A SERVICE ORIENTED WIRELESS SENSOR NETWORK FOR POWER METERING

A. Bagnasco, P. Buschiazzo, L. Carlino, A. M. Scapolla DIBE – University of Genoa, Via Opera Pia 11/a, 16145 Genova, Italy

Abstract

Wireless Sensor Networks hold the promise of innovative applications in the area of monitoring and control. This paper presents a distributed architecture to control a grid of Wireless Sensor Networks. Each Wireless Sensor Network is managed by a gateway device that uses a Web Service interface to provide basic functionalities for delivering data collected by the sensors. The sensor nodes have been designed with features of low cost, easiness to use, and offer specific support to industrial applications. The article examines how these nodes can be used in a real world scenario, an automatic meter reading system, where there is the need to monitor energy consumption of several buildings in a wide commercial area.

Keywords:

Wireless Sensor Networks, Web Services,

1.

INTRODUCTION

In the last years Wireless Sensor Networks (WSN) have been studied by many researchers and new communication standards have been developed to fit power constraints. IEEE 802.15.4 [8] and ZigBee [18] are among these. Commercial systems and hardware devices in particular low-cost processors, miniature sensors, and low-power radio modules, are becoming available. Many companies, like Texas Instruments, Atmel and Ember have developed Systems-on-Chip (SoC) that reduce the cost and the power consumption of the network nodes, whose batteries should last several years. They are standard compliant and the interoperability of products by different vendors is guaranteed [14].

2

A Service Oriented Wireless Sensor Network for Power Metering

In a typical situation, a WSN is composed of a large number of sensor nodes which are deployed in a region where there is the need to monitor a physical phenomenon. If the region is large, inaccessible or dangerous, this deployment could be random and the WSN nodes have to self-organize and to collect data automatically. Since the networks can be composed of hundreds or thousands of nodes, many algorithms and protocols [10][16], available in the literature, target network issues, like distributed routing, congestion control and reduced power consumption. WSNs for environmental monitoring and process estimation are investigated in [4], with attention to the trade-off between quality of service in process estimation and network life-time. Potentially WSNs support a broad spectrum of monitoring applications like habitat monitoring, object tracking, military surveillance, industrial plant controlling, fire detection, and many others. The absence of physical cables reduces the impact of the systems and their installation costs. One of the emerging application fields is automatic energy meter reading. As the European, Asian and North American electricity grids come under increasing demand, the focus on curtailing peak load asks to building owners and building automation stakeholders to participate in load shedding strategies. Besides this, electricity suppliers want to improve their services and enhance internal operation management by means of a real-time energy consumption indication from the meters. This paper presents the service oriented implementation of a WSN platform for monitoring power meters. The next sections describe the architecture of an Automatic Meter Reading (AMR) system for power metering and detail the hardware structure of a single node of the WSN.

2.

SYSTEM ARCHITECTURE

A WSN is typically composed by a set of low-power sensing nodes, deployed in the environment and managed by a more powerful node, called base-station or gateway. The gateway brings the acquired data to the upper level of the communication infrastructure, so it is often equipped with GSM/GPRS/UMTS, Ethernet or Wi-Fi connectivity. In the real world the number of WSNs for data acquisition, and thus the number of the gateways, tends to be large. Moreover, we have to deal with different kinds of gateways, provided by different vendors or integrators. It is important to find a standard and flexible way of communication to collect data from these gateways. This paper proposes the adoption of a Service Oriented Architecture (SOA) [9], a loosely coupled model, where clients ask services to the

3 The gateway gateway servers, without a specific knowledge about them. The service interfaces, described by WSDL (Web Service Description Language), hide the specific hardware and software architecture used in the WSN. The implementation of Web Services on limited-resources devices is not trivial [6] [11], but it is feasible at gateway level because these devices, the WSN access points, have more resources, in term of energy, computational power, and storage, than the simple sensor nodes. We have studied a two-layer architecture, where the lower layer provides basic functionalities for the exchange of data between the gateway and the sensor nodes using the ZigBee protocol, and the upper layer aggregates and export data via the Web Service technology. From the WSN point of view the gateway is the network coordinator, while from the application point of view, it is a data service provider.

Figure 1.

The system architecture

This approach allows aggregating many WSNs and to build Grid based infrastructures [3].

