Abstract
Abstract
Considering the fact that the main part of electrical energy losses is related to the distribution level; therefore, controlling and monitoring of this part of the network is necessary to reduce energy losses which is made possible through distribution automation. Distribution automation systems need decision making regarding the responsibilities being performed by operators. On the other hand, there is the possibility of incorrect decision making by operators due to the large volume of information and decision making time shortage during automation steps which can be created by incorrect information received from the network and consequently causes the incorrect control command sending to distribution network devices. Also, introducing the Distributed Generations (DGs) in the protection system of distribution network, it is possible for the protective devices to have an incorrect operation which affects the operators’ decisions performed by the automation system. In this paper, a new Petri-Net modeling method is used to make the distribution network monitoring and automation possible and also to help the operator in correct decision making. In addition, the protective devices’ operation is determined more reliably using the attained data from the automation system.
Introduction
Nowadays, increasing penetration of DGs in electrical energy distribution network development causes complexity and difficulty in distribution automation monitoring. Large numbers of feeders in addition to application of several DGs show that traditional manual monitoring cannot be implemented correctly in presence of these complexities and hence correct decision making is not possible during automation steps. Incorrect control command sending, because of false identification of operators in control center, causes incorrect switching and decreases the reliability of electrical energy distribution systems [1]. Therefore, a software system to help the operator for correct decision making is essential because of high impedance faults which are not identified by protective devices and increase in information volume in the control center and also lack of time for decision making.
So far, different methods have been proposed for electrical system monitoring in several investigations. In a paper [1], a Petri Net model of protective device operation and identification is presented which performs automation system monitoring using fault indicator alarms and protective relay alarms. Considering DG presence in the system, these alarms are more and more affected and consequently the monitoring action will be affected [2]. Nevertheless, the proposed Petri-Nets model has a shortcoming in monitoring. For example, if no fault occurs (absence of token in the places corresponding to indicators I1 to I5), this model shows fault existence in section 1, incorrectly. A method based on Petri-Nets is proposed for distribution networks’ protective system monitoring in presence of DG and the over current protective relay fault identification is performed using Petri-Nets as a case study [3, 4, 3, 4]. Fault current and protective devices’ coordination curves are used, but the procedure implemented for fault current in the models is not clearly determined. In addition, for correct fault identification by the proposed model, complete information of fault occurrence in line (token in place p 4) should be available. This actually contradicts the fault identification goal that is based on the information of control center. Herewith, it is not clear that fault has occurred because of the information of control center or the operator has made a mistake due to receiving incorrect information which is caused by unreliable information or lack of protection coordination. In [5] authors focused on ring distribution network protection based on Petri-Nets in presence of DG. The goal was to propose a backup protective structure for DG in network in which Petri model is used for correct DG protection monitoring. In this work, the authors do not pay attention to coordination procedure of protective devices after adding the structure, and only a model of protective system operation procedure has been presented using Petri-Net. Also, they do not discuss how the control center receives information, and the proposed models are for particular case studies which do not have enough accuracy and effectiveness. None of these investigations has proposed a complete model of network monitoring and identification procedure for incorrect protective devices operation in presence of DG.
Here, applying the modeling based on Petri-Nets, the distribution network automation monitoring in presence of DGs is provided to help operators to make correct decisions. In this method, using the attained data from automation system including substation voltage and distribution feeders current, protective devices’ operations are determined with high accuracy and reliability.
Petri-Nets
Petri-Nets (PN) theory was created by Carl Adam Petri in 1962 [6]. Basically, Petri-Nets are introduced to illustrate logical relation and different presentations of concurrent sequential actions in discrete event dynamic systems. The Petri-Nets are ideal tools to analyze and model the hierarchy and sequential behavior of different systems such as computer systems, industrial systems, and power systems. In the following subsections, Petri-Nets theory is introduced, briefly.
The structure of Petri-Nets
A Petri Net is composed of four parts: places (P), transitions (T), input function (I) and output function (O) [7]. The input and output functions are described based on the places and transitions. These functions are bridge links between the transitions (T) and places (P). The structure of a Petri-Net such as C is defined as C = (P, T, I, O) where the set of places and transitions are expressed as relations (1).
Where p i indicates the ith place and t j indicates the jth transition.
P ∩ T = Φ: The set of places and transitions which are separate.
I : T → P ∞ is the input function, a mapping from transitions to the bags of places which is specified as input places to transition and P ∞ determines the bags of places.
O : T → P ∞ is the output function, a mapping from transitions to the bags of places which are specified as output places from transitions.
Two types of nodes are defined in the graph of a Petri-Net: A circle indicates a place and a rectangular expresses a transition. Also, by a number of directed arcs (arrows) from the places to transitions (input places of transition) and other directed arcs from the transitions to places (output places of transition), the places and transitions are connected to each other. A simple graph of Petri-Net has been shown in Fig. 1.
In Fig. 1 p 1, p 2, p 3, p 4, and p 5 are the places and t 1, t 2, and t 3 are the transitions. The black points in places p 1 and p 4 are tokens. The structure of Petri-Nets is static, and its dynamic characteristics are determined by the firing of transitions and the movement of tokens. The firing of transition transmits the tokens from the input places of transition to its output places [6–10].
