TCP/IP Communication Aspects in Monitoring of a Remote Wind Turbine

Transcription

1 TCP/IP Communication Aspects in Monitoring of a Remote Wind Turbine Tapio Sokura, Taneli Korhonen, Mikael M. Nordman, and Matti Lehtonen Power Systems and High Voltage Engineering Laboratory P.O.Box HUT, Finland {tapio.sokura, taneli.korhonen, mikael.nordman, matti.lehtonen}@hut.fi ABSTRACT Today there is a growing interest in utilizing different distributed energy resources (DER) in the power system. A problem that the system operators, owners and power utilities have to consider is thus how this technology can be integrated within the existing information systems. Another emerging problem is what type of telecommunication architecture should be used for remote monitoring and control of these units that normally are distributed over a large geographical area. One possible solution is to use the commercially available TCP/IP based telecommunication infrastructure and extend it with different technologies when needed to reach a distant DER unit. The commercial TCP/IP telecommunication network poses, however, concerns regarding the reliability, response time and level of security that can be achieved. This paper addresses different protocol aspects and presents a pilot study where a wide area internet and a local packet switched radio network were used to monitor parameters of a wind turbine located approximately 1000 km away from the control center. In the pilot system, the applicability of the protocol based on the IEC draft standard as well as a protocol based on a proprietary solution were assessed. Main emphasize was put to the evaluation of the system architecture as well as the reliability and response characteristics of the tested communication solution in the pilot installation. Based on the test results, the general usability of a commercial wide area internet and packet switched radios as a communication means in monitoring of distant wind power plants is discussed. INTRODUCTION One of the most important aspects that will affect the success and cost-competitiveness of distributed energy resources integrated within the power system, is the development and usability of communications and control technologies. Because distributed energy resources typically are small and geographically spread over a large area, the units must be linked together as well as monitored and controlled from a remote control center [1]. This requires that a robust and lowcost communication infrastructure is available and that appropriate communication protocols are developed. Today, the most known communication protocol is TCP/IP. TCP/IP was developed by research groups in the USA in the late seventies and became during the next twenty years the de facto standard for communication between computers and information systems [2]. TCP/IP is today used in all relevant telecommunication systems and is also considered to be the preferred platform for future communication in large distributed control infrastructures. However, TCP/IP 1

2 defines only the network and transport protocols. Hence, the application objects, data format and exchange rules must be specified by some other related standard. Well-known commercially used application protocols are e.g. FTP for file transfer and HTTP for transfer of web contents. Appropriate application protocol standards for TCP/IP based communication networks in the power systems domain, are currently under development. For example, the joint efforts of IEC, IEEE and EPRI has resulted in the proposed standard IEC 61850, which defines functions, data objects and protocol mappings for substation automation (and determines thus generic interfaces for distribution automation as well). This joint standardization effort has also been extended to consider e.g. distributed energy resources. It seems thus possible that in the future there can be a commonly agreed TCP/IP based communication platform for the management, control and integration of power systems and distributed energy resources. Because the distributed energy resources are geographically spread over a large area, it may not be economically feasible to purchase proprietary physical communication links. The communication resources can instead be leased from a telecommunications operator, e.g. an internet service provider. However, there are two major concerns with this arrangement. Typically the first issue considered relates to the level of data security that can be offered. Although the communication can be implemented using secure virtual private networks (a link protected by means of encryption and authentication providing rather reliable and secure data transmission [3]), the trust in this service should be delegated from the power system operator to the telecommunications network operator. These trust relationships are today not sufficiently strong and the operators lack tools to efficiently manage them. The second major concern relates to the performance and the reliability of the TCP/IP based network. The leased resources are shared with other traffic and although prioritization can be implemented to avoid congestion in network routers, these types of service agreements are not very common today (and also difficult to implement in networks with several operators involved). In addition, because the TCP protocol was not developed for real-time applications, it does not provide means to control acknowledgements nor does it provide means to efficiently control the response time (e.g. the timeout mechanisms of the TCP retransmission scheme varies with the quality of the link [2]). Using UDP and application level acknowledgements can however mitigate this problem. In this paper, a pilot study of a system using a leased wide area communication link from an internet service provider to monitor a remote wind turbine is discussed. The pilot system is used to assess the performance characteristics of the communication system and its behavior when different communication protocols are used. Emphasize is also put to the performance of the interface link between the wind turbine and the wide area network. This link is implemented with radio modems. The protocols tested are the draft standard IEC , which is developed for the remote monitoring of wind power plants, and a proprietary protocol, which is developed for test purposes only. The paper is structured as follows: In Section 2, a short introduction to the current standardization work in communication protocols for distributed energy resources, is given. Main emphasize is put to the communication standard for monitoring and control of wind turbines and wind power plants. Section 3 discusses the general architecture of the communication system used in the pilot project. Section 4 is dedicated to evaluation of the pilot system, applicability of the communication protocols in this particular environment, and to the traffic tests that were made in the system. Section 5 concludes the paper. 2

