Analysis of Data Quality Issues in Wide Area Monitoring and Control Systems

Transcription

1 2010 IREP Symposium- Bulk Power System Dynamics and Control VIII (IREP), August 1-6, 2010, Buzios, RJ, Brazil Analysis of Data Quality Issues in Wide Area Monitoring and Control Systems Kun Zhu, Moustafa Chenine, Johan König and Lars Nordström Industrial Information and Control Systems Royal Institute of Technology, KTH Sweden Abstract Synchronized phasor measurement based Wide Area Monitoring and Control (WAMC) system is becoming a reality with international research and development both in academia and industry. Timely and accurate data with high resolutions holds great promise for more responsible and advanced grid control and operation. Currently, most of the research focuses on the different control schemes and applications. A relatively less addressed aspect is the dependency of the WAMC system on the performance of the Information and Communication Technology (ICT) infrastructure, without whose support the projected functionalities of the WAMC systems will not be achieved. Possible input delay brought by the complex data transfer and processing processes in WAMC systems are presented at the beginning of this paper. Thereafter, simulations where delayed phasor measurements are fed to a typical WAMC application - Static Var Compensation (SVC) are conducted iteratively to detect its maximum tolerated delay. Based on this maximum tolerated delay, the consequences of missing certain data in the phaor packets have been studied. Furthermore, data quality requirements to design a reliable SVC function are analyzed based on simulation results. In conjunction with the requirement discussion, the paper is concluded by proposing a robust ICT architecture to mitigate the aforementioned data quality issues. Index Terms WAMC systems, ICT infrastructures, PMU, phasor measurement delay, incompleteness, remote signal based SVC I. Introduction There is an international interest and implementation drive, in both academia and industry, on the prospects of Phasor Measurement Unit (PMU) based monitoring and control technology [1] [2]. These systems promise to offer more accurate and timely data on the state of the power system increasing the possibilities to manage the system at a more efficient and responsive level and apply wide area control and protection schemes [3]. Such systems are needed in the modern electrical power system, where transmission expansion is limited by monetary and environmental regulations. The high dependency on the aforementioned systems and their ability to provide correct functionality put very high requirement on the supporting ICT infrastructures to ensure the quality of service. A. Purpose The major purposes of our work are to demonstrate that insufficiency of ICT systems runs the risk of jeopardizing system function execution and elicit requirements on data quality aspect for WAMC system. In this paper, the remote signal based SVC function is studied as a WAMC application case. The consequence of delayed and uncompleted input data for this SVC case is demonstrated and the corresponding requirements are analyzed. Furthermore, ICT architecture is suggested for this specify control scheme to enhance the robustness of the WAMC system. B. Outline The rest of the paper is structured as follows: Section II presents an overview of WAMC system architectures outlining the main components of such systems. Section III the possible delays in WAMC communication. Section IV describes the SVC simulation models implementation and Section V explains the parameters in simulation scenario. The analysis of simulation results and a discussion can be found in Section VI and the paper is concluded in Section VII. II. Wide Area Monitoring and Control Systems A. WAMC system components A WAMC system includes 4 basic components: PMUs, PDC, PMU application systems and finally communication network [4]. Fig. 1 illustrates the logical architecture of WAMC systems.

