability indices for each load point and whole system

Initialization, set T=0

Generate TTR for faulted agent

Agent faulted Agent repaired

Generate TTF for repaired agent

T=T+t

Fig. 9. Flow chart of the reliability evaluation.

402 G. Li et al. / Applied Energy 167 (2016) 397–406

where TTFTþt and TTFT are the Time to Failure at time T + t and T, respectively. If TTFTþt 6 0 and TTFT > 0, meaning the agent state transfers from up-state to down-state at time T + t, and thus, a new TTR for the agent is then generated based on Eq. (2). If an agent in a down-state, its state can be updated by:

TTRTþt ¼ TTRT � t; ð6Þ

where TTRTþt and TTRT are the Time to Repair at time T + t and T, respectively. If TTRTþt 6 0 and TTRT > 0, meaning the agent state

transfers from down-state to up-state at time T + t, and thus, a new TTF for the agent is then generated based on Eq. (1). For dis- tributed generators, the states are updated based on their reliability models in Section 2.2.

After states of low hierarchy agents updated, go to step B, to detect the states of high hierarchy agents.

E. Convergence criteria

Go to step B until the simulation precision reaches a given con- vergence criteria.

F. Reliability indices calculation

Calculate reliability indices for various energy supplies.

3.3. Reliability indices of integrated energy system

To quantify the reliability performance of integrated energy systems, reliability index system is necessary. There are some

F1

F2

F3

F4

23 24

26

25 27

29

28 30 31

32

33

LP1 LP2 LP3 LP4 LP5 LP6 LP7 34

35

36

37

LP8 LP9 38

39

40

48

49

43

51

52

46

54 LP10 LP11 LP12 LP13 LP14 LP15

LP16 LP17 LP18 LP19 LP20 LP21 LP22

50 53 55

56

57

41 42 44 45

47

58

23 26

24 25

Detail

Fuse

Fig. 10. Topology of RBTS Bus 2.

G. Li et al. / Applied Energy 167 (2016) 397–406 403

widely used reliability indices in reliability evaluation of power systems, and they can be divided into two types: load indices and system indices. The first kind reflects the reliability perfor- mance of particular load point in a system. These indices include: the outage rate k (occurrence/yr), outage duration r (h/occ.) and annual outage time U (h/yr). System indices reflect the reliability performance of the whole system, and include: System Average Interruption Frequency Index (SAIFI), System Average Interruption Duration Index (SAIDI), Average Service Availability Index (ASAI), Customer Average Interruption Duration Index (CAIDI), and Expected Energy Not Supplied (EENS).

Detailed definition for these indices can be found in [24]. In this paper, various energy demands in IES are equivalently transferred into electricity, and thus, reliability performance of IESs on cooling and heat can also be quantified using the indices mentioned above.

4. Case studies and analyses

4.1. Results and analyses based on Bus 2 in RBTS

To demonstrate the presented models and methods, reliability evaluation results in this paper are compared with standard test results. Case studies are firstly conducted on Bus 2 in Roy Billinton Test System (RBTS), which is a widely used reliability test system [25]. The topology of RBTS Bus 2 is shown in Fig. 10, reliability parameters can be found in Ref. [25], and corresponding two- hierarchy agent modeling on the AnyLogic platform is shown in Fig. 11.

The coefficient of variation of index EENS was chosen as conver- gence criterion, which is set to be 1 � 10�4. Reliability evaluation results for indices SAIFI, SAIDI, EENS, and ASAI are shown in Fig. 11, and Table 1 lists the evaluation result comparison.

As shown in Table 1, the results in this paper are close to those in Ref. [25], the maximum error is less than 7%. It is worth noting that the reliability evaluation is conducted based on stochastic simulation using Monte Carlo simulation method, which means

that there will be random errors. Therefore, these results can vali- date the models and methods presented in this paper.

4.2. IES reliability evaluation results and analyses

The presented models and methods are also conducted on a test IES, which is shown in Fig. 12. The test system is constructed based on RBTS Bus 2, in which a wind turbine, a gas system, a cooling sys- tem, and a heating system are integrated. As shown in Fig. 12, the wind turbine is connected to node I. Gas is transported via gas pipelines, and provided to gas-fired boiler, in which high- temperature steam is generated. The steam is then transported via heat distribution pipelines, and supplied to steam turbines (connected to node II and node III, respectively) and heat loads.

