Graph Theory-Based Feeder Automation Logic for Low-End Controller Application

Transcription

1 Graph Theory-Based Feeder Automation ogic for ow-end Controller Application Fang Yang, Zhao i, Vaibhav Donde, Zhenyuan Wang, James toupis Abstract---This paper presents a graph theory-based feeder automation logic that is applicable to the low-end feeder automation controller, including programmable logic controllers (PCs) and intelligent electronic devices (IEDs). IEC - standard PC programming languages are used to implement the entire logic engine in oftpc programming environment. This logic engine can provide following functions required by the realtime feeder automation application: () a dynamic update of the system configuration and the load profile and () a generic logic for fault location detection, fault isolation, and power restoration after the occurrence of a permanent fault in the distribution network. The logic engine implemented using PC languages enables the low-end controller to perform complex feeder automation functions and facilitates end users (usually field electrical engineers) to easily understand and customize the logic. In addition, the oftpc technique makes the logic engine development independent of any proprietary PC or IED hardware. The proposed feeder automation logic is demonstrated by example distribution feeder systems. Index Terms Electric power distribution system, feeder automation, intelligent electronic devices (IEDs), fault detection and isolation, power restoration, IEC - PC programming languages, oftpc technology. E I. INTRODUCTION ECTRIC utilities in U often rely on a trouble call system where customers can report outages to the utility. More specifically, when a permanent fault occurs and customers experience a power outage, a customer(s) may call the utility and report the power outage. After receiving the power outage report, the utility may send a crew to the field to investigate the fault location and figure out and implement a switching scheme to first isolate the fault and then restore service to as many impacted customers as possible while the faulty feeder part is being repaired. Many utilities have deployed feeder switching devices (reclosers, circuit breakers, and so on) with intelligent electronic devices (IEDs) for monitoring, protection and control applications. When IEDs are used in connection with network communication and distributed control, feeder automation is enabled. That is, the feeder system operating condition can be monitored and controlled by a feeder automation controller and multiple IEDs that are equipped Authors are with ABB Inc. U Corporate Research Center in Raleigh, NC /9/$. 9 IEEE with switches in the feeder network. The IEDs send system information to the controller and in response the controller executes the feeder automation logic that () identifies a fault location, () isolates the faulty feeder section, and () restores power for the unfaulty out-of-service area. The fault isolation and power restoration control commands are then sent to IEDs, which implement the switch status change accordingly. As a result of this automated procedure, the power outage duration can be reduced and the distribution system reliability can be improved significantly [-]. Based on the information provided by IEDs, automated fault location identification and fault isolation are relatively easy to achieve. In contrast, automated power restoration becomes a challenging task due to operating constraints, load balancing, and other practical concerns. Many automated power restoration algorithms have been proposed in previous literature, such as heuristic search-based techniques [-8], artificial intelligent-based algorithms [9-], analytical-based algorithms [, ], and algorithms combining two or more of these techniques [, ]. Most of these algorithms were developed for planning analysis or executing in distribution control centers to aid system operators with appropriate decisions. These algorithms are not suitable for the real-time feeder automation application that runs in controllers located in substations. In addition, these algorithms were generally implemented in advanced programming environment that provides comprehensive mathematic libraries, data structures, and many other highly developed functions. These features make the development of complex algorithms moderately easy and efficient. However, the logic developed in such environment is usually difficult for end users such as field electrical engineers to understand and customize. Field electrical engineers generally prefer IEC- programmable logic controller (PC) programming languages, such as sequential function chart, function block diagram, structure text, and so on. uch PC languages, however, do not provide many advanced features as offered by advanced programming environment and languages. Therefore, the development of efficient algorithms suitable for PC languages to realize complex feeder automation logic becomes very critical. Based on the characteristics of PC languages, this paper presents a feeder automation logic that applies the graph