The structure of Petri-Nets and the process of firing of transitions not only can be shown by graphical methods but also can be expressed and analyzed through matrix operations. For this goal, we need to construct the main matrixes consist of the structure matrix C, the marking vector M, and the transitions firing vector U. Complete information about the procedure and Petri-Nets analysis by matrix operation is available in [7–16].
Distribution automation and DG effect
Automation system
According to the IEEE standard, the distribution automation system is a system that makes distribution system capable of remote monitoring, coordination distribution devices, and imposing controlling command on them in real time from long distances. Therefore, information receiving, system analyzing, and sending connect and disconnect control commands to switches and processors should be done rapidly and without fault. Reducing the outages time, energy losses, and the number of switch operation in addition to essential analysis, power flows, presentation of the network parameters’ information, and the faults (for decision making) are the outcomes of a proper system automation. Figure 2 shows an automated distribution feeder test system.
This system consists of two common feeders with several line switches which are used to divide feeders into different parts. These line switches are controllable by remote terminals’ units (RTU) which are installed in line switches’ locations [17]. Since Petri-Nets are used in real time applications, they will be very useful for decision making in the automation system.
The effect of DG in distribution network protection system
Here, two problems have been discussed: protective blindness and incorrect trip which are caused by the effects of DG entrance to distribution network. According to Fig. 3, when a fault occurs in the feeder which DG is connected to, DG providing part of the fault current, the network contribution in fault current decreases, and it causes protective blindness in the distribution network [2–4]. The network contribution decreasing in the fault current when it becomes lower than the current setting of the feeder relay does not react to the fault occurrence. According to Fig. 5, if the fault current is lower than the threshold current of the feeder relay (1000 ampere), it does not react.
Also, the incorrect trip occurs when DG connected to the feeder contributes to the fault current of another adjacent feeder which is connected to a substation [2–4]. This situation is shown in Fig. 4. In Fig. 4, with regard to Fig. 5 curve, if relay R2 current is greater than its threshold current (1000 ampere), it will do unnecessary trip.
Distribution automation monitoring and protective system
The goal of distribution automation monitoring especially in protection is to identify devices with incorrect action or ones that do not act where they should act based on the received information from the distribution grid and automation system. In other words, it must create a kind of software protection through communication system. This needs the analysis of network’s voltage and current information in addition to a fast real-time decision maker. The presented technique in [18] can be used to analyze network information. Since monitoring devices are installed in the automated distribution network, the operator uses them for decision making in the distribution network. In other words, the operator uses the operational information of protective devices received from the automation system and also parameters such as voltage and current to identify the network disturbances and make decisions. It is probable that the operator makes mistakes in analyzing information or identifying their incorrectness (like DG’s effect on protection system’s operation that distorts the received information) and therefore makes a wrong decision. In such condition, utilizing a software method to protect and monitor the distribution network is necessary. Hence, here, we present a method to protect and monitor the distribution network using the proposed technique in [18]. This method is implemented based on terminal voltage obtained by the PQ measurement device. Now, using this technique and Petri Nets, we can monitor the operation of the protective system and distribution automation system to recognize their wrongperformance.
Figure 6 shows the flowchart of the modified technique presented in [18] (Part I) and Petri Net models of fault type identification (Part II) and monitoring (Part III) which are separated with dashes where the Petri Net inputs are terminals’ measured voltage angle variation and modal transforming which are used to determine the line-to-earth fault. Also, the measured current of each feeder leaving the substation is used to determine the correct operation of primary relays on that feeder. In Petri Net shown in Fig. 6, all linking arcs among the places and transitions have the weight of 1. Places Im, I, PASa, PASb, PASc, GR, and UNGR are the input places of the Petri Net model. Place Im has token when Imax1 (the maximum current of one of the phases of feeder 1) is more than Imax2 (the maximum current of one of the phases offeeder 2).
Place I has token when Imax is more than Ith (the threshold current of feeders’ relays) if Imax2 is less than Ith and vice versa. Places PASa, PASb, and PASc have tokens when the real value of these quantities is more than a specific number (here for the studied distribution network is determined 0.08 after trial and error). This number is to distinguish between faults and load variations which are determined based on trial and error. Also, the presence of token in places GR and UNGR through modal transforming of the measured voltages and determining V0 are as [18]. The presence of token in places AG, BG, CG, ABG, BCG, ACG, AB, BC, AC, and ABCG show the type of occurred faults, respectively. Place p17 is just an auxiliary place with no physical concept and has token when at least one kind of fault has occurred in the network. Place p18 has no physical concept and just helps in distinguishing the faulty feeder. Place “No Fault Feeder” shows the initial conditions of feeders and no token in it, means no fault in feeders. The presence of token in place “DGs” shows the effect of distribution generation, and the presence of token in it means fault occurrence on the feeder. The presence of token in place “Relays” indicates the relays’ correct operation. Places Feeder1 and Feeder2 show faults on feeder 1 and feeder 2, respectively. Place “Send Signal Trip” shows the sending trip command from the protection software. Transitions t1, t2, … , t10 do not have physical concept and are used only for communication among places and distinguishing types of faults according to Petri Nets. Transition t11 indicates the effect of fault on feeder. Transitions t12, t13, t14, t15, and t16 do not have physical concept and are utilized for communication between inputs and the monitoring model based on the principals of Petri Nets. There are two types of nodes in Petri-Nets: places and transitions; and no places or transitions can be directly connected. In Fig. 6, each of PASa, PASb, PASc, GR, UNGR places are independently related to transitions t1, t2, … , t10, and none of t1, t2, … ,t10 are connected to each other. To avoid the model complication, from each PASa, PASb, PASc, GR, UNGR places, one output is connected to all transitions t1, t2, … ,t10, so each path should be considered independently for example an independent path from place PASa to t1 and another independent path to t2. Some of the places are connected to transitions with small circles which indicate NOT which create enabling conditions for that transition if there are no tokens there.