In the test IES, reliability parameters for RBTS Bus 2 are the same with Section 4.1. For the wind turbine, historical wind speed data from an actual wind farm is adopted to fit the series of Wei- bull distributions. Cut-in, rated, and cut-out wind speeds for the wind turbine are 3 m/s, 12 m/s and 25 m/s, respectively. Thus, parameters A, B, C in Eq. (4) are 0.121528, �0.07841, and 0.012635, respectively. Reliability parameters for gas, cooling and heating systems are listed in Table 2. These parameters are assumed to follow exponential distributions.

In this paper, cooling and heat loads are converted to equivalent electric power, the total cooing loads is 28 MW, and heat loads is 14 MW. The rated power for the wind turbine is 1.5 MW, those for stream turbines at node II and node III are 0.4 MW and 0.8 MW, respectively.

Based on all the parameters mentioned above, reliability evalu- ation results for different types of loads based on the AnyLogic platform are listed in Table 3. Here, results for electrical load are synthesized results, since there more than one electrical loads in the test system.

As shown in Table 3, the evaluation results of SAIFI indicate the average numbers of interruptions that electrical, cooling, and heat loads would experience per year are 0.231, 0.848, and 0.454,

Fig. 11. Two hierarchy agent modeling for RBTS Bus 2.

Table 1 Reliability evaluation result comparison.

SAIFI (interruptions/customer/yr) SAIDI (h/customer/yr) CAIDI (h/customer interrupted) ASAI EENS (MW h)

Ref. [25] 0.248 3.61 14.55 0.999588 37.74 This paper 0.231 3.59 15.52 0.999667 35.95 Difference (%) �6.85 �0.55 6.67 0.01 �4.74

Fig. 12. Schematic diagram of the integrated energy test system.

404 G. Li et al. / Applied Energy 167 (2016) 397–406

Table 2 Reliability parameters for gas, cooling and heating systems.

Failure rate (occ./yr) Repair time (h)

Gas pipeline (per km) 0.065 5 Heat pipeline (per km) 0.065 5 Gas-fired boiler 0.025 300 Steam turbine 0.03 200 Absorption cooling plant 0.03 200

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

Load 5 Load 11 Load 13 Load 18

U (h

ou r/

ye ar

)

Original system IES formed

Fig. 13. Index U for typical loads.

G. Li et al. / Applied Energy 167 (2016) 397–406 405

respectively. While, the evaluation results of CAIDI indicate that the average interruption durations for electrical, cooling, and heat loads would be 15.143, 77.348, 67.311 h, respectively. Similarly, reliability evaluation results of other indices have also quantita- tively revealed the reliability performance of the IES from different aspects. These evaluation results can provide valuable references for designing or operating IESs with acceptable reliability performance.

It can be seen from Table 3 that electrical loads have the best reliability performance; the Expected Energy Not Supply (EENS) index of electrical loads is 35.65 MW h. The reliability of cooling load is the worst among the three types of loads, the EENS index reaches to 1836.54 MW h per year. The reasons are:

(a) The total demand of electricity loads is about 12 MW in the test IES, which is less than those of cooling (28 MW) and heat (14 MW). Thus, the EENS index of electricity could be much smaller;

(b) Due to low economic benefit, usually, cooling and heat loads have no backup energy sources. Thus, the reliability perfor- mance of cooling and heat supply may be worse than that of electricity supply.

4.3. Impact factor analyses on IES reliability

Reliability performance of IESs are affected by various factors, impacts of the integration of distributed generation, gas, cooling and heating systems on power distribution system are firstly ana- lyzed, and then, impacts of the interaction between sub-systems are also discussed.

A. Impacts of sub-system integration

As shown in Fig. 12, along with establishment of the test IES, distributed generation, gas, cooling and heating systems are integrated into original system, and a wind turbine was connected to node I in feeder F1, and two steam turbines were connected to node II and node III in feeder F3, respectively. They will have effects

Table 3 Reliability indices for different types of loads.

SAIFI (interruptions/customer/yr) SAIDI (h/customer/yr

Electrical load 0.231 3.498 Cooling load 0.848 65.591 Heat load 0.454 30.559

Table 4 Reliability evaluation results of feeder F1 and F3.

SAIFI (interruptions/customer/yr) SAIDI (h/custome

F1 (original system) 0.237 3.5272 F1 (IES formed) 0.225 3.335 Difference (%) �5.06 �5.45 F3 (original system) 0.237 3.697 F3 (IES formed) 0.237 3.635 Difference (%) 0.00 �1.68

on the reliability of feeders F1 and F3. Corresponding evaluation results are listed in Table 4.