2 theory to () dynamically update the system configuration and the load profile and () provide generic logic for fault location detection, fault isolation, and power restoration after the occurrence of a permanent fault in the distribution network. The logic engine can execute in real-time in low-end (as well as high-end) feeder automation controllers located in the substation. The logic engine implemented using PC languages can be easily understood and customized by field engineers. generated based on both the connectivity matrix and the realtime system switch status. pecifically, in an incidence matrix, any entry corresponding to an upstream node is represented by positive one (), and any entry corresponding to a downstream node is represented by negative one (-), as shown in Equation. II. METHODOOGY This section presents the graph theory-based feeder automation logic engine implementation. Figure shows a hardware infrastructure of a centralized feeder automation system, which includes a master controller located in the substation and multiple IEDs associated with switches in the feeder network. The overall feeder automation logic engine flow diagram is shown in Figure, which includes functional modules such as system information collection, update system configuration and load profile, fault location detection, fault isolation, power restoration, and so on. In this work, these functional modules are achieved using the depth-first and breadth-first search strategies in the graph theory and analytical techniques. The implementation of theses functions is illustrated using an example system shown in Figure. This example feeder network includes three substations (ub~), nine switches (switch ~9), and seven loads (~). In the normal operating condition, substation supplies energy to loads ~ and, loads and are supplied from substations and, separately. witches and 8 work as normally open tie switches to maintain the radial configuration of the feeder system, and all other switches are in the close status. A. ystem information collection The system information collection includes polling IEDs for information, obtaining switch status (e.g., open, close, lockout), counter values for reclosers, electric parameters such as currents, voltages, and real/reactive power. Communication between the master controller and slave IEDs can be through various communication protocols such as Modbus, DNP, and so on. B. Dynamic system configuration update The feeder system topology can be represented in the logic with a system connectivity matrix, which inclues the component connection relationship, wherein the rows represent the switches and the columns represent the load/source nodes. In the matrix of Equation for the network shown in Figure, if a switch is connected to a node, the corresponding entry is one (). Otherwise, the entry is zero (). Note that the connectivity matrix does not include the system component upstream and downstream relationship. uch a relationship may change as a switch status changes, For example, a change in the system configuration may change a relative upstream and downstream relationship among components. To reflect such system configuration information, a system incidence matrix is dynamically Figure. Hardware infrastructure of a centralized feeder automation system tart Collect system information Update system configuration and load profile No A permanent fault occurs? Yes Generate isolation logic end isolation control signals to IEDs Generate restoration logic end restoration control signals to IEDs Figure. The flow diagram of overall feeder automation logic engine Figure. An example feeder network

3 witches Matrix ty Connectivi Node = 9 8 () witche Matrix Incidence Node = 9 8 () The generation of such incidence matrix is done based on a depth-first search strategy. Figure depicts a flow diagram illustrating a stack based depth-first search procedure: tep. The search starts from an unexamined source node k, the source node is pushed into a stack vector and the stack top index is increased by accordingly. tep. Find the downstream closed switch i connected to the node k at the stack top, and find the downstream node j connected to this closed switch i. tep. Push the node j into the stack vector and increase the stack top index by, this node j is the downstream node of the previous node k, then change the connectivity matrix entry [ i, j ] from to -. tep. Repeat steps and until no more closed switch can be found. Then pop out the stack top node and reduce the stack top index by. tep. If the stack top index is not zero (the stack is not empty), repeat steps ~ until the stack is empty. tep. Repeat steps ~ until all the source nodes are examined. This stack based depth-first search can be performed without complex vector and matrix operations and can efficiently update the dynamic incidence matrix. C. Dynamic load profile update In this work, a load in the feeder system is represented in terms of the load current magnitude, which can be calculated from the measured switch current magnitudes using a breadthfirst search method. This method can be generalized to calculate the load represented by real/reactive power based on real/reactive power measurements from IEDs. The flow diagram of the breadth-first search technique is shown in Figure, and the search steps are described as follows: tep. A queue vector, a queue head index, and a queue tail index are initialized with zeros. tep. The search starts from an unexamined source node; all closed switches connected to this node are pushed into this queue, and adjust the queue tail index accordingly. tep. Find the downstream load node i that is connected to the switch at the