November 2009

Special Section: Wireless

A real mesh

Low-cost wireless sensors test performance, battery life

Fast Forward

  • Researchers demonstrate effectiveness of wireless sensors in power plant.
  • Sensors feature self-forming mesh network, low power for lower costs.
  • Focus on radio performance and battery life proves a success.
 
By Cyrus W. Taft, Aaron J. Hussey, Teja Kuruganti, John N. Sorge, and Asis Nasipuri

Since low-cost, battery-powered, wireless sensors have the potential to radically alter traditional methods of equipment monitoring in power plants, researchers from EPRI, Oak Ridge National Laboratory, Southern Company Services, University of North Carolina at Charlotte, and Taft Engineering set out to demonstrate the effectiveness of wireless sensors in a power plant. Focusing on radio performance and battery life, they wanted to provide guidance for someone deploying a wireless sensor network in a plant on the allowable distance between sensors and what to expect for battery life. The results were a 15-minute powerup of 12 motes communicating with the base station and no obvious problems.

Mesh1

Historically, the normal practice in power plants and other process industry facilities has been to hardwire important sensors from their location in the plant to a central equipment room where monitoring systems are installed to collect the data from the sensors. This arrangement worked well, but the cost of installing all the wiring to connect each sensor to the central location was high, often more than the cost of the sensor itself. This high cost has discouraged companies from installing all the sensors they need to fully monitor their plants.

A research project co-sponsored by Southern Company and EPRI to demonstrate wireless sensors in a power plant environment focused on sensors' radio performance and battery life. Each wireless device, or mote, consists of two small circuit boards, one for the processor and radio and the other for sensors or data acquisition. The research team deployed two types of processors and radios and three types of sensor boards. The sensor boards had several built-in sensors for temperature, ambient light intensity, barometric pressure, sound, magnetic field, and acceleration in x and y directions, along with general purpose data acquisition capabilities. The wireless communication conforms to the IEEE 802.15.4 standard for low power, low data rate sensors with mesh networking capabilities. The team gathered results of a radio frequency site survey conducted before the installation of the sensors.

Wireless sensors

Early wireless sensors used proprietary protocols over a point-to-point radio link and required power high enough to make operation on batteries impractical. In the last few years, a new class of wireless sensors features a standard physical communication layer, a self-forming mesh network, and power use low enough to enable battery life of a year or more. These new wireless sensors have the potential in the next few years to dramatically change not only the wireless sensor landscape but the process monitoring landscape.

Before the power industry can accept any new technology, tests must occur in a realistic power plant environment. Southern Company and the UNC Charlotte conducted two sets of tests focused on wireless communication and battery life. Before testing the sensors in the plant environment, the research team tested them in an office environment to gain understanding of the operation of the sensors and the mesh network.

Mesh2
When an unidentified interference was present intermittently in all the frequency bands, it raised the noise floor by 43 dBm. The signal duration was 0.78ms, but the percentage of time it was present was not determined. The source of the noise signal was not identified, and it did not appear to interfere significantly (if at all) with the mote testing.

Mesh3
The spectrum analyzer monitored six frequency bands at each location until no new peaks appeared. The bands were 1MHz to 30 MHz, 30 MHz to 100 MHz, 100 MHz to 500 MHz, 500 MHz to 1 GHz, 1GHz to 2 GHz, and 2 GHz to 3 GHz. This results in a complete scan from 1 MHz to 3 GHz. The laptop records the frequency and power level at designated increments and records the maximum signal present during the total scan.

Plant radio frequency site survey

Before installing any new wireless network technology such as a wireless sensor network, quantify the existing radio frequency (RF) environment to determine if potential interference exists. It is also useful to test the signal propagation characteristics of the facility where you will install the wireless equipment to ensure reliable communication. The U.S. Department of Energy Extreme Measurement Communication Center at Oak Ridge National Laboratory will perform both tests at the E.C. Gaston Electric Generating Plant. Plant Gaston currently does not have any wireless sensor networks installed.

The team used two instruments to record and analyze ambient RF signals. An Aeroflex CS65040 Broadband Signal Recorder and Generator recorded the signals in the time domain, and a Rohde & Schwarz FSH3 spectrum analyzer provided spectral information about the signals. Two special antennas provided the inputs to the analyzers. The monitoring equipment was mounted on a wheeled cart and moved to four sites in the plant. The test site locations were 1) on the turbine deck near the unit 4 turbine, 2) on the mezzanine level under the unit 5 high pressure turbine, 3) on the mezzanine level under the unit 5 generator, and 4) on the base level near a boiler feed-pump turbine.