As shown in Table 4, after the IES formed, the SAIFI, SAIDI, and EENS indices of feeder F1 are obviously improved, which reduces 5.06%, 5.45%, and 9.75%, respectively. These results show that the integration of the wind turbine can effectively improve the reliabil- ity performance of IES on electricity supply.

For feeder F3, the improvements on all the reliability indices are slight, the most obvious improvement occurs on EENS index, which is only 2.43%. These results indicate that the integration of the two steam turbines has no significant improvement on the reliability performance of feeder F3. The reason is that the rated power of the steam turbines are less, they can only improve the reliability performance of loads directly connected to them. Annual outage time U for several load points are illustrated in Fig. 13.

It can be seen from Fig. 13 that the index U of load points those are close to steam turbines or wind turbine (Load 5, 11, and 13) reduces obviously, while U of load points far from these turbines (Load 18) has almost no changes.

The results and analyses above have revealed the importance of a reasonable capacity configuration on IES’s reliable operation. Reliability performance should be an important constrain when optimizing the capacity configuration in IES planning. Therefore, reliability evaluation is the foundation of IES planning.

B. Impacts of the interaction between sub-systems

In IESs, there may be interactions between sub-systems. As shown in Fig. 12, valves of the heat pipeline are driven by electric- ity from the power distribution system, thus, failure in the power distribution system may affect the cooling and heating systems.

) CAIDI (h/customer interrupted) ASAI EENS (MW h)

15.143 0.999669 35.65 77.348 0.992512 1836.54 67.311 0.996511 427.83

r/yr) CAIDI (h/customer interrupted) ASAI EENS (MW h)

15.072 0.999580 13.397 14.822 0.999621 12.091 �1.66 0.00 �9.75 15.599 0.999556 12.089 15.338 0.999566 11.795 �1.67 0.00 �2.43

Table 5 Reliability evaluation results of cooling and heat loads under different cases.

SAIFI (interruptions/customer/yr) SAIDI (h/customer/yr) CAIDI (h/customer interrupted) ASAI EENS (MW h)

Cooling load Without interaction 0.844 65.227 77.283 0.993 1826.36 Considering interaction 0.909 69.241 76.173 0.992 1938.75 Difference (%) 7.70 6.15 �1.44 �0.10 6.15

Heat load Without interaction 0.454 30.599 67.397 0.997 428.38 Considering interaction 0.519 34.115 65.732 0.996 477.61 Difference (%) 14.32 11.49 �2.47 �0.10 11.49

406 G. Li et al. / Applied Energy 167 (2016) 397–406

In this paper, length of power line 1 and 2 are assumed to be 0.75 km, failure rate is 0.065 occ./yr/km, and repair time is 5 h. Reliability evaluation results for different cases are listed in Table 5.

It can be seen from Table 5 that after considering the interaction between sub-systems, reliability indices SAIFI, SAIDI, and EENS of cooling and heat loads significantly increase. Consistent with these, index ASAI reduces 0.1%. The reason for this phenomenon is that extra interruptions caused by the interaction between sub- systems will be captured in the reliability evaluation, and lead to a more accurate evaluation. Here, index CAIDI decreases about 2%, which means that the duration of the extra interruptions caused by the interactions is smaller than those of the original interruptions.

These results show that interactions between sub-systems decrease the reliability level of IES, thus, independent electricity supply for cooling and heating systems may be necessary. How- ever, because of a low dependence on cooling and heat energy for most customers, in IES planning, extra investment on the inde- pendent electricity supply also needs to be taken into account.

5. Conclusions

In this paper, integrated energy system structure and reliability evaluation models have been introduced, and reliability indices for IESs have been presented. Furthermore, this paper has introduced an IES reliability evaluation approach by combining system recon- figuration based on smart agent communication with sequential Monte Carlo simulation. Finally, reliability evaluation models and methods of IESs have been conducted on a test system, and case studies and analyses have described.

Below are the main conclusions drawn from the results: (1) the presented models and methods can provide reasonable evaluation results on the reliability performance of IESs; (2) because of backup source and system reconfiguration, electricity loads have better reliability performance than those of cooling and heat loads. How- ever, the necessity of providing backup sources for cooling and heat loads needs to further consideration; (3) the interactions between sub-systems have obvious impacts on the reliability per- formance of IESs, which should be properly taken into account in the planning and operation of IESs.

References

[1] European Commission. The fifth framework programme [EB/OL]. <http://ec. europa.eu/research/fp5>.

[2] European Commission. The sixth framework programme [EB/OL]. <http://ec. europa.eu/research/fp6>.

[3] European Commission. The seventh framework programme [EB/OL]. <http:// ec.europa.eu/research/fp7>.