queue head, and find all the closed switches that connected to this node, sum up the currents flowing through these closed switches, and calculate the difference between the current of switch at the queue head and the summed currents, which is the load current at node i. tep. Add all closed switches found in step in the queue tail, adjust the queue tail index, and increase the queue head by. tep. Repeat steps and until the queue head index is the same as the queue tail index, i.e., the queue is empty. tep. Repeat steps,, and until all source nodes have been examined. imilar to the stack-based depth-first search, this queuebased breadth-first search can avoid complex vector and matrix operations and update load profile dynamically. In the example system, the IED measurements of switch current magnitudes are listed in the Table, and the calculated load currents are listed in Table. Figure. The flow diagram of a stack-based depth first search for updating system incidence matrix D. Fault identification and isolation logic The fault location identification and fault isolation logic are generated automatically based on the incidence matrix by searching downstream of the lockout switch and comparing the reclosing counter values before and after the fault. For example, if a permanent fault occurs at load node in the example feeder system, the upstream switch goes to lockout status after a reclosing sequence. Based on the lockout status and the increased reclosing counter value of this switch, the downstream node can be identified as the fault location, In addition, the other switch connected to this load node (switch in this case) can be also identified as the isolation switch that should be opened to isolate the faulty section. The search for the downstream node of the lockout switch, i.e., the fault

4 location, and the isolation switch using the incidence matrix is shown in Figure, and the post-fault isolation system configuration is shown in Figure. As a result of this fault and switch lockout and isolation, loads - and lose their power supply, this area is referred to as the unfaulty outof-service area. Figure. Post-isolation system configuration Figure. The flow diagram of dynamic load profile update Table. IED measurements of switch current magnitudes (Ampere) witch No. 8 9 Current(Ampere) Table. Calculated load current magnitudes (Ampere) oad No. Current(Ampere) 9 9 Figure. The search procedure for the fault location and isolation switch E. Power restoration logic The automatic generation of power restoration logic includes two stages: in stage one, a depth-first search technique is used to search all possible restoration sources/paths; in step, a reverse search procedure is developed using an analytical method to find one or multiple restoration path(s) to restore as much load as possible while balancing the load level in each restoration path. This power restoration logic is illustrated as follows: tage : earch possible restoration sources/paths The search for possible power restoration sources/paths starts from isolation switch. All the downstream nodes and connected switches are searched and stored. The search stops at the normally open tie switches. Based on the post-isolation system configuration shown in Figure, this search procedure will stop at the two normally open tie switches and 8. These two switches lead to two possible restoration sources: substations and. tage : earch one or multiple restoration sources/paths based on system operating constraints Note that in this work only the capacity limits of sources (substations) and switches in terms of current magnitudes are included in the operating constraint set. To simplify the problem, all other operating constraints such as voltages and real/reactive power flow are assumed to be within normal limits after the power restoration. These constraints can be included in the constraint set similarly to determine the restoration sources/paths. Among all possible restoration sources/paths, the equivalent capacity margin (ECM) of each restoration/path down to the normally open tie switch is first calculated. Based on such ECM information, the source/path that has the largest ECM is selected to restore load in the unfaulty out-of-service area. After picking up certain amount of load, the selected source/path may have less ECM than others, then another restoration source/path with the largest ECM is selected instead to continue the restoration work. This procedure repeats until the entire load in the unfaulty out-of-service area is restored or all the restoration sources/paths have run out of their capacities. When a joint node that connects to more than one restoration source/path is encountered, the restoration source/path with highest ECM is selected to continually pick up the load while other restoration sources/paths stop the power restoration just before the joint node. This can

