Healer reinforcement for smart grid using discrete event models of FLISR in distribution automation

Abstract

Since a power outage can impose damages to the economy, the value of lost load is much more than the cost of electricity not sold. Hence, efforts should be devoted to achieve a highly reliable energy delivery. Although a large number of power system outages occur in power distribution system, the FLISR mechanism is inefficient in many traditional distribution systems. This leads to late awareness of fault and outages, thus remaining the affected parts of the system without electricity for considerably long periods. Hence, first steps to inject intelligence into the smart grid should be started from the distribution system. In this study, a discrete event system is proposed to model the smart distribution grid operation, specifically to monitor the status of the system, using Petri Nets. Moreover, the proper actions for fault clearance and restoration of the network to the normal state can be obtained using the presented model. The operator is supported by information on fault and the performance of protections. The presented model also helps in reducing the time of fault clearance, thus increasing the system reliability. The proposed model enhances the self-healing ability of smart grid, by reinforcing the healer system, i.e. the FLISR process.

Keywords

FLISR self-healing healer reinforcement Petri Net smart grid distribution automation

1 Introduction

Supplying electricity to customers with the least outages and highest levels of reliability has always concerned electric utilities. The value of lost load is much more than the cost of electricity not sold, as a power outage for even few seconds can impose damages to the economy and social welfare. Hence, efforts are devoted to achieve a highly reliable energy delivery.

After the introduction of the Smart Grids in 2008, this kind of grid became of great interest of the power industry experts, since it promised too many benefits for the power industry. More effective demand response programs, reducing environmental pollution, optimal use of the equipment and eliminate or delay on investments, loss reduction, implementing retail market, self-healing, and distributed generation are of examples of these benefits [1]. Self-healing is of smart grid properties, and it means that the failure occurred in the network could automatically be restored. Self-healing can be performed in component layer, system layer or as healer reinforcement method [2].

Several studies have been conducted in field of Self-healing of grid, through FLISR (fault location, isolation and service restoration) and grid monitoring. In [3], automation system monitoring and service restoration procedures have been performed using Petri Nets. Ref. [4] presents a Petri Net model for locating fault in the radial networks in the presence of distributed generations. A survey of fault diagnosis methods in distribution systems with distributed generations is presented in [5]. In [6], Petri Net is used for grid modelling and fault diagnostics and the effect of fault in a feeder on an adjacent feeder is checked. In [7], artificial neural networks are used for error detection within the power station. A combination of fuzzy and expert systems methods for fault diagnosis in the range of transmission system is used in [8]. In [9], a Petri Net model is proposed for power system and it is used for exact location of the fault when incomplete or uncertain information from fault indicators and protective relays are received. Ref. [10] solves the distribution system restoration problem using dynamic programming with state reduction. Ref. [11] deals with smart grid protection and security issues. Loop restoration scheme (LRS) as a special feeder automation (FA) scheme and two common types of automatic control system (ACS) of LRS are presented and used in [12].

FLISR as the cornerstone of strategies for developing smart grids is described in [13]. In [14] a quantitative evaluation framework for ranking restoration plans with their performance indexes, using the analytical hierarchy process-based fuzzy-grey approach is proposed. Ref. [15] considers time constraints in restoration problem using a fast reasoning expert system mechanism. Refs. [16 –21] build up a multi-agent system (MAS) for service restoration. Smart Grid concept and technologies to construct a Self-healing framework for use in distribution systems are applied in [22]. More types of service restoration and FLISR techniques are discussed in [23 –28].

An automated FLISR scheme can be regarded as a part of the Self-healing property of the smart grid, acting as the immune system for the smart grid. The self-healed smart grid can detect the occurring or emerging failures, isolate them and restore the lost services. Self-healing is expected to work without –or with least–human intervention. Apparently, the faster and more efficient healer system, the less duration of outages is reckoned, leading to a substantially more reliable power distribution system.

Since the vast majority of failures occur in distribution networks, the need for further and more appropriate research on FLISR, in the distribution network, appears. Smart operation of the distribution network, and penetration of distributed generation, have led to complexity of monitoring of automation process. Thus, the traditional monitoring methods are no longer responsive by themselves, increasing the possibility of wrong decisions in operation of these networks. Hence, the need for more sophisticated methods for monitoring and automation systems to achieve higher levels of performance is inevitable [2].