3 INTRODUCTION TO THE RELEVANT STANDARDIZATION WORK In the power systems domain, the actors (manufacturers, operators, utilities, etc.) have always faced problems of what communication and data modeling protocols to provide to the customers for monitoring and control of their devices. Distributed energy resources face the same problems and proprietary solutions have generally been as common in this area as in the rest of the domain. However, as an extension to the power system standardization effort UCA (Utility Communications Architecture), a new and international framework for defining interconnection and data communication for distributed resources is evolving. It has the work name UCA-DER (Utility Communications Architecture Object Models for Distributed Energy Resources) [4]. The approach taken in the standardization work is to define generic object models of devices, of functions and for the exchange of information within a system. That is, a real device (e.g. a fuel cell) is modeled as a logical device containing data objects that determine the information and attributes that can be exchanged and managed as well as the functionality that the devices can support. The data exchange is realized with a client-server relationship between the devices and with several different communication services. Typical services being supported are conventional request-response data transactions, reporting and logging. Hence, UCA-DER adapts the same modeling approach as used in the standard for substation automation, IEC [5], which is frequently discussed among utilities and electrical engineering companies today. One subpart of the general standardization effort considers wind power plants. This standardization work has been given the name IEC Communications for monitoring and control of wind power plants [6]. It defines the object models and information exchange models of units and devices in wind power engineering as well as for auxiliary services. Consequently, in the standard, object models for e.g. wind turbine rotor information (WROT), wind turbine generator information (WGEN), wind turbine alarm information (WALM), and wind power plant meteorological information (WMET), are given. In addition, different communication topologies and services are determined. These are given on an abstract level and specific performance requirements are not specified (due to the assumption that a wide range of applications and communication systems will be integrated). Nevertheless, the major topologies proposed, allow the protocol and data servers to be implemented in three locations, at a local supervisory computer, at the controllers directly connected to the wind power plant components, and as a mix of these two. The major communication services proposed for wind power plant management are authorization (e.g. authentication service), client initiated control operations, monitoring (both request-response and spontaneous reporting) and different management functions (e.g. time synchronization and configurations management). The communication services and the application data contents are in the real implementations mapped to some application protocol on top of preferably TCP/IP. Four different mapping profiles are proposed in the standard. These are ISO 9506 (MMS) + ASN.1, or XML + HTTP, or DNP3.0, or IEC /104. [6] Generally it can be expected that the MMS and the XML based profiles will gain most attention and the DNP3.0 and IEC 101/104 standards will be used in the transition from currently developed systems to the systems developed in the future. 3