3.

THE GATEWAY

The gateway collects and stores the information acquired by the WSN nodes in a local memory, delivers data to the upper control layer and forwards configuration commands to the end nodes. This multi-threaded job requires a complete operative system. The Automatic Meter Reading system, presented in this paper, uses gateways that are implemented on the Fox

4

A Service Oriented Wireless Sensor Network for Power Metering

Board [5] platform. It is a ready to use Linux embedded board, based on an Axis ETRAX 100LX microprocessor [2]. The ETRAX 100LX processor has a 100 MIPS 32 bit RISC architecture with a 32MB RAM and 8 MB of Flash memory. The Fox Board offers several types of communication channels: up to 4 serial ports, two USB 1.1 interfaces and a 10/100 Ethernet interface. Fox Board also supports MMC/SD memory cards, to increase the persistent storage. Despite the original GNU/Linux distribution of Fox Board has a native web server, the adoption of the more sophisticated foXServe distribution was preferred. It contains an embedded version of the Apache Web Server and uses the SQLite library for high-level data storage methods. The SQLite library simplifies the implementation of the database of the data acquired from the WSN, and the reading procedures running in the Fox Board can use common SQL queries. The gateway manages all the nodes of the network through a radio module connected to an RS-232 port of the Fox Board. It stores two kind of information: data acquired from sensors and events, like alarms. The WSN sends three kinds of events to the gateway: local events, simple notifications and critical notifications. Local events are typically associated to well known occurrences, so they are directly managed by the gateway. Notifications are forwarded to the server. The critical ones usually represent alarms that need human intervention, so it’s necessary to notify them immediately. All the Web Services deployed on the gateway have been realized using the gSOAP toolkit [15]. gSOAP allows an easy development of Web Services by C/C++ language and provides a transparent SOAP Application Program Interface (API) that maps C/C++ native data types to XML schemas. The Web Services realized with gSOAP are compact and fast, so they can be deployed in embedded systems such as the FoxBoard.

4.

THE WSN NODE

Engineers have designed many sensor nodes (commonly called “motes”) for WSNs; some relevant examples are: Mica [7], MicaZ, Telos [12] from University of Berkeley, Imote developed by Intel Research, and EyesIFX by Infineon. These nodes and some other sensor boards with limited capabilities are often used as prototypes for research activities. They are good development platforms for the research community. There are many studies on new network protocols which employ these devices, but they are seldom used in real world applications, because they are expensive and sometimes use proprietary communication protocols, making difficult to achieve a complete interoperability between different devices. These considerations

5 The WSN node suggested designing a WSN node which could be totally customized to the application under development. The System-on-Chip approach has been adopted instead of conceiving a specific radio module and a separate computational unit. During the preliminary stage of the design different commercial products have been considered: Telegesis ETRX2, Xbee series 2 and Meshnetics ZigBit modules [17]. The final choice was ZigBit, since it is totally ZigBee compliant, while ETRX2 and Xbee series 2 use modified packet structures. ZigBit can be used as both a separate IEEE 802.15.4 radio modem and a complete system on chip. The ZigBit chip includes an ATmega1281 microcontroller and an AT86RF230 [1] radio from Atmel. The microcontroller has 8 kB RAM, 4 kB EEPROM, and 128 kB of program memory to host both the ZigBee protocol stack and the user program. The entire protocol stack is built on the TinyOS operating system and Meshnetics provides API functions to manage the reception and the transmission of data through the network. This way, application developers can build ZigBeebased wireless solutions, dedicated to their applications, without a deep knowledge of TinyOS and the nesC language. The same approach has been used, for example, by Texas Instruments with the release of API in the ZStack, and by Ember in the EmberZNet ZigBee protocol stack. The architecture of the node consists of a central unit which contains the Meshnetics ZigBit module, and one or more external sensor boards to acquire and condition analog signals and sensor data. This modular design guarantees a general purpose system. It allows to select the hardware required by the target application, avoiding unnecessary sensors or components. This way the power consumption and the dimension of the node are reduced at the minimum. Since industrial applications often use serial communication (like the Modbus protocol) and differential form of signaling (like RS422/RS485) we have developed a specific hardware module to manage these communications. In this board there is a low-power half-duplex RS422/RS485 transceiver that, instead of using termination resistors, uses a Schottky diodes termination. Unlike other termination types, Schottky diodes do not attempt to match the line impedance, but they simply clamp the overshoots and undershoots caused by reflections. Voltage excursions are limited to the positive rail plus a Schottky diode drop in one direction, and to ground less the same voltage drop in the other direction. Schottky-diode terminations waste little power, because they conduct only in presence of overshoots and undershoots. On the other hand, standard resistor terminations, ith or without failsafe bias resistors, draw power continuously.