In this study, a discrete event is presented to model the smart distribution network operation, specifically to monitor the status of the system, using Petri Nets. Moreover, the proper actions for fault clearance and restoration of the network to the normal state can be obtained using the presented model. The salient feature of this model is its comprehensiveness in including faults of both middle and low voltage parts of network. Moreover, the operator is supported by information from fault location, fault type (transient or permanent,) and the performance of main and backup protections. The presented model also helps in reducing the fault clearance time, thus increasing the reliability of the system. The proposed model enhances the Self-healing ability of smart grid, by healing/reinforcing the healer system. Ref. [29] describes healer reinforcement.

Rest of the paper is organized as follows: Section II deals with introducing Petri Net. In the next two sections the under study distribution network is introduced and then its proposed Petri Net model is derived and discussed. In Section V, software is designed based on Petri Net models in order to be run beside other software existing in Control Centre (CC). And in Section VI the model is validated within several scenarios. Section VII completes previous sections models and software so that backup switching and its monitoring became possible. Section VIII concludes the work.

2 Petri Net

2.1 Petri Net basics

A Petri Net consists of three basic components: place with a circular symbol, transition with a symbol of rectangle and arcs which represent connections between places and transitions (Fig. 1). Transitions represent the modeled system events, and places and the token inside them represent post and pre conditions of an event.

If a transition is enable (the number of tokens in each input place of that transition at least should be as much as the weight of its corresponding arc,) then it has possibility to fire or occur and if it fires, the number of tokens in each of its input places will be reduced equal to the weight of the corresponding arc. It deposits in each output place of that transition the number of tokens equal to the weight of corresponding arc.

As an example in Fig. 1, T1 is an enabled transition, Fig. 2 showing it after being fired.

The matrix m×1 is defined as state matrix for Petri Net S (with m transitions and n places). Which its ith entry shows the number of tokens existing in ith place at kth time instant and its state evolution is stated as follows: $q_{s} [k + 1] = q_{s} [k] + B . x [k]$ (1)

B is called incidence matrix and is defined as: $B = F^{+} - F^{-}$ (2)

F^- (i,j) is the weight of the arc from place Pi to transition Tj, and F+ (i, j) is the weight of the arc from Tj to Pi. The m×1 matrix X[k] is called control vector (fire vector) and defines priority of transition firing in the process. In each stage of the process, only the entry related to each sequence is 1 and other entries are zero.

2.2 Failure diagnostic algorithm using Petri Net

If an error occurs in the under study system or process, this error is reflected in the Petri Net model of the process. Thus it is possible to reach to the error occurred in the system/process. The method presented in [2, 30] is borrowed to diagnose a failure in a Petri Net. According to the mentioned method, some additional places as monitoring places are added to the main Petri Net, thus forming a new Petri Net called H, which for all time instances has the following relation with the main Petri Net S: $q_{h} [k] = [\begin{matrix} I_{n} \\ G \end{matrix}] q_{s} [k]$ (3)where In is the n-dimensional identity matrix, G is a d×n matrix called generator matrix, and d is the number of added places. Considering (3), the state equation of Petri Net H can be written as: $q_{h} [k + 1] = q_{h} [k] + [\begin{matrix} B \\ G B \end{matrix}] X [k]$ (4)

If a failure occurs at time instantk-1 in the system/process under study, the next state of Petri Net can be expressed as: $q_{f} [k] = q_{h} [k] + e [k]$ (5)where is the state matrix of Petri Net h without failures and e [k] is the effect of the failure. In order to detect failures in a Petri Net, a parity check matrix is defined and calculated as: $S \equiv [\begin{matrix} - G I_{d} \end{matrix}]$ (6)and s [k] which is called error syndrome matrix at time instant k is written as: $s [k] \equiv [\begin{matrix} - G I_{d} \end{matrix}] \begin{matrix} \end{matrix} q_{f} [k] = [\begin{matrix} - G I_{d} \end{matrix}] \begin{matrix} \end{matrix} e [k]$ (7)whenever the error syndrome matrix is nonzero, an error has occurred in the under study system. A failure can be detected, if this matrix is nonzero. It can be identified, if the mentioned syndrome can uniquely be expressed by columns of parity check matrix. In this case the number of each column shows the failure place numbers [30].

3 The under study distribution network

3.1 The legacy distribution system

Figure 3 shows a smart distribution network, to which the proposed Petri Net based healer reinforcement method applied.

The network is divided into eight medium voltage (MV) sections; each area also has a number of 20 kV/400V transformers (stations) installed on it, feeding the connected loads on the low voltage (LV) parts of the network. A distributed generation (DG) source is intended in Section 2.2. Thus this area of the network can provide electricity for its loads both in grid-connected and island mode.