2 are taken into consideration in this paper. The communication infrastructure in this part has been simplified as illustrated in Fig. 2. Fig. 1: Layers and components of WAMC system. The WAMC system interface with the power system at Layer 1 through the PMUs connected to substation bus bars or power lines. Layer 2 is known as the Data Management layer and this is where the PMU measurements are collected and sorted by the PDC producing a single time synchronized dataset that is forwarded to the applications systems on Layer 3. The application systems represent the real time PMU based application functions that process the time synchronized PMU measurements provided by the previous layer. The information transactions between these three layers are realized by a communication network. B. Communication network in WAMC systems The communication infrastructure is a critical component in the architecture of WAMC systems. This is because PMU devices in most cases are geographically distributed over a territory. The PMU devices are connected to a central control center or several control centers over a communication network depending on the operation concept. Therefore the communication network is a possible bottleneck in the architecture of these systems, since the data quality of PMU measurements collected from the remote sites would highly depend on the communication infrastructure s capabilities and architecture. III. Delay in WAMC Systems Real time WAMC systems require stringent time requirements in order to indentify, analyze and respond to emergency phenomenon in the power system, according to [4] and [5]. The delay of the WAMC system can be traced back to the involved processes and components according to the system architecture presented in the previous section. Specifically, data propagation, PDC sorting, calculation for decision making and the control actuation processes Fig. 2: Utility communication model, showing substation LAN and control center LAN connected by Wide Area Network (WAN). The PMUs or controllers are assumed to reside in substations, connected to a Local Area Network (SS LAN). The SS LAN is in turn connected via a substation router to a Wide Area Network (WAN). The application that utilizes the PMU data resides at a central location connected to a Local Area Network (CC LAN). Since the application is dependent on data from several PMUs the data from the PMUs are first passed to the PDC for sorting after which they are sent to the application. Thereafter, the actions generated by application are transferred via the WAN to the substation where the controllers are located. It is necessary to clarify, that depending on the needs from the applications, the PDC and Applications do not necessarily to be implemented at the control center. A. PDC sorting function The most basic function of the PDC is to collect or receive phasor measurements from connected PMUs and to sort them according to the GPS time stamp. Once a time stamp set, i.e. all phasors with the same time stamp, is complete the PDC forwards the phasor set to the applications consuming the data. The PDC can also have other functionality, such as error checking and archiving for offline and historical data analysis [6]. In this paper only the time synchronization functionality is considered. Generally, there are no specified published algorithms and most vendors consider the algorithms as confidential. On the other hand, the main time synchronization algorithm can be derived from descriptions [7] and the actual requirement of the sorting and synchronization task. In this generic algorithm the PDC will group together measurements from the same time stamp in to a set. This is done as the measurements arrive to the PDC. To minimize delays a time-out per time stamped buffer is added. The time-out would be the amount of time the buffer is actively waiting for the rest of the phasor 2

3 measurements with the same time stamp. The countdown to the time-out is initiated when the first phasor measurement of a certain time stamp arrives at the PDC. The PDC assigns this newly arrived measurement to a new buffer, and begins the time-out counter for that buffer. In this case when the time-out is up the PDC will forward the set without waiting for the rest of the phasor measurements to arrive. The time out parameter introduces a ceiling in terms of the delay that can be experienced and is more deterministic. If the PMU communication network were to experience abnormal delays or packet loss then the waiting time parameter insures that the PDC forwards the phasor measurements in an acceptable time range without waiting for the delayed measurements to arrive. On the other hand, the issue of incomplete data arises where the phasor measurements forwarded to the applications would be incomplete and missing certain measurements from PMUs on the grid. Fig. 3 below illustrates