4 SYSTEM ARCHITECTURE OF THE PILOT STUDY At the Olos fell in Lapland, located approximately 1000 km north of Helsinki, 5 wind turbines, each with the size of 600 MW, are located, see Figure 1. In this environment, the purpose of the pilot study was to test the applicability of a wide area communication network (refereed to as an internet) and different protocols in remote monitoring of one of the wind turbines. A standard ADSL connection of 256 kbps was chosen for the implementation of the internet link. This connection was leased from an internet service provider and connected clients located in the Helsinki area with a data server located in Muonio close to the Olos fell. The communication between the server and the controllers, which managed the interface to different wind turbine sensors, was implemented with packet switched radio modems having the speed of bps. The distance between the server in Muonio and the controllers at the Olos fell was approximately 6 km. The general system architecture and the monitored parameters are depicted in Figure 2, and a snapshot of a view presented to clients (in a standard Web browser) is shown in Figure 3. Figure 1: Map of Finland where the location of the wind turbine and server (at Olos) as well as the location of the client (Helsinki) are shown at the left-hand side, and a picture of the measurement arrangement and thus the location of the controllers is shown at the right-hand side. WAN TCP/IP Controller A Wind speed Temperature Humidity Web client Server Controller B Energy of wind turbine 2 Figure 2: The architecture of the test system. 4

5 Figure 3: A snapshot from a Web client of the wind speed and the generated energy. Three different protocol configurations were developed, see Figure 4. The first protocol was based on the IEC logical node object WMET that was mapped to the application protocol IEC according to the Annex B in [6]. Of the available WMET attributes, the wind speed, ambient temperature and humidity were selected. Encapsulated in the IEC application data frame, this application data had a total length of 52 Bytes. With the TCP/IP header (40 Bytes) and the PPP header (5 Bytes) added the length of an IEC type protocol message was thus 97 Bytes (excluding low-level framing and byte stuffing). The second protocol was a proprietary protocol developed for assessing different performance parameters of the communication network as well as the performance of the IEC based protocol. The content of the proprietary protocol is shown in Figure 4. It has a header of 11 Bytes and a maximum data payload of 244 Bytes. This protocol was configured to support TCP/IP and PPP with a total header length of 56 Bytes, or alternatively a plain serial transport protocol (total header length 11 Bytes). The plain serial protocol and the PPP protocol were used only in the radio communication between the controllers and the server (see Figure 2). The internet had its own low-level protocol layers that are dependent on the network owned by the service provider. The following data exchange services were developed: Periodic data transactions initiated by a controller (controller server communication). Spontaneous data transactions initiated by a controller upon violation of some threshold (controller server communication). Periodic request-response transactions initiated by some remote client (client server communication or client controller communication with the server as a transparent node forwarding the messages). Spontaneous request-response transactions initiated by some remote client (client server communication or client controller communication with the server as a transparent node forwarding the messages). Time synchronization. 5

6 Configuration A (IEC ) Radio/WAN Framing PPP/- IP TCP IEC IEC Configuration B (TCP/IP protocol) Radio/WAN Framing PPP/- IP TCP Proprietary Configuration C (Serial protocol) Radio Framing Sender [1] Receiver [1] Length [1] Type [1] Seq Nr [1] Timestamp [4] Data [0-244] CRC [2] Figure 4: The three different protocol configurations (A), (B), and (C) used in the test system. Hence, the server could not spontaneously send critical messages (e.g. alarms) to a client. This limitation was because the clients were intended to be web clients without advanced features like e.g. server push implemented. SYSTEM PERFORMANCE EVALUATION In remote monitoring and control of wind power plants, the requirements on the communication system are rather rigorous, especially if control sequences are allowed. Hence, although the IEC standard does not give specific performance requirements, it can generally be stated that it is expected that the data transfer is reliable, that the response time is predictable and that messages are rarely lost or duplicated. However, for monitoring purposes only, the performance requirements can be slightly relaxed, especially if remote alarm management is not considered a function of high priority. To evaluate the applicability of an internet type network (generally considered a broadband telecommunications network although there are different opinions regarding the definition of broadband), packet switched radio modems and different protocols, several performance tests were made in the pilot installation. In the first test session, the average round-trip delay of request-response transactions between a controller and the server (the radio link), the server and a remote client (the internet link), as well as a controller and a remote client (denoted the total link) were measured. In this test session, the serial proprietary protocol (depicted C in Figure 4) and the TCP/IP based proprietary protocol (depicted B in Figure 4) were assessed for different sizes of the application data. The IEC based protocol was not considered, as it in these tests simply is a specific case of the TCP/IP based protocol. The test results are depicted in Figure 5, where the left-hand side shows the measured protocol throughput characteristics and the righthand side shows the round trip delay characteristics. The serial type protocol performed, as expected, slightly better than the TCP/IP based protocol. The different length of the packet header (header lengths of 56 Bytes compared to 11 Bytes) mainly caused the performance differences. For a standard remote monitoring application, however, the measured performance of both protocol types was acceptable. 6