The presence of this distributed generation will help operating this section in island mode in some fault cases, thus the number and duration of outage for the healthy part of the system is expected to be reduced.

The available recloser in the network which is coordinated with the sectionalizer helps in clearing the transient faults at least time and before its fault current enters the sub-transmission station 63/20 [kV]. The breaker in beginning of feeder has a digital relay: Once the errors occur the digital relay will produce pick-up and operation signals and if the relay trips it will produce trip signal.

Along the chain of protection operation, the “pickup” and “operation” of relay elements are more related to the fault than the trip of relay and opening of circuit breaker, since the fault directly creates the “pickup” and “operation” of relay elements. On contrary, the trip of relay and opening of circuit breaker are more affected by pickup and operation of relay elements. In these effects, the trip of relay and opening of circuit breaker may be more affected by uncertainty indices. For example, even if an element of relay successively responds to a fault by “pickup” and “operation”, but it is likely that the contact of relay trip is defective because of electrical or mechanical problems [31]. Use of information with higher reliability contains more guarantees. Thus, in determining whether or not the relay has seen an error, signals with higher reliability (pick-up and operation) are also used in this study.

3.2 The associate communication link

An effective IT infrastructure –serving as a communication link–is essential for monitoring the smart distribution network [32, 33]. Remote terminal units (RTUs) are used for switches. A two-way radio communication network is used for data and control command transmission between the Control Centre and local automation equipment. This scheme is shown in Fig. 3. Information is collected at the Control Centre and the control commands are sent from there.

4 The Petri Net model

In the proposed scheme, Petri Nets are used to model the FLISR as discrete event systems. The proposed model monitors all the procedures of FLISR automatic operation, and supporting the operator with suggestions for proper actions. It can work fully autonomous when needed, as well.

In addition to modeling of transient and permanent faults, and checking faults of the low voltage side, backup protections have also been modeled. Furthermore, smart fault indicators have been applied, in order to rapidly detect the fault location.

Moreover, auxiliary support models have also been developed to implement backup switching of fault isolating equipment. Using this facility, the healer itself is reinforced to not only detect the unexpected operation of any equipment during the FLISR procedure, but also choose another way for continuing the monitoring and fault restoration procedure. Backup switching will be discussed more in part 8.

The sequences described in part II are used to provide Petri Net model of desired process. Expected events in the process constitute transitions of the Petri Net. In our process, these events are FLISR sequences. Depending on how and where the fault occurred is slightly different for different areas. Information available at the Control Centre will also form the model places. These available data –i.e. the received signals at the Control Centre–can be found in Table 1.

The set of desired Petri Net places are divided into several groups, based on the under study sections of distribution network and type of fault occurred (transient or permanent,). Each group is related to participating equipment for that type and zone of fault. The proposed block diagram and models for monitoring of sample smart distribution network of Fig. 3 are shown in Figs. 4 to 6, respectively.

The part related to fault in the LV zones of network is plotted separately (Fig. 3). Note that some arcs have a small circle on their end –inhibitor arcs–and change the enabling rule in a way that existence of tokens in the corresponding input places prevents the transitions to be enabled [34].

After determination of permanent fault in eight sections, the Petri Net model for each section can be evaluated separately to reduce calculation complexity.

In general, the proposed model works in such a way that when a fault occurs in a LV part, its relevant model runs and clears the fault. Only if the main protection and first backup protection does not work, it will interact with other parts of the model. In the case these two mentioned protections fail, the fault transfers to the related MV section and so the proper actions for fault clearing in this MV section will be continued by other related parts of model.

5 Grid monitoring software

The Petri Net model is a graphical model that makes it easier to understand the system behaviour which is modelled by. However, mathematics is needed to evaluate the behaviour and performance of systems. Although Petri Net model is plotted graphically, yet it uses matrix mathematical calculations. On the other hand, computer software is needed to use the achieved graphical model in distribution automation monitoring. It can be run beside other Control Centre software. Therefore, computer software is designed based on mathematical relations of Petri Nets. Thereafter, the performance of the proposed model is evaluated using this software. The software is designed in MATLAB software environment.