this algorithm. T PDCi = f(t TO,T WANi ) Where T TO represents the time-out parameter in the PDC sorting algorithm, and T WAN represents the data propagation delay from the PMUs to the PDC [8]. In the implemented algorithm the PMU packet with the longest transmission time, T WANi, determines the maximum time that all packets will have to wait T w = Max (T WANi ) i=1..n Finally, this gives the complete expression for T PDCi. T PDCi = T sort +Min(T w T TO ) To study the characteristics of T PDCi in different system architectures and with different settings for T TO in networks experiencing different amounts of delay a simulation study as described in the following section was performed. IV. SVC Test Bed Implementation This chapter describes the test platform implementation, including the reference network and the WAMC system components. MATLAB SIMULINK is used as platform for the power system simulation part in this paper. The communication delays and incompleteness values obtained from simulations implemented in the OPNET communication systems simulation environment. The OPNET simulations studied the ICT aspects of WAMC systems, specifically the communication network and the PDC sorting algorithm, the details could be found in [8]. A. Reference network Fig. 3: Time stamped buffers in the PDC holding measurements in a linked list representation. B. Data propagation time and PDC delay With an assumption that the transmission delay in modern high speed LAN is negligible in comparison to the delay experienced in a WAN, the data propagation time can be treated as equal to the transmission time through WAN including routing delays, T WAN. In the PDC sorting algorithm described in the previous section, the amount of time a packet is delayed in the PDC depends on the following parameters: When it arrives (early packets have to wait until all packets from the same timestamp have arrived) How many PMUs are sending (more PMUs to wait for) How long to wait for the last PMU packet for that time stamp The time-out setting. This linked-list sorting algorithm can be described as: Four-machine two-area system from Kundur is used as reference network model. The model consists of two fully symmetrical areas that are connected by two parallel 230 kv transmission lines. Two identical round rotor generators are located in each area respectively. The network model and its detailed parameters can be found in [9] and [10]. B. Remote signal based SVC A Static Var Compensator (SVC) is deployed for voltage regulation controlling the reactive power injections at the regulated bus. There are different models to describe SVC dynamic performance. In these simulations, a simple Automatic Voltage Regulator (AVR) model is adopted and its scheme can be found in [11]. The SIME method is applied here to transform the trajectories of this four-machine power system into the trajectory of a single machine equivalent system [12]. In this case, Gen1 and Gen2 are assumed to be critical machines that are responsible for loss of synchronism, since the power is flowing from area 1 to area 2 in steady state. POD signal and voltage measurement from the 3

4 regulated bus are the inputs to this SVC. There is one PMU placed at each generator bus to collect rotor speed ω and acceleration power P a, the mismatch between the mechanical powers that turbine takes in and the generated power. All ω and P a from 4 generators are used to compute POD signal. The voltage at the regulated bus is measured by another PMU. V. Simulations scenarios and Parameters In WAMC systems, there is a centralized computation for decision making which can be either integrated with the actuator or located remotely at the control centers. This implies all the involved data has to be uploaded to a remote centralized point and then sent back as orders to the controlling equipment all over the network. Unlike in the local control in the distributed system, the delay in wide-area power systems could reach a significant number depending on the ICT system architecture and implementation [13]. Thus, it is very important to take performance of the supporting ICT infrastructure into consideration in design and implementation of a WAMC system. A. Simulation parameters for the ICT system In the simulation, the SVC control actuation is based on the information from 4 PMUs placed at generation buses together with 1 PMU located at bus 9. The conceptual layout of the ICT infrastructure is illustrated Fig. 4: B. PDC delay scenarios and parameters The 4 PMU scenario simulation results from [8] are adopted. Two parameters where varied in order to create a set of scenarios based on these variations. The parameters that were varied in the ICT system scenarios were: The delay experienced by measurements (from the PMU) traveling to the PDC (T WAN ). The time-out (T TO ) that determines the maximum waiting time at the PDC. The main