7 Figure 5: The left-hand side shows the throughput characteristics and corresponding maximum number of queries that can be achieved with the proprietary protocol using the simple serial protocol stack and TCP/IP. The right-hand side shows the average round trip delays (controllerserver, server-client, and controller-client) achieved with the serial and the TCP/IP based protocols. In addition, the delay characteristics of the internet network were measured (with standard TCP/IP ping and traceroute messages). The internet delay was on average 46 ms, with a maximum deviation of ±10%. This delay mainly (approximately 70%) originated from the last link between the ADSL modem at the server end and the internet gateway it communicated with. Generally, a total of 10 internet gateways connected the client site with the server. Of these 4 were part of the general university network in Finland (the test client was located at Helsinki University of Technology) and 6 were part of the internet service provider s network. Next, the reliability of the radio network was assessed. As in the previous test session, the serial protocol and the TCP/IP proprietary protocol were tested with different size of the application data. The protocol messages were sent between a controller and the server without using application layer acknowledgements. Hence, application layer retransmissions did not occur. The results of the test runs with the serial protocol are denoted in Figure 6. The test results of the TCP/IP protocol are, however, not shown because no packets were lost with this protocol (because TCP handles retransmissions). Nevertheless, with the plain serial protocol the probability of not succeeding was almost always below 1%, and this probability did not depend on the message size. Hence, it was concluded that some messages were lost but with retransmissions this problem could be mitigated. The radio network did therefore not decisively degrade the communication performance of a remote monitoring application. As the first test session was dedicated to measurements of the average round trip delays, the third session considered the distribution of the round trip delays for request-response transactions implemented with the different protocol configurations. The aim with these measurements was to assess the predictability of the performance achieved with the TCP/IP protocol in the pilot system. Especially the characteristics of TCP/IP at the radio modem link were of interest, as this link was assumed to have less predictable performance than the internet link. However, in the 7

8 same test session, also information of the internet link was collected. Hence, protocol messages (request-response transactions) were sent between a controller and a remote client using the serial protocol (i.e. without TCP/IP at the radio link and with TCP/IP at the internet link) and the TCP/IP protocol (TCP/IP at both the radio and the internet link). Three different sizes of application data were tested. The test results are shown in Figure 7. A clear difference can be noted in the distribution of round trip delays between the serial protocol and the TCP/IP protocol. Because both protocol configurations used TCP/IP at the internet link, the difference is assumed to originate from the radio link. The explanation for the distribution can thus be found in the way that TCP handles retransmissions and timeouts. TCP adjusts the value of its retransmission timeout according to the time elapsed between a message transmission and the reception of the acknowledgement. Therefore, if the quality of the link is bad, the retransmission timeout grows and can finally become significantly long. Consequently, both the number of retransmissions needed and the extended timeout affect the duration between successful transactions. [2] This phenomenon is seen in the results shown in Figure 7. Accordingly, it can also be seen that the size of the application data did not affect the behavior (almost equal distributions for 0, 100, and 240 Bytes of application data) and that the internet link had a rather predictable performance. Figure 6: Lost packets probability for different packet data lengths at the radio link (controllerserver). Figure 7: Distribution of the round trip delay (controller-client) for the serial proprietary protocol (left-hand side) and the TCP/IP proprietary protocol (right-hand side) and for different packet data lengths. 8