Generally different information is communicated between a distribution network and Control Centre (see Table 1), in which Petri Net based software receives this information and enables corresponding place of each in the Petri Net model. If this information is consistent with the expected and pre-defined pattern of the Petri Net model, the model goes to the next step and waits for further information. Vice versa, if the information does not match the pattern, model stops and gives an error message to the operator. Outgoing messages of the model are as follows:

Declaration of fault zone

Declaration of error step in automation steps

Declaration the failed signals

Declaration of nature of fault (permanent/transient)

Declaration of the phase that a transient fault is cleared by recloser

Declaration of the type of protection used in low voltage fault clearance

Specific declaration of main feeder digital relay trip fault

Declaration of adjacent feeder free capacity to accept main feeder loads

During the process of distribution automation, the model continues until no error receives from the distribution network or control. There are two probabilities for having wrong status in this system:

A signal is wrongly transmitted to the Control Centre, e.g. the status of a switch is not transmitted to the Control Centre or is unexpectedly received (deficiency in communication system,) or the switch really has a wrong status (deficiency in switch or breaker).

Wrong command decisions are sent from Control Centre to the distribution network. In this case, the model stops immediately and appropriate error messages gained from the model are shown in the Control Centre’s display. And the operator will be notified of the status of process and wrong conditions happened.

6 Evaluation of the proposed model

6.1 First type scenario; normal operation

6.1.1 Fault in Section 1

A fault is simulated in Section 1 and it is assumed that adjacent feeder does not have enough capacity to provide the full lost loads of the faulted feeder, yet is able to feed the downstream parts of S3. Further, the trip signal of main feeder relay is not received. However the pickup and operation signal are received (Fig. 7).

As shown in Fig. 7, the program is working properly. It states that the fault is on Section 1, and in the second line, mentions that the adjacent feeder can only partially provide power for the faulted feeder. Moreover, in the third line states that an error has occurred in fault occurring sequence and R1 trip signal has not received. Nevertheless, the software has continued the routine. This notice, announces the operator that possibly a deficiency is in relay that has not transmitted the trip signal.

The island operation advantage of Section 2.2 is illustrated in this case. It is seen that during the fault, Sections 3, 3.1 and 3.2 are fed through adjacent feeder. Moreover, the loads on Section 2.2 are fed by DG. Further, during repair time, the only loads of Section 2.1 are in outage. Hence, the system reliability is increased.

6.1.2 Fault on low voltage part no: 8 (LV8)

It is assumed that main protection fails and fault has removed by first backup protection (Fig. 8).

6.1.3 Transient fault in Section 3.1, and second operation of recloser

Figure 9 shows the output for this situation.

6.2 Second type scenario: Operation under abnormal conditions

6.2.1 Permanent fault in Section 2

The Petri Net model continues step by step, along with real network till the last step (normal operation), while step by step receiving of information from the Control Centre. The fault-relevant part of the model is executed, after determining the fault location and its nature. The other parts of model play no role in simulation. Thus it is possible to, based on the fault location; write a small incidence matrix for that part. In other words, write a small matrix for each section instead of considering a large matrix for entire model [2].

Therefore, in current assessment conditions, which is simulation of a permanent fault in Section 2, its special incidence matrix, B2, is needed. B2 is calculated as in Equation (8):

$\begin{matrix} t 1 & t 2 & t 3 & t 4 & t 5 & t 6 & t 7 \end{matrix}$ $B = [\begin{matrix} - 1 & 0 & 0 & 0 & 0 & 0 & 0 \\ - 1 & 0 & 0 & 0 & 0 & 0 & 0 \\ 1 & - 1 & 0 & 0 & 0 & 0 & 0 \\ 0 & 1 & - 1 & 0 & 0 & 0 & 0 \\ 0 & 1 & - 1 & 0 & 0 & 0 & 0 \\ 0 & 0 & 1 & - 1 & 0 & 0 & 0 \\ 0 & 0 & 1 & - 1 & 0 & 0 & 0 \\ 0 & 0 & 0 & 1 & - 1 & 0 & 0 \\ 0 & 0 & 0 & 1 & - 1 & 0 & 0 \\ 0 & 0 & 0 & 0 & 1 & - 1 & 0 \\ 0 & 0 & 0 & 0 & 1 & - 1 & 0 \\ 0 & 0 & 0 & 0 & 0 & 1 & - 1 \\ 0 & 0 & 0 & 0 & 0 & 1 & - 1 \\ 0 & 0 & 0 & 0 & 0 & 0 & 1 \end{matrix}] \begin{matrix} Sec 2 \\ R . Cc \\ R . Co \\ S 2 o \\ S 3 o \\ B 2 c \\ TSc \\ B 2 o \\ TSo \\ S 2 c \\ S 3 c \\ R . Cc \\ B 2 c \\ P 1 \end{matrix}$ (8)

