By Douglas J. Smith IEng, Senior Editor
In order to be competitive electric generating companies are now incorporating state-of-the-art technology in the management and operations of their generation, transmission and distribution systems. Management is purchasing sophisticated condition and performance monitoring systems for use in everyday operations, and some are also utilizing the internet, or their own intranet, to centrally monitor the operations of power plants.
As electric power companies face competition for the first time, many of them are realizing the importance of real-time condition and performance monitoring of power plant equipment. Real-time condition and monitoring systems provide power plant operators and engineers with data that immediately allows them to determine the need for maintenance or if there is the potential for major equipment failures.
For many years power plants have regularly collected and stored data from power plant operations. Unfortunately, in only a few cases, has all of the data been analyzed. However, this is changing and plants are now installing software that allows them to analyze all of the data in real-time. Rather than adding new monitoring and diagnostic systems electric generating companies are installing software to integrate data currently being collected from a variety of in-house monitoring systems.
SmartSignal Inc., Lisle, Il. recently released eCM2.0, an advanced equipment condition monitoring software product. This software uses empirical mathematical techniques to understand the complex relationship between correlated pressures, temperatures, flows, vibrations, speeds, voltages, current and other sensor outputs from a variety of plant equipment.
According to the developer of the technology, SmartSignal, the eCM2.0 system incorporates three proprietary software engines: A set-up engine, a signal engine and an alert engine, Figure 1. After the software has been installed the set-up engine’s algorithms use reference data representing normal equipment operation to create an empirical model of the equipment’s normal operation.
Based on real-time signals, and interfacing with the set-up engine model, estimated signals are determined for various sensors. These signals indicate the estimated value of each sensor during normal operation of the equipment. The final step compares the estimated signals with the real-time signals. Unlike some systems that alert the operator when a set point has been exceeded, the higher sensitivity of the eCM2.0 enables it to identify operating problems while the equipment is still operating well within its design limits.
Trending with eCM2.0
A 161 MW cogeneration plant supplies process steam and approximately 50 percent of the electricity to U.S. Steel’s Gary, Ind. facility. In late 2000 through early 2001, plant engineers started to notice an increase in vibration levels on the steam turbine and some problems with heat exchanger efficiencies. Over time this started to limit the output of the steam turbine.
Within any 24-hour period, the manufacturing facility’s electricity and steam requirements vary considerably. As a result, the cogeneration plant’s load is never constant. Since the load is inconsistent, reliable trending of increasing vibration of the steam turbine, and the fading efficiencies of the heat exchangers, was not possible.
During a four-week outage to implement manufacturer’s recommended modifications to the steam turbine, an inspection and evaluation of the turbine revealed that the increased vibration and loss of capacity was due to sodium build-up on the blades. Following the removal of the sodium build-up, and completion of the manufacturer’s recommended modifications, the unit was returned to service. Since that time the steam turbine has not experienced any vibrations and the unit has been able to operate at full capacity.
Because the engineers were unable to adequately trend, predict and diagnose problems in the operation of the cogeneration plant without dis-assembling the turbine, U.S. Steel decided to research technology that would give the plant the ability to detect equipment deterioration before it became a problem. After researching different options, U.S Steel purchased an eCM2.0 on-line diagnostic system from SmartSignal in 2000.
Utilizing the new software, the plant is now able to continuously monitor the turbine, steam extraction system, generator, condenser and the bearings. In use the eCM2.0 software continually checks and analyzes the flow, valve positions, temperatures, amps, voltages and vibrations of the cogeneration plant. Signal inputs from the sensors are sampled at one-minute intervals. After analysis the data is stored on a plant-wide historian. A total of 60 points are monitored.
Although the system is monitoring in real-time it does not alarm if a trend is detected. The eCM2.0 software is an early detection