[4] Hongjie JIA, Yunfei MU, Xiaodan YU. Thought about the integrated energy system in China. Elect Power Construct 2015;01:16–25.

[5] Park Sung Ku, Ahn Ji-Ho, Kim Tong Seop. Performance evaluation of integrated gasification solid oxide fuel cell/gas turbine systems including carbon dioxide capture. Appl Energy 2011;88:2976–87.

[6] Liu Xuezhi, Mancarella Pierluigi. Modelling, assessment and Sankey diagrams of integrated electricity–heat–gas networks in multi-vector district energy systems. Appl Energy 2015.

[7] Wei Zhinong, Chen Sheng, Sun Guoqiang, Wang Dan, Sun Yonghui, Zang Haixiang. Probabilistic available transfer capability calculation considering static security constraints and uncertainties of electricity–gas integrated energy systems. Appl Energy 2015.

[8] Loh Poh Chiang, Zhang Lei, Gao Feng. Compact integrated energy systems for distributed generation. IEEE Trans Industr Electron 2013;60(4):1492–502.

[9] Jiang Zhenhua, Yu Xunwei. Modeling and control of an integrated wind power generation and energy storage system. Power & energy society general meeting, IEEE; 2009. p. 1–8.

[10] Hughes JW, Von Dollen DW. Developing an integrated energy and communications systems architecture: the initial steps. In: Power systems conference and exposition, IEEE PES. IEEE; 2004. p. 1651–4.

[11] Martinez-Mares Alberto, Fuerte-Esquivel Claudio R. A robust optimization approach for the interdependency analysis of integrated energy systems considering wind power uncertainty. IEEE Trans Power Syst 2013;28 (4):3964–76.

[12] Hajimiragha A, Canizares C, Fowler M, Geidl M, Andersson G. Optimal energy flow of integrated energy systems with hydrogen economy considerations. In: Bulk power system dynamics and control-VII. Revitalizing operational reliability, 2007. iREP Symposium. IEEE; 2007. p. 1–11.

[13] Jin Hongguang, Gao Lin, Han Wei, Li Bingyu, Feng Zhibing. Integrated energy systems based on cascade utilization of energy. Front Energy Power Eng Chin 2007;1(1):16–31.

[14] Sheikh Shaya, Malakooti Behnam. Integrated energy systems with multi- objective. Energytech, IEEE; 2011. p. 1–5.

[15] Zheng Zhonghai, Fu Lin, Di Hongfa. Incremental evaluation method-an integrated method to assess urban energy system. In: International conference on Management and Service Science (MASS), IEEE; 2010. p. 1–5.

[16] Cartes David, Ordonez Juan, Harrington Julie, Cox Daniel, Meeker Richard. Novel Integrated Energy Systems and control methods with economic analysis for integrated community based energy systems. Power engineering society general meeting; 2007.

[17] Munoz Jorge, Jimenez-Redondo Noemi, Perez-Re Juan, Barquin Julian. Natural gas network modeling for power systems reliability studies, vol. 4. In: Power tech conference proceedings. 2003 IEEE Bologna; 2003. p. 8.

[18] Qi Yudong, Liu Zhiguo, Song Geimei. CCHP and reliability of electricity supply. Adv Mater Res 2011;250:3173–6.

[19] Wang Jiangjiang, Fu Chao, Yang Kun, Zhang Xutao, Shi Guohua, Zhai John. Reliability and availability analysis of redundant BCHP (building cooling, heating and power) system. Energy 2013;61:531–40.

[20] Petcu Adrian, Faltings Boi. A scalable method for multiagent constraint optimization. In: Proceedings of the 19th international joint conference on artificial intelligence, IJCAI-05; 2005.

[21] Ilya Grigoryev. AnyLogic 7 in three days: A quick course in simulation modeling. 2nd ed.; 2014. <http://www.anylogic.com/books>.

[22] Billinton Roy, Ronald A, Norman R. Reliability evaluation of power systems. 2nd ed. New York: Plenum Press; 1996.

[23] Bie Zhaohong, Zhang Peng, Li Gengfeng, Hua Bowen. Reliability evaluation of active distribution systems including microgrids. IEEE Trans Power Syst 2012;27(4):2342–50.

[24] Billinton Roy, Li Wenyuan. Reliability assessment of electric power systems using Monte Carlo methods. New York: Plenum Press; 1994.

[25] Allan RN, Billinton R, Sjarief I, et al. A reliability test system for educational purposes-basic distribution system data and results. IEEE Trans Power Syst 1991;6(2):813–20.