State matrix and input matrices for this case can be written as follow: $\begin{array}{l} q_{s} [0] \\ = S e c 2 R . C c R . C o S 2 o S 3 o B 2 c T S c B 2 o T S o S 2 c S 3 c R . C c B 2 c P 1 \\ [\overset{︷}{1 1 0 0 0 0 0 0 0 0 0 0 0 0}] \end{array}$ (9)

$x |_{i = 1} = {[\begin{matrix} t 1 t 2 t 3 t 4 t 5 t 6 t 7 \\ \overset{︷}{1 0 0 0 0 0 0} \end{matrix}]}^{T}; i = 1, \dots,$ (10)X_i is calculated for i = 1 (i.e. first sequence) and for each sequence, the entry of X_i will be “1” and other entries will be “0” depicting the priority of the sequences. The method used in this study obtaining the generator matrix G, is the method described in [35]. Based on this method, the overall equation for obtaining G is:

$G = [\begin{matrix} 1 & 1 & 1 & \dots & 1 \\ 1 & 2 & 3 & \dots & n \\ 1 & 2^{2} & 3^{2} & \dots & n^{2} \\ ⋮ & ⋮ & ⋮ & ⋱ & ⋮ \\ 1 & 2^{d - 1} & 3^{d - 1} & \dots & n^{d - 1} \end{matrix}]$ (11)

In this equation, n is the number of main places of model, d is number of added places to the Petri Net and are called monitoring places. The number of monitoring places depends on the number of the failures which are expected to diagnose concurrently every sequence. If we expect from the model to diagnose h failures in every sequence, then the number of the monitoring places which should be added to main Petri Net is d = 2h [5].

Therefore, to obtain generator matrix for fault in Section 2 (G₂), and since n = 14, m = 7 and assuming that at most 4 failures are expected to diagnose concurrently (h = 4), then d = 8 and have: $G = [\begin{matrix} 1 & 1 & 1 & \dots & 1 \\ 1 & 2 & 3 & \dots & 14 \\ 1 & 2^{2} & 3^{2} & \dots & 14^{2} \\ ⋮ & ⋮ & ⋮ & ⋱ & ⋮ \\ 1 & 2^{7} & 3^{7} & \dots & 14^{7} \end{matrix}]$ (12)and the parity check matrix S2 can be extracted like: $S_{2} = [- G_{2} ⋮ I_{8}]$ (13)

6.3 Two Possibilities for the Second Type Scenario:

In this part, two possibility of second type of scenario that will stop the model, are designed to evaluate the accuracy of model in diagnosis of this false conditions.

6.3.1 First possibility

Suppose that the following information which means occurring a permanent fault in Section 2, are received in Control Centre:

Signal of fault on Section no2: Sec2 = 1

Signal of 3th closing of Recloser: R.Cc = 1

q_f [0] should be extracted first. For this, using (3) and (9) yields:

$\begin{matrix} q_{f} [0] \\ = \underset{[1}{Sec 2} \underset{1}{R . Cc} \underset{0}{S 2 o} \underset{0}{S 3 o} \underset{0}{B 2 c} \underset{0}{TSc} \underset{0}{B 2 o} \underset{0}{TSo} \underset{0}{S 2 c} \underset{0}{S 3 c} \underset{0}{R . Cc} \underset{0}{B 2 c} \underset{0}{P 1} | \underset{2 3, 9, \dots 129]^{T}}{monitoring places} \end{matrix}$ (14)q_f is calculated by (3) in every sequence and then is updated by received data in the Control Centre and monitors the process in real-time. For example, in first sequence, according to received data in the Control Centre, no change is made in q_f [2]. A comparison between updated q_f and q_f of before update (which is expected from normal operation of model with no errors) is needed in each sequence in order to diagnose faults and wrong statuses. This comparison is done with error syndrome matrix, s_f [k]. This matrix is calculated by (7) and based on updated q_f.