Case studies
After the simulation of the network in Fig. 2 in DIGSILENT software and creating different types of faults, the proposed method has been studied. In this network, two DGs in feeder 2 have been considered, and the network’s information is given in appendix-A [19]. Here, the proposed method is analyzed by three scenarios. In section 6 of feeder 2 in Fig. 2, a single phase short circuit (phase A to earth) is created (fault F1) where the fault current is 1517 ampere. The current flows through relay 2 in which the network contribution in the fault current is 1038 ampere, and DGs contribution is 479 ampere (DG1 contribution is 259 ampere, and DG2 contribution is 220 ampere). With regard to Fig. 5 curve, the relay current is lower than its setting current, and therefore the relay does not have any reaction to this fault. In section 1 of feeder 1 in Fig. 2, a three phase short circuit is created (fault F2) in which the fault current is 6592 ampere. The current flows through relay 1 in which the network contribution of the fault current is 5465 ampere and DGs contribution is 1127 ampere (DG1 contribution is 730 ampere, and DG2 contribution is 397 ampere) which is passed through relay 2. With regard to Fig. 5 curve, both relays will react to this fault in which the reaction of relay 2 is unnecessary. In Section 3 of feeder 2 in Fig. 2, a phase to phase to earth short circuit (phase B and phase C to earth) is created (fault F3) that the fault current is 4056 ampere. The current flows through relay 2 in which the network contribution of the fault current is 3045 ampere, and DGs contribution is 1011 ampere (DG1 contribution is 611 ampere, and DG2 contribution is 400 ampere), with regard to Fig. 5 curve; the current of relay 2 is greater than its setting current, so the relay will react to this fault.
By using DIGSILENT software, three-phase voltage of substation and three-phase current of two feeders for one of scenarios 1, 2, and 3 have been determined in three cycles and transferred to MATLAB software as inputs of the proposed method. The sampling frequency of aforementioned signals in MATLAB software is 20000 HZ. The results for the above three scenarios are determined as Table 1. With regard to Table 1, the situations of input places to Petri Nets Model will be according to Table 2.
With regard to the situation of input places of Table 2, the type of fault for scenario 1 and the fault of diagnostic model (part II) is determined as Phase A to earth fault; therefore, due to the monitoring model of the fault existence in feeder 2 and absence of token in relays’ places and the token presence in DGs place show the weakness of protective system’s operation.
Therefore, the fault occurrence in feeder 2 and the lack of sensation of the fault by relay 2 due to DG effect has been determined as shown in Figs. 7–11.
With regard to the situation of input places of Table 2, the fault type for scenario 2 is determined by the fault diagnostic model (part II) as three-phase fault, and according to the monitoring model, the fault existence in feeder 1, the absence of token in relays’ places, and the token presence in DGs’ places show the weakness of the protective system’s operation, here, showing the unnecessary sensation of fault by relay 2 due to DG effect based on Figs. 12–16.
Regarding the situation of input places, the situation of Table 2, the fault type for scenario 3 is determined by the fault diagnostic model (part II) as phase to phase to earth (BCG). Based on the monitoring model, the fault existence in feeder 2 and the token’s absence in relays’ places and the token’s absence in DGs places show the correct protective system operation which here show the operation of relay 2 for fault clearing according to Figs. 17–21.
Only in scenario 3, receiving the protective operation information from the automation system is correct while in scenarios 1 and 2 because of DG effect it is not correct.
Conclusion
Monitoring and control of electrical distribution networks are essential for optimal operation of distribution networks which need automation systems. Distribution automation systems need decision making by operators in order to perform their tasks. Especially, due to increase in distribution generations’ (DGs) penetration in the distribution network, the feeder is not radial which means that the protection system cannot properly fulfill the previous tasks. Accordingly, it is desirable for the protection system to have an efficient performance in the presence and absence of distributed generation’s resources since it needs more flexibility of the protection system. In this paper, a new method using Petri-Nets has been proposed for automation systems’ monitoring to help operators using the substation’s voltage and the information of feeders’ currents to identify the correct performance of the automation system. The proposed approach brings about more reliable and accurate steps of automation performed by operators.
Appendix A
The distribution network’s loads and lines (Fig. 2) information are according to Table 3 and two DGs’ information is shown in Table 4.