5 guarantee the restoration is performed to the maximum extent for the load at the joint node and beyond it. The power restoration based on this analytical method can restore as much load as possible and meanwhile balance load to each restoration source/path in terms of its available capacity. This load restoration method is illustrated using the post-isolation system shown in Figure. Based on the possible restoration source/path obtained from stage, the ECMs of substation down to the switch (path ) and substation down to the switch 8 (path ) are compared, following cases may occur: Case : The ECM of one substation is significantly larger than the other. For example, the ECM of path is larger than that of path, the switch is assumed to be closed and the switch is assumed to be opened to restore load from substation. After this restoration, if the ECM of path is still larger than the other, then switch is assumed to be closed and switches and are assumed to be opened to restore load. This procedure repeats until the rest of the unsupplied load has been picked up or the path uses up its capacity. If after picking up load, the ECM of path is less than the other, then path is selected instead to restore load by the assumption of closing switch 8 and opening switch. After this restoration, the ECMs of path and are compared, and the larger one is selected to continue the restoration of the load at the joint point () and beyond. Case : If the ECMs of path and are about the same, then each of them starts to pick up the load respectively, when they reach the joint node of, the path that has larger ECM is selected to continue the load restoration. If case occurs, and path has larger ECM after paths and restore loads and, respectively, path continues the restoration of loads and. o the finalized restoration plan is to open switch, and then close switches and 8. The post-restoration system configuration is shown in Figure 8. automation logic. This system includes six substations (ub~ub), seventeen switches (R~R), and twelve loads (~). In the base case configuration, loads ~,,, and are supplied from ub, load 9 is supplied from ub, load 8 is supplied from ub, load is supplied from ub, and load is supplied from ub. witches R, R9, R, R, and R are normally open tie switches. The measured switch current magnitudes (in terms of ampere), pre-defined switch capacity limits, and the load current magnitudes are shown in Figure 9. Figure 9. An example feeder system When a permanent fault occurs to load, switch R goes to a lockout status after a reclosing sequence, and switch R is determined as the isolation switch which is opened to isolate the faulty section of this feeder circuit. The postisolation system configuration is shown in Figure, in which loads ~,,, and in the unfaulty out-of-service area are no longer served from ub. Figure 8. Post-restoration system configuration By using PC programming languages, each of the functional modules implemented using structure text can be packaged in one or more function blocks, and the program flow is controlled by the sequential function chart. The sequential function chart and function blocks facilitate end users to understand the overall logic engine and allow the logic to be further customized. III. CAE TUDY In this section, a more complex feeder system shown in Figure 9 is used to demonstrate the proposed feeder Figure. Post-isolation configuration of the example system The power restoration logic is then executed. This restoration logic considers both substation and switch capacity limits. All substation capacity limits are assumed to be amp, and all switch capacity limits are shown in Figure 9. The first stage of the power restoration logic is to

6 search all normally open tie switches starting from isolation switch R, related sources, and intermediate switches. The search results are listed in Table. Table. All possible restoration sources, related tie switches, and intermediate switches Normally open tie R R9 R R switch Related sources ub ub ub ub Intermediate switches R R NA R ECM 9 8 In the second stage of the power restoration procedure, the ECMs of each restoration source/path down to the normally open tie switch are calculated and listed in Table. After ub picks up load, its ECM becomes amp, and after ub picks up load, its ECM becomes amp. By comparing the ECMs of the sources ub, ub, and ub, which have a joint node (), the ECM of ub ( amp) is the highest among the three sources. Therefore, resource ub is selected as the source to restore load at and beyond the joint node. The restoration solution then includes () open switch R and () close switches R and R. After source ub picks up loads and, its ECM becomes amp, and after source ub picks up load, its ECM becomes amp. As ub and ub have a joint node (), and ub has larger ECM than ub, ub is selected to continue to restore power to loads and. The restoration solution obtained is to () open switches R and R and () close switches R9, R and R. The post-restoration system configuration is shown in Figure, in which load is restored by ub, loads,, and are restored by ub, and loads,, and are restored by ub. We assume that all other operating constraints are within limits after the restoration. All the unserved loads in the unfaulty out-ofservice area are restored and these loads are balanced to each restoration source/path. Figure. Post-restoration configuration of the example system IV. CONCUION A graph theory-based feeder automation logic is developed using IEC- PC programming languages. This logic engine has following features: () it is applicable to low-end (as well as high-end) controllers located in distribution substations to execute real-time feeder automation application, () it can dynamically update the system configuration and the load profile and provide generic logic for fault location detection, fault isolation, and power restoration, () implemented