Suppose that automation sequences have been correctly executed until fifth sequence, q_f can be calculated as:

$\begin{matrix} \overset{S e c 2 R . Cc R . C o S 2 o S 3 o B 2 c T S c B 2 o T S o S 2 c S 3 c R . C c B 2 c P 1}{︷} \\ q_{f} 1 = [0 & 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 | & 1 & 3 & 9 & \dots & 2187]^{T} \\ q_{f} [2] = [0 & 0 & 0 & 1 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 | & 2 & 9 & 41 & \dots & 94509]^{T} \\ q_{f} [3] = [0 & 0 & 0 & 0 & 0 & 1 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 | & 2 & 13 & 85 & \dots & 1103479]^{T} \\ q_{f} [4] = [0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 1 & 0 & 0 & 0 & 0 & 0 | & 2 & 17 & 145 & \dots & 6880121]^{T} \\ q_{f} [5] = [0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 1 & 0 & 0 & 0 & 0 & 0 | & 2 & 21 & 221 & \dots & 29487171]^{T} \end{matrix}$ (15) where q_f [5] has been updated by received information in Control Centre as below:

Open status of B2 = 1

Open status of TS = 1

Close status of S2 = 0

Close status of S3 = 0

So error syndrome matrix can be written as:

S_{f} [5] = \begin{matrix} [0 & 4 & 76 & 1090 & 169234 & 1977976 & 22607050]^{T} \end{matrix}

(16) It can be written by its linear combination as:

S_{f} [5] = S_{2} (:, 8) + S_{2} (:, 9) - S_{2} (:, 10) - S_{2} (:, 11)

(17)It states failures in places 8th, 9th, 10th and 11th. It is seen that in this sequence, Control Centre has sent a close command to S2, S3 and the close status of this sectioners should be received in this sequence, but are not received due to internal error in sectioners or error in communication system. Also open status of B2 and TS are received unexpectedly. These wrong statuses are diagnosed by the software and it is shown in Fig. 10.

6.3.2 Second possibility

Suppose that following signals in third sequence have been received which imply open status of B2 and TS:

Open status of B2 = 0

Open status of TS = 0

Matrix q_f [3] after calculation and updating with the information received in the Control Centre, will be as follows: $\begin{matrix} \overset{Sec 2 R . Cc R . Co S 2 o S 3 o B 2 c TSc B 2 o TSo S 2 c S 3 c R . Cc B 2 c P 1 monitoring places}{︷} \\ q_{f} [3] = [0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & | 2 & 13 & 85 & \dots & 1103479]^{T} \end{matrix}$ (18)Therefore, calculating error syndrome matrix for third sequence yields: $S_{f} [3] = [\begin{matrix} 213 85 559 3697 24583 164305 1103479 \end{matrix}]^{T}$ (19)It can be written by its linear combination as: $S_{f} [3] = S_{2} (:, 6) + S_{2} (:, 7)$ (20)Based on rules, it states that failures have been occurred in places 6th, 7th which means close status of B2 and TS has not been received. Figure 11 shows the software output for this condition.

7 Backup switching

From the various stages of the FLISR process, isolating faults is more critical and must be implemented properly and in the shortest possible time. Hence, in this part, the study attempts to consider the possibility of a malfunction in the equipment which are responsible for fault isolation, choose another approach to isolate the fault and hence FLISR process continues.

Consider a fault in previous model and the proposed Petri Net model runs to manage the fault, but it sees an error in isolating sequence. Figure 12 shows an example of this situation.

Due to not reception of open signal of S3, there are two possibilities; first, the switch is really damaged and unopened, second, it is not damaged and also open signal is produced. However, the signal has got error when it is transmitted to the Control Centre. In other words, communication system has got error.

Here, we assume a perfect communication platform and the mentioned deficiency returns to the equipment itself. This assumption is reasonable because of the generally high reliability of communication systems. On the other hand, in many errors occur during the transmission of the signal, at the receiving side (Control Centre), using methods such as error checking, existence of these errors can be discovered. Moreover, using methods such as Hamming code it is possible to correct these errors and the actual signal transmitted can be achieved.

Depending on how many sides a fault-prone section can be fed, one or two equipment participates in isolating that section. Auxiliary models have been developed to implement backup switching of fault isolating equipment. Using this facility, the healer itself is reinforced in order to not only detect and inform of non-expectation operation of any equipment during FLISR procedure, but also choose another approach for continuing the monitoring and fault restoration procedure. Using simulation results, improved performance of base monitoring and fault management program is shown.

Figure 13 shows a sample of these auxiliary models. That is for conditions described in Fig. 12.

7.1 Auxiliary models evaluation and their performance simulation

7.1.1 Fault in Section 1; non-operating of recloser in isolating fault

Figure 14 shows the augmented software output and Fig. 15 shows the base software output. As evident from the above figures, the base model after diagnosing interruption in operation due to lack of signal reception of recloser opening, stops from working. But augmented model in addition to including the base program looks for an alternative way to continue the automation monitoring process and has provided an alternative way.