parameter that is varied in the simulation is the time-out time. Table 1 lists the values chosen for the PDC time-out. Table 1 : Time-out (T TO) parameters used for the PDC model T TO time-out parameters (milliseconds) The time-out time was used in conjunction with two main delay parameters. The delay parameters are used in the simulation to introduce delay in the transmission of the phasor measurement packets from the PMU to the PDC. The delay parameter is a mean and variance input to a normal probability distribution function which generates random values from the distribution. These values are assigned to the every packet that is sent to the PDC. Table 2 below illustrates the two delay parameters used in the simulation. Table 2: Normal Delay Distribution Parameters Mean Variance Base Fig. 4: ICT system architecture for WAMC system in the simulation, including information flows within the processes. The phasor data are collected from the PMUs which are geographically scattered and transferred via the Wide Area Communication network to a PDC. PDC sorts the data with the same time stamp into a packet and forward the packet to App where the centralized coordination and computation are executed. The generated control signal is then sent to the SVC controller through the Wide Area Network. The Base mean and variance are selected from previous work aimed at studying transmission delay in communication networks for wide area monitoring and control systems [14]. This delay parameter is selected to be the standard or base in the simulations. The mean and variance were calculated for usage, as values to represent the case of abnormal delays that could be experienced in high speed power system communication networks. The extended delay was applied to a specific number of PMUs and not to all PMUs in any given scenario. This was done to represent the possible delay only from a subset of PMUs that could have been a result from various network or hardware conditions, such as increased traffic on the network segment where the PMUs with extended delay are placed. Finally, using the combination of delay and time-out parameters, the number of PMUs on the network was varied. This resulted in 12 scenarios with 4 PMUs. 6 4

5 scenarios each with a value from the time-out parameters outlined in Table 1, and the delay parameter with a normal distribution with the Base values in Table 2. The other 6 scenarios had the same parameter expect that one of the PMUs was delayed with the extended normal distributed delay using the delay parameters. Table 3 below illustrates this configuration. Scenario Table 3: 4 PMU scenario set parameter settings Delay Parameter Time-Out Parameter (T TO ) number of PMUs with Delay 1 Base 15ms 0 2 Base 25ms 0 3 Base 30ms 0 4 Base 35ms 0 5 Base 45ms 0 6 Base 50ms ms 1 25ms 1 30ms 1 35ms 1 45ms 1 50ms 1 Voltage at Bus 9 (p.u) Time (s) No time delay time delay of 47.5ms time delay of 69ms Fig. 5: Voltage profiles at the regulated bus given delayed phasor data from all PMUs It can be observed in Fig. 5 that the oscillation grows as the delay increases. When the delay reaches 69ms, the oscillation cannot be cancelled by SVC. Therefore, for this specific network and WAMC system, the maximum tolerated delay is 69ms which puts the latency requirement supporting ICT system. To investigate robustness for POD signal and voltage measurement to their corresponding delay, the following simulation are conducted regardless of the ICT architecture presented in the previous sections. It is assumed that only the POD signal calculated by measurements from 4 PMUs at generator buses is experiencing the delay while voltage measurement is perfect in time. The simulation results are in Fig. 6: 1.15 C. Simulation parameters for the disturbance A three-phase solid ground fault is introduced at the transmission corridor at the time t = 1. And the clear time t c here is set as 200ms, which means at the time 1.2 the transmission line is reclosed. VI. Analysis Iterative simulations are conducted applying increasing delays until the SVC function fail. Thereafter, the performance requirement to design a robust remote signal based SVC control; in this case the maximum tolerated delay could be decided as assumed in Section V. A. Delayed data Voltage at Bus 9 (p.u) Time (s) No time delay time delay of 47.5ms in POD Fig. 6: Voltage profiles at the regulated bus given delayed POD signal and perfect voltage measurement. It can be observed that the system even experienced a smaller oscillation when the PMU measurement suffers a time delay of 47.5ms comparing to the case when all measurements are free of delay. To investigate the impact of POD signal delay, it is necessary to detect a threshold for the delay beyond which the system will collapse. However until the delay reaches a value of 300ms, which actually is of a very low probability to be experienced in the 4 PMU simulation scenarios according to [8], the system remains stable. It can be concluded that the POD is quite robust to its input delay. 5