9 In the last test session, the distribution of the round trip delay of request-response transactions with the IEC based protocol was compared to the distribution of the round trip delay of request-response transactions with the other two protocols having the same application data contents. To assess the distribution characteristics the following protocol configurations were made (see also Table 1 for the total size of the data packets at different communication links): The proprietary serial protocol with application data 12 Bytes and standard header 11 Bytes. The application data size is determined by the data needed to send the wind speed, temperature, and humidity as well as related timestamp information with the protocol. The proprietary TCP/IP protocol with application data 12 Bytes and standard header 51 Bytes (TCP/IP header + proprietary header). The application data size is determined by the data needed to send the wind speed, temperature, and humidity as well as related timestamp information with the protocol. The IEC protocol mapped to IEC and TCP/IP. A total of 52 Bytes application data was sent (three objects, wind speed, temperature and humidity according to the specification). The test results are given in Figure 8, where the left-hand side shows the general distribution and the right-hand side shows a zoomed version of the distribution (for 100 ms to 500 ms) of the round trip delays. The results conform well to the results shown in Figure 7. Accordingly, the IEC based protocol performance was similar to the performance measured for a general protocol using TCP/IP (this was also expected). Therefore, it is verified that the TCP protocol may not be appropriate for control applications where the response times must be short and predictable, especially if communication links with lower quality are used (e.g. radio links experiencing noise from the environment). However, for remote monitoring applications the achieved performance is satisfactory, also with a protocol based on the IEC standard proposal. DISCUSSION Using the available telecommunications infrastructure as a means to accomplish remote monitoring of distributed energy resources seems to be rather attractive and cost-efficient. However, there are several problems that must be considered before an implementation can be realized. For example, data security must be managed, a low-cost solution for the communication between the resource and the interface gateway of the telecommunication network must be found (generally called the last mile problem), appropriate application protocols and data modeling should be standardized and the performance of the communication network should be assessed. Table 1: Size of the data packets for different protocols used in the measurements of the round trip delay distribution. Protocol Radio link [Bytes] Proprietary (serial) Proprietary (TCP/IP) IEC IEC Internet link (without low-level protocol headers) [Bytes]

10 Figure 8: Distribution of the round trip delay (controller-client) with different protocols. In this paper a pilot system for remote monitoring of a wind turbine is discussed. A wide area communication network (internet) is used to connect remote clients to a data server, which further communicates with the wind turbine controllers via radio modems. The main motivation for developing the pilot system was to assess the system architecture and the applicability of the communication means as well as the proposed standard for wind power management, IEC Different communication performance tests were executed in the pilot environment. According to the experience gained from the developed system and the executed test sessions, it can be concluded that the IEC based protocol performance meets the requirements for remote monitoring applications. When retransmissions were used, packets were not lost. The internet response time was rather predictable and the general round trip delays for data transactions was satisfactory. However, using the (non-optimized) TCP/IP protocol at the radio link caused some unpredictable delays. The pilot study presented in this paper is a pre-study of a real system and more extensive traffic tests should be made to verify the behavior under different conditions. In addition, the IEC standard proposes several different protocol mappings of which only the IEC has been tested. This configuration should be compared to e.g. an XML based solution, and to a solution using e.g. GPRS instead of the ADSL internet link. However, the results of the conducted tests give confidence in the applicability of IEC and TCP/IP as a future platform for remote monitoring of wind power plants. REFERENCES [1] E. Petrie, and D. Julian, Microgrids unleash true power of dispersed energy, IEEE Transmission & Distribution, August [2] Comer, Internetworking with TCP/IP, vol 1: Principles, Protocols, and Architecture. 3 rd ed., Engelwood Cliffs, New Jersey, USA, Prentice Hall, 1995, ISBN: [3] Venkateswaran, R., Virtual private networks, IEEE Potentials, vol 20, issue 1, February- March

11 [4] EPRI, Draft UCA-DER Object Models, October [5] IEC-TC57-WG10/11/12, IEC 61850: Communications Networks and Systems in Substations, International Electrotechnical Commission, Draft IEC 61850, [6] Draft IEC , Communications for monitoring and control of wind power plants, 4 July,