It is worth noting that the fault-monitoring process is still considered. And if an error occurs in the alternative approach, which it is less likely, the software detects and announces the error as before.

8 Conclusions

In this study, evaluating of proposed Petri Net models for a smart distribution network was discussed. Transient faults as well as fault occurring in low voltage side of distribution network and fault clearing in this voltage level, backup protection modelling and so on are of other things that have been considered.

It is very important to know the free capacity of the adjacent feeder to choose an appropriate strategy for fault recovery procedures in the network. The study has seen this important issue. Furthermore, auxiliary models have been also developed to implement backup switching of fault isolating equipment. Using this facility, the healer itself is reinforced. The accuracy of the model was evaluated by assessing its performance in different scenarios. And the results of software demonstrate its success in implementation of all expectations.

References

Fereidunian

, Lesani

, Zamani

M.A.

, Kolarijani

M.A.S.

, Hassanpour

, Mansouri

S.S.

, A Complex Adaptive System of Systems Approach to Human–Automation Interaction in Smart Grid, in Contemporary Issues in Systems Science and Engineering (eds Zhou

, Li

H.-X.

and Weijnen

), John Wiley & Sons, Inc., Hoboken, NJ, USA, 2015. doi: 10.1002/9781119036821.ch12

Tsai

, Pen

, Application of BDI-based intelligent multi-agent systems for distribution system service restoration planning, Euro Trans Electr Power (2011), 1783–1801.

, Liu

, Novel multiagent based load restoration algorithm for microgrids, Smart Grid, IEEE Transactions on2(1) (2011), 152–161.

, Zhang

, Sutanto

, A hybrid multiagent framework with Q-learning for power grid systems restoration,Power Systems, IEEE Transactions on26(4) (2011), 2434–2441.

Solanki

J.M.

, Khushalani

, Schulz

N.N.

, A multi-agent solution to distribution systems restoration, Power Systems, IEEE Transactions on22(3) (2007), 1026–1034.

Zidan

, El-Saadany

E.F.

, A cooperative multiagent framework for self-healing mechanisms in distribution systems, Smart Grid, IEEE Transactions on3(3) (2012), 1525–1539.

Kumar

, Das

and Sharma

Jaydev

, Multi-objective, multi-constraint service restoration of electric power distribution system with priority customers, Power Delivery, IEEE Transactions on23(1) (2008), 261–270.

Donde

, Wang

, Yang

, Stoupis

, Short-term load forecasting based capacity check for automated power restoration of electric distribution networks, Transmission and Distribution Conference and Exposition, 2010IEEE PES, 2010, pp.1–8.

Zidan

, El-Saadany

E.F.

, Service restoration in balanced and unbalanced distribution systems with high DG penetration, Power and Energy Society General Meeting, 2011 IEEE, 2011, pp.1–8.

10.

Hadjicostis

C.N.

, Verghese

G.C.

, Power system monitoring using Petri Net embeddings, Generation, Transmission and Distribution, IEE Proceedings-147(5) (2000), 299–303.

11.

Ashouri

, Jalilvand

, Noroozian

, Bagheri

, A new approach for fault detection in digital relays-based power system using Petri Nets, Power Electronics, Drives and Energy Systems (PEDES) & 2010 Power India, 2010 Joint International Conference on2010, pp.1–8.

12.

Fereidunian

, Lucas

, Lesani

, Rahmani

, Wymore

A.W.

, A Policy-Driven Method for IT Infrastructure Selection in Power Distribution Automation System, International Review of Electrical Engineering, Part B5(2) (2010), 671–682.

13.

Fereidunian

, Nordman

, Lesani

, Lucas

, Lehtonen

, A systems approach to information technology (IT) infrastructure design for utility management automation systems, Iranian Journal of Electrical and Electronic Engineering2(3&4) (2006), 91–105.

14.

Zurawski

, Zhou

, Petri Nets and industrial applications: A tutorial, Industrial Electronics, IEEE Transactions on41(6) (1994), 567–583.

15.

Ren

, Mi

, Zhao

, Yang

, Fault diagnosis for substation automation based on Petri Nets and coding theory, Power Engineering Society General Meeting, 2004 IEEE1 (2004), 1038–1042.

16.

Tsai

M.S.

, Development of an object-oriented service restoration expert system with load variations, Power Systems, IEEE Transactions on23(1) (2008), 219–225.

17.

, Song

, Wang

, Multi-agent approach for service restoration of microgrid, Industrial Electronics and Applications (ICIEA), 2010 the 5th IEEE Conference on, 2010, 962–966.