6

5.

A Service Oriented Wireless Sensor Network for Power Metering

APPLICATION SCENARIO

The WSN node, described in the previous section, has been designed and tested for the application of power monitoring at the Porto Antico area of Genova. This is a wide commercial and recreational area planned by Renzo Piano for the Columbus Exhibition in 1992 in Genova (Italy), Power load and energy consumption of the utilities in the area must be monitored through a number of energy meters, located in different places. A control room in the headquarter building supervises all these devices. The area is very large (130,000 m2) and there are many sparse buildings. WSNs fit well communication needs. Furthermore, since power meters are already on, wireless nodes avoid the installation of additional cables or electrical equipments inside the cabins. The tiered multi-hop structure of ZigBee-based networks is well suited for this scenario. It is possible to use the nodes connected to the wall sockets as routers, which synchronize the other battery powered nodes of the room. The network layer of the ZigBee protocol stack, uses the Ad-hoc On-demand Distance Vector (AODV) routing protocol [13], which enables a multi-hop data transmission across the network; so data packets can cross different routers to reach the control room. Tests have been done on the collision avoidance of IEEE 802.15.4, to be sure that nodes could communicate even in presence of Wi-Fi access points or other electromagnetic interferences in the area. In the cabins there are two kinds of electricity meters: the Circutor CVMBD-RED-C420 and the Vemer Energy-400 PWR. The former is used inside the high power cabins and it is accessible through a RS485 interface by using a Modbus based protocol. The latter is a simpler energy meter that has an opto-isolated line, which provides a pulse every unit of KWh consumed. The sensor devices of the CVM meters, are equipped with an additional board to convert UART signals into RS485 differential signal levels. By using the Modbus protocol we can access also to internal registers of the meter that gives information about the peak load and the active/reactive power consumption. This information is useful for a real time analysis of the system. To reduce the power consumption of the nodes, a periodic polling of the CVM meters is scheduled and the nodes stay in sleep mode for the rest of the time. The sensor devices, which monitor Vemer Energy-400 meters, use the interrupt lines of the microcontroller to wake them up when a pulse is received. The devices are in sleep mode most of the time to save their energy and periodically send information.

7 Application scenario Each cabin is equipped with a WSN that acquires data from the meters. The gateway of each WSN exposes services that can be used by end users to collect data from power stations. To preserve battery life and to improve reliability, every node remains in a sleep state for the most of the time and wakes up periodically to send data. So, if the gateway doesn’t receive any packet from a specific node for a predefined amount of time, it alerts the management staff that there is a problem in the network, and there is the need to verify if a node or a communication link is broken. Every WSN node saves data in its internal EEPROM memory, and if the radio transmission to the coordinator is successful (that means that the remote node receives an acknowledge packet), the EEPROM location is cleared. Instead, if the transmission is unsuccessful, the node goes on incrementing the value of the measured pulses, until the communication link is restored. By using this approach, it is guaranteed that the system doesn’t lose any pulses from the energy meters. Since every nodes wakes up periodically to send data to the gateway, there is the need to define its optimum duty cycle, to preserve battery life, and to guarantee a good resolution of the system. For example, if WSN nodes wake up for 1 second, and sleep for 30 seconds, with 2 standard AA alkaline batteries (the capacity of a single AA battery is 2500 mAh), every nodes consumes about 19.4442 mA every 31 s. This means that, depending on the number of pulses received, the batteries will last for about 332 days in optimal working conditions (few loss packets, 0 dBm transmission power, good communication link, ideal batteries, etc.). An improvement could be obtained if the node wakes up only for half a second (which is a sufficient time to send data to the coordinator); in this situation the batteries life is about doubled. The following graph shows the estimated battery life of a single node, depending on its active and sleep periods.