6 In order to study the time delay robustness for voltage measurement at regulated bus, PMU voltage measurement at Bus 9 is corrupted with a delay of 47.5ms while other inputs are assumed to be free of delay. The simulation results are shown below in Fig 7: Voltage at Bus 9 (p.u) Fig 7: Voltage profiles at the regulated bus given perfect POD signal and delayed voltage measurement. The oscillation grows as the delay increases. By iterative simulations, we find that tolerated latency for reference voltage input is below 65ms. Comparing the simulation results from the delayed POD case, it is obvious that voltage measurement is much more sensitive to its delay comparing to the other measurements. B. Data incompleteness Time (s) As described in the previous section, phasor data that arrives later than the selected PDC waiting time will be ignored which consequently results in data incompleteness. To enable the calculation for WAMC control applications, there is need to replace the empty data slot in the phasor packet with a value. The most straight forward method is to adopt the latest available data. As the delay processes, the data incompleteness problem will ultimately turn into a data quality issue, in this case the performance of SVC functionality will be jeopardized. The dependency of this remote signal based SVC system reliability and selection of PDC waiting time, T TO, can be expressed as: 1 No time delay time delay of 47.5ms in reg-vol time delay of 65ms in reg-vol Where T represents the maximum delay to ensure reliable SVC function, S is the phasor data resolutions and P(T TO ) is the probability of data incompleteness for every PMU. All the phasor measurements are assumed to be taken independently from dedicated PMUs. When a stochastic model of T W is used, the probability of losing data can be illustrated in Figure 8. Fig. 8: Dependency of data incompleteness probability and the selection of PDC waiting time. According to the study of input signal delay robustness, the voltage measurement has a dominating impact on the overall system performance whereas the other PMU data that is used to calculate POD signal has very limited impact to the system reliability. Taking a data resolution of 30 samples per second, the reliability of this specific SVC application can be expressed as a function of probability of voltage measurement data incompleteness: 1 1 The simulation result from previous section shows that an overall delay larger than 65ms will lead to the failure of SVC function. In paper [14], the minimum delay that could be experience in wide area Communication network can be roughly estimated as 15ms (the approximation is made by using the mean value of delay distribution instead of the stochastic delay model). The largest possible PDC delay due to waiting can be calculated as 50ms, the further increase of time-out parameter will lead to SVC collapse. In [8], data incompleteness probability results are for the entire data packet which contains 4 phasors. The probability of losing this specific voltage measurement is a quarter of this value. The reliability of SVC function in association of PDC waiting time selection is presented in Table 4: 6

7 Scenar io T TO Table 4: SVC reliability Probability for losing voltage data SVC function reliability 1 15ms 6.25% 99.61% 2 25ms 5.5% 99.70% 3 30ms 3% 99.91% 4 35ms 2.5% Larger than 99.99% 5 45ms 1% Larger than 99.99% 6 50ms 1% Larger than 99.99% 7 15ms 11.25% 98.73% 8 25ms 8.5% 99.28% 9 30ms 6.25% 99.61% 10 35ms 2.5% 99.88% 11 45ms 1% 99.96% 12 50ms 1% 99.97% C. Discussions The simulation results indicate that inadequate supporting ICT system runs the risk of invaliding the WAMC systems. One possibility to enhance its robustness is to redesign the ICT architecture. Since SVC is located at the regulated bus in this specific case, it is possible to have a dedicated communication link between SVC and PMU with a low cost. If the calculation process is integrated with the SVC controller, the voltage measurement can be treated as local signal whose latency can be significantly reduced. The suggested ICT infrastructure which contains both local and wide area measurement is presented in Fig. 9: Fig. 9: A suggested ICT system architecture for WAMC systems to improve system robustness including information flows within the processes. The major argument here would be that there is a time skew between the POD signal generated from the 4 PMU signals and the voltage measurement since the former still has to pass the Wide Area Network and PDC sorting process which inevitably introduces a significantly higher delay comparing to the locally collected measurements. Assuming the local PMU measurement are with the delay of 10ms while the other PMU data passing