using PC languages, the logic engine is easy to be understood and customized by field engineers. Further improvement can be done in the following two aspects: () more work can be done to improve the representation of the system configuration without using the matrix form and () advanced power restoration scheme can be developed to include more practical concerns such as minimizing switch actions and so on besides load balancing. V. REFERENCE [] [] G. Ockwell, Implementation of Network Reconfiguration for Taiwan Power Company, IEEE PE General Meeting,. [] D.M. taszesky, D. Craig, C.Befus, Advanced Feeder Automation Is Here, IEEE Power & Energy Magazine, ept./oct.. [] J. Fan, X. Zhang, Feeder Automation within the cope of ubstation Automation, Power ystem Conference and Exposition, Nov.. [] V.. Devi, and G. Anandalingam, Optimal Restoration of Power upply in arge Distribution ystems in Developing Countries, IEEE Transactions on Power Delivery, Vol., NO., January 99. [] J.. Wu, K.. Tomsovic, and C.. Chen, A Heuristic earch Approach to Feeder witching Operations for Overload, Faults, Unbalanced Flow and Maintenance, IEEE Transactions on Power Delivery, Vol., No., Oct. 99. [] T. Taylor and D. ubkeman, Implementation of Heuristic earch trategies for Distribution Feeder Reconfiguraiton, IEEE Transactions on Power Delivery, January 99. [8] Y. Moon, B. Cho, H. Park, H. Ryu, B. Ha, and. im, Fault Restoration Algorithm Using Fast Tracing Technique based on the Tree-tructured Database for the Distribution Automation ystem, IEEE PE ummer meeting,. [9] C. iu,. J. ee, and.. Venkata, An Expert ystem Operation Aid for Restoration and oss Reduction of Distribution ystems, IEEE Transactions on Power ystems, Vol. No., May 988. [] C. Huang, Multi-objective ervice Restoration of Distribution ystems Using Fuzzy Cause-Effect Networks, IEEE Transactions on Power ystems, Vol. 8, No., May. [] Y. T. Hsiao and C. Y. Chien, Enhancement of Restoration service in Distribution ystems Using a Combination Fuzzy-GA Method, IEEE Transactions on Power ystems, Vol., Nov.. [] Y. Ke, Distribution Feeder Reconfiguration for oad Balancing and ervice Restoration by Using G-Nets Inference Mechanism, IEEE Transactions on Power Delivery, Vol. 9, No., July. [] J. Wu, A Petri-Net Algorithm for Multiple Contingencies of Distribution ystem Operation, IEEE Transactions on Power ystems, Vol., No., August 998. [] Y. Ke, Distribution Feeder Reconfiguration for oad Balancing and ervice Restoration by Using G-Nets Inference Mechanism, IEEE Transactions on Power Delivery, Vol. 9, No., July. [] Q. Zhou, D. hirmohammadi, W. E. iu, Distribution Feeder Reconfiguration for ervice Restoration and oad Balancing, IEEE Transactions on Power ystems, Vol., No., May 99. [] K. N. Miu, H. D. Chiang, B. Yuan, G. Darling, Fast ervice Restoration for arge-cale Distribution ystems with Priority Customers and Constraints, IEEE Transactions on Power ystems, Vol., No., August 998. [] C. Chen, C. in, and H. Tsai, A Rule-Based Expert ystem with Colored Petri Net Models for Distribution ystem ervice Restoration, IEEE Transactions on Power ystems, Vol, No., Nov..

7 VI. BIOGRAPHIE Fang Yang (M ) joined ABB Corporate Research in Raleigh, North Carolina in, where she is currently a r. R&D Engineer. Her research interests include distribution automation and voltage/var optimization, power system reliability assessment, the application of artificial intelligent techniques in power system control, and the application of microprocessor technique in power system monitoring, protection and control. Zhao i (M 8) joined ABB Corporate Research in Raleigh, North Carolina in, where he is currently a oftware Architect. His research interests include the application of software technologies in process automation and power systems, performance analysis, and information system design and tuning. Vaibhav Donde (M ) joined ABB Corporate Research in Raleigh, North Carolina in, where he is currently a Consulting R&D Engineer. Prior to joining ABB, he had a postdoctoral appointment at awrence Berkeley National aboratory (-). He holds Ph.D. () and M.. () degrees, both in electrical engineering from the University of Illinois at Urbana-Champaign and a B.E. degree (998) in electrical engineering from V.J.T.I., Mumbai, India. He has worked with TATA Consulting Engineers, Mumbai ( ). His technical interests include power system analysis, modeling and simulation, power distribution systems and automation, hybrid dynamical systems and nonlinear control. Zhenyuan Wang (M ) joined ABB Corporate Research in Raleigh, North Carolina in, where he is currently a Principal Consulting R&D Engineer. His research interests include electric power equipment condition monitoring/assessment/diagnosis, system monitoring, control and automation for a smart grid. His experiences include asset management IT applications in the electric power industry, power system transient analysis, substation/distribution automation, and data integration/warehousing/mining applications. James toupis (M 99) is a Principal Consulting R&D Engineer in the Power Technologies Department for ABB s U Corporate Research Center located in Raleigh, North Carolina. Jim has been employed at UCRC for years, and his research has been focused in the areas of distribution and feeder automation, wireless communications, power system protection and control, and event detection and classification.