8

A Service Oriented Wireless Sensor Network for Power Metering Figure 2.

Estimated battery life

As you can see from the graph, if the active period is long, then increasing its sleep period doesn’t extend battery life. Instead, when the active period is short, controlling the sleep period duration gives an improvement, even of one hundred days, on the life of the node. According to this, the energy monitoring system can use an adaptive strategy and change the sleep period of every node, depending on the energy consumption of the meters connected to them. So, if the electricity consumption of a utility is little, the system can change the duty cycle of the node, to save batteries. This adaptive control can be done from the management station, that can produce statistics of the current consumption of the different utilities, and then, depending on this analysis, it can send messages to the nodes, to change their duty cycle.

6.

CONCLUSIONS

In this paper we have presented a Wireless Sensor Network application for automatic meter reading that is been developed for monitoring the energy consumption and the peak loads of many different buildings in a wide commercial area. The proposed architecture allows controlling WSN devices using an embedded web server that provides a Web Service interface as well as common services on the Internet. The project was founded by the “Parco Scientifico e Tecnologico della Regione Liguria”. The author would like to thank Giorgio Allasia and Giovanni Giannotta for the technical contributions.

REFERENCES [1] “AT86RF230 ZigBee/IEEE 802.15.4-transceiver datasheet”, Atmel Corporation, San Jose, Ca. [2] “Axis ETRAX 100LX Datasheet”, Axis Communications, Sweden.

9 Conclusions [3] Andrea Bagnasco, Davide Cipolla, Arianna Poggi, Anna Marina Scapolla, “ServiceOriented Architectures for distributed cooperative instrumentation grids”, 2nd International Workshop on Distributed Cooperative Laboratories (INGRID07), S. Margherita Ligure, Italy, 2007. [4] D. Dardari, A. Conti, C. Buratti and R. Verdone, “Mathematical evaluation of environmental monitoring estimation error through energy-efficient Wireless Sensor Networks”, IEEE Transaction on Mobile Computing, Vol. 6, 2007. [5] “Fox Board, a Linux Core Engine in just 66x72 mm”, ACME Systems, Rome, Italy. [6] R. Glaschick, B. Oesterdieckhoff, C. Loeser, “Service Oriented Interface Design for Embedded Devices”, 10th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA), 2005. [7] J. Hill, and D. Culler, “Mica: A Wireless Platform for Deeply Embedded Networks”, IEEE Micro, vol. 22, pp. 12-24, Nov. 2002. [8] “IEEE 802.15-2006 standard for Low Rate Wireless Personal Area Networks”, Institute of Electrical and Electronics Engineers, 2006. [9] F. Jammes, H. Smit, “Service-Oriented Paradigms in Industrial Automation”, IEEE Transactions on Industrial Informatics, 2005. [10] H. Karl, and A. Willig, “Protocols and Architectures for Wireless Sensor Networks”, Wiley Interscience, John Wiley and Sons Inc, 2005. [11] B. Oesterdieckhoff, C. Loeser, I. Jahnich, R. Glaschick, “Integrative approach of Web Services and Universal Plug and Play within an AV scenario”, 3rd International IEEE Conference on Industrial Informatics (INDIN), 2005. [12] J. Polastre, R. Szewczyk, and D. Culler, “Telos: Enabling Ultra-Low Power Wireless Research”, Proc. of the 4th Int. Symposium on Information Processing in Sensor Networks (IPSN), Los Angeles, California, 2005. [13] “RFC-3561”, Ad-hoc On-demand Distance Vector. [14] K. Tuan, “Zigbee System-on-Chip (SoC) Design”, High Frequency Electronics, Jan. 2006. [15] R. A. van Engelen and K. A. Gallivan, “The gSOAP toolkit for web services and peer-topeer computing networks”, Proceedings of the 2nd IEEE/ACM International Symposium on Cluster Computing and the Grid, 2002. [16] R. Verdone, D. Dardari, G. Mazzini, and A. Conti, “Wireless Sensor and Actuator Networks: Technologies, Analysis and Design”, Elsevier, 2007. [17] “ZigBit OEM module product datasheet”, Meshnetics Corporation, Phoenix, Az. [18] “ZigBee Specifications”, ZigBee Alliance, Dec. 2006.