Data from these tests is presented as screen shots from the instruments. Only a few screen shots are included in the report as examples of the information obtained from the tests.

The signal recorder has an anomaly that appears as four continuous wave (CW) signals, first at -15 MHz from center, second at -10 MHz from center, third at +5 MHz from center, and fourth at +30 MHz. The signal at -15 MHz pulses, while the others are constant. They are present at these frequency offsets in all plots and in all frequency bands. Ignore these signals.

The team set up the signal recorder to collect data in four frequency bands: 500 MHz, 900 MHz, 2.4 GHz, and 5.8 GHz. The bandwidth for all bands is 130 MHz. The plots are the result of replaying the signals recorded during the tests. The average noise floor across the different locations ranged from -79 dBm to -85 dBm in the different frequency bands.

Southern Company mote testing

The project plan for mote testing was to do the initial familiarization and setup and a short test in an office environment and then deploy the sensors in the plant for longer term testing. Testing the wireless sensors in an office environment before testing them in the plant was attractive because the office environment was more controlled, more convenient, and less costly. The objectives of all the tests were:

  • Familiarization with the hardware and software being tested.
  • Document the range of the low power radio system.
  • Verify the operation of the mesh network and its adaptability to network changes.
  • Measure the battery usage and estimate battery life.
  • Develop a method for measuring temperature with an external thermocouple.
Mesh6
The data from the spectrum analyzer for the 30 MHz to 100 MHz frequency range. Peaks on the right side of the graph indicate the presence of RF signals from local television and FM radio stations.

 

Mesh5
The 500 MHz to 1 GHz range: The team detected no communication signals in any of the ISM bands, so there is little chance of interference with the wireless sensor network under test.

The team set up the initial wireless sensors in an office where the radio environment is friendlier than in a plant. It developed the thermocouple input system using the low level analog input data acquisition board in the office. The monitoring program monitored the sensors and saved the collected data. The team estimated battery life at about 100 days based on one week of operation. The team learned during the office testing that the monitoring program could only collect data from a single type of sensor board at one time. Since two types of sensor boards were seeing use in the project, another program was needed. The mote vendor provided a new gateway product that supported many types of sensor boards at the same time.

The gateway is a small computer running the Linux operating system, which interfaces with the base station on one side and an Ethernet network on the other. The focus of the plant environment testing was to determine how well the sensors communicated in the relatively unfriendly radio environment of a power plant. The goal was to provide guidance for someone deploying a wireless sensor network in a plant on the allowable distance between sensors and what to expect for battery life.

The wall of the control room in which the base station was located is made of sheet metal. To ensure good communication through the wall, the team positioned one mote just outside the control room. The bulk of the remaining motes were at least 100 feet away under the unit 5 turbine.

The computer communicates with the gateway in one of three ways over an Ethernet network. The gateway has a web server built-in, so a web browser in the computer can login to the gateway and view sensor and network data. You can also make a connection using a secure shell (ssh) program on the computer. Finally, you can attach the disk drive on the gateway as a network drive (using Samba on the gateway) to the computer for sharing data files.

The research team distributed 12 motes around the area under the unit 5 turbine. This area has many large steam pipes and structural steel elements that have the potential to obstruct the radio signals from the wireless sensors. In most cases, there was no clear line of sight between motes. Because the team only planned the test to last a couple of months, it placed the motes in plastic zip-lock bags to protect them from dust and water.

The research team powered up the wireless sensor network equipment, and within 15 minutes, most motes were communicating with the base station. A mesh network was established, and all but two motes that could not communicate with the base station directly were able to communicate through the mesh. After the two motes were relocated slightly, they began communicating. The team left the motes in these locations for several weeks, recording sensor data and mesh network performance data continuously. There were no obvious problems with the mote communication, but the mesh structure did change occasionally for unknown reasons.

ABOUT THE AUTHORS

Cyrus W. Taft, PE is president of Taft Engineering, Inc., in Harriman, Tenn. (cwtaft@taftengineering.com) Aaron J. Hussey is a project manager at the Electric Power Research Institute (EPRI) in Charlotte, N.C. (ahussey@epri.com) Teja Kuruganti is with Oak Ridge National Laboratory in Oak Ridge, Tenn. (kurugantipv@ornl.gov) John N. Sorge, PE is a research engineer at Southern Company Services, Inc., in Birmingham, Ala. (jnsorge@southernco.com) Asis Nasipuri is an associate professor and graduate program director in the Department of Electrical & Computer Engineering at UNC Charlotte (anasipur@uncc.edu).

RESOURCES