18.

J.S.

, Lee

T.E.

, Lu

C.S.

, Su

S.D.

, An Autonomous Decision Approach for Fault Allocation and Service Restoration in Electrical Distribution Systems by Multi Agent System, Hybrid Intelligent Systems, 2009 HIS ’09 Ninth International Conference on, vol. 3, 2009, pp.89–94.

19.

Shahsavari

, Fereidunian

, Lesani

, A healer reinforcement approach to Smart Grid Self-healingby redundancy improvement, Electrical Power Distribution Networks (EPDC), 2013 18th Conference on, 2013, 1–9.

20.

Saki

, Fereidunian

, Lesani

, Kolarijani

M.A.S.

, (2011), Distribution automation monitoring using Petri Nets, pp. , Control, Instrumentation and Automation (ICCIA), 2011 2nd International Conference on56–61.

21.

Calderaro

, Hadjicostis

C.N.

, Piccolo

, Siano

, Failure identification in smart grids based on petri net modeling, Industrial Electronics, IEEE Transactions on58(10) (2011), 4613–4623.

22.

YaJie

, Xiaoli

, XiaoHui

, Survey and development on fault diagnosis in distribution network with distributed generators, Advanced Power System Automation and Protection (APAP), 2011 International Conference on, vol. 1, 2011, pp. 13–17.

23.

Calderaro

, Galdi

, Piccolo

, Siano

, A Petri Net based protection monitoring system for distribution networks with distributed generation, Elsevier Journal on Electric Power Systems Research79 (2009), 1300–1307.

24.

Rodriguez

, Rementeria

, Martin

J.I.

, Lafuente

, Muguerza

, Perez

, A modular neural network approach to fault diagnosis, Neural Networks, IEEE Transactions on7(2) (1996), 326–340.

25.

Tan

J.C.

, Crossley

P.A.

, McLaren

P.G.

, Fuzzy expert system for on-line fault diagnosis on a transmission network, Power Engineering Society Winter Meeting, 2001 IEEE, vol. 2, pp.775–780.

26.

Shahsavari

, Mazhari

S.M.

, Fereidunian

, Lesani

, Fault indicator Deployment in distribution systems considering available control and protection devices: A multi-objective formulation approach, IEEE Trans on Power Systems29(5) (2014), 2359–2369.

27.

Shahsavari

, Fereidunian

, Mazhari

S.M.

, A joint automatic and manual switch placement within distribution systems considering operational probabilities of control sequences, International Transactions on Electrical Energy Systems (ETEP), Online first, 2014.

28.

Fereidunian

A.R.

, Lesani

, Lucas

, Distribution system reconfiguration using pattern recognizer neural networks, International Journal of Engineering (IJE), ISSN: 1025-2495, Transactions B: Applications15(2) (2002), 135–144.

29.

Fereidunian

, Healer Reinforcement: A Cybernetic Approach to Self-healing in Smart Grid, In Proc of the Smart Grid Conference (SGC 2015), NRI, Tehran, Iran, 2015.

30.

Ashouri

, Jalilvand

, Noroozian

, Fault diagnosis modeling of power systems using Petri Nets, Power Engineering and Optimization Conference (PEOCO), 2010 4th International, 2010, 313–317.

31.

Perez-Guerrero

, Heydt

G.T.

, Jack

N.J.

, Keel

B.K.

, Castelhano

A.R.

, Optimal restoration of distribution systems using dynamic programming, Power Delivery, IEEE Transactions on23(3) (2008), 1589–1596.

32.

Boroomand

, Fereidunian

, Zamani

M.A.

, Amozegar

, Jamalabadi

H.R.

, Nasrollahi

, Moghimi

, Lesani

, Lucas

, Cyber Security for Smart Grid: A Human-Automation Interaction Framework, In Proc of the IEEE PES ISGT, Gothenburg, Sweden, 2010.

33.

Kazemi

, Firuzabad

M.F.

, Sanaye-Pasand

, Lehtonen

, Impacts of automatic control systems of loop restoration scheme on the distribution system reliability, Generation, Transmission & Distribution, IET3(10) (2009), 891–902.

34.

Mamo

, Mallet

, Coste

, Grenard

, Distribution automation: The cornerstone for smart grid development strategy, pp. , Power & Energy Society General Meeting, 2009 PES ’09 IEEE, 2009, pp.1–6.

35.

Chen

W.H.

Quantitative decision-making model for distribution system restoration, Power Systems, IEEE Transactions on25(1) (2010), 313–321.