through WAN and PDC sorting are experiencing a delay of 47.5 ms, simulations is carried out and results are presented in Figure 10: Voltage at Bus 9 (p.u) Fig. 10: Voltage profiles at the regulated bus given delayed phasor data from all PMUs with the suggested ICT architecture. The simulation results reveal that the SVC successfully damped the oscillation after the breaker reclosure even though a time skew exists between the POD signal and voltage measurement. It is due to the fact that POD signal is quite insensitive to its measurement delay. Hence a timely voltage measurement is sufficient to guarantee the WAMC system s performance. The SVC function reliability reaches a value larger than % with this architecture. VII Conclusion time delay of 47.5ms for POD and 10ms for reg-vol Time (s) Though the conclusion can only be limited to this specific network and wide area control scheme, however simulation presented in this paper can still raise the awareness of the dependency of WAMC system on their supporting ICT infrastructure. In this paper the consequence of WAMC system failure has been demonstrated when the PMU measurements are exposed to delayed or uncompleted input data. Based on the simulation results, it is realized that input signals of WAMC applications may have different robustness to their corresponding input latency depending on their functionalities. Specifically talking about this case, there are two input variables in SVC control scheme, termed POD signal and voltage at the connected bus. POD is mostly in charge of oscillation damping while the voltage measurement at the connected bus is the main reference for SVC to adjust its tunable reactance. According to the results in this paper, the later one has a dominate impact on the overall system performance 7

8 comparing the POD signal. Since the voltage measurement is more sensitive comparing to its POD counterpart, it is cost efficient to prioritize it in resources distribution to meet the data quality aspect requirements of the supporting ICT systems. [14] M. Chenine, E. Karam, L. Nordström, Modeling and Simulation of Wide Area Monitoring and Control Systems in IP-based Networks, in Proc of the IEEE PES General Meeting, July Although time delay caused by data transmission cannot be ignored no matter dedicated or shared communication scenarios are used, the time spent on computing and decision making in application and also function execution at control devices still take a large proportion of total response time Therefore, the WAMC system would have a major development with the technologies improvements both in hardware and software. References [1] A.G. Phadke, The Wide World of Wide-Area Measurements, IEEE Power and Energy Magazine, vol. 2, no. 4, September/October pp [2] D. Karlsson, M. Hemmingsson, and S. Lindahl, "Wide Area System Monitoring and Control," IEEE Power & Energy Magazine, vol. 2, no. 5, September/October 2004, pp [3] A.G. Phadke and J.S. Thorp, Synchronized Phasor Measurements and Their Applications ISBN: , Springer, Jan, [4] C. Marinez, M. Parashar, J. Dyer, and J. Coroas, Phasor Data Requirements for Real Time Wide-Area Monitoring, Control and Protection Applications, CERTS/EPG, EIPP-Real Time Task Team. January realtime_group_ data_requirements_draft5_jan26_2005.pdf [5] M. Kim, M.J. Damborg, J. Huang, and S.S Venkata, Wide-Area Adaptive Protection Using Distributed Control and High-Speed Communication, in Proc of 2002 the 14th power system computation conference, Sevilla, Spain. [6] C. Marinez, M. Parashar, J. Dyer, and J. Coroas, Phasor Data Requirements for Real Time Wide-Area Monitoring, Control and Protection Applications, CERTS/EPG, EIPP-Real Time Task Team. January [7] R. Moxley, C. Petras, C. Anderson, K. Fodero. Display and Analysis of Transcontinental Synchrophasors. Schweitzer Engineering Laboratories, Inc. 2004, retrieved form: [8] M. Chenine and L. Nordström Investigation of Communication Delays and Data Incompleteness in Multi-PMU Wide Area Monitoring and Control Systems in Proc of 2009 International Conference on Electric Power and Energy Conversion systems, Sharjah, UAE. [9] P. Kundur, "Power System Stability and Control," McGraw-Hill, Example 12.6, pp. 813, 1994 [10] M. Klein, G. J. Rogers, S. Moorty and P. Kundur, "Analytical investigation of factors influencing PSS performance," IEEE Trans. on EC, Vol. 7, No 3, pp , September [11] IEEE Special Stability Controls Working Group, Static Var Compensator Models for Power Flow and Dynamic Performance Simulation, IEEE Trans. on Power Systems, Vol. 9, No. 1, February [12] M. Pavella, D. Ernst and D. Ruiz-Vega, "Power System Transient Stability Analysis and Control," Kluwer Academic Publishers, [13] W. Hongxia, K. S. Tsakalis, and G. T. Heydt, Evaluation of Time Delay Effects to Wide-Area Power System Stabilizer Design, IEEE Trans. on Power Systems, Vol. 19, No. 4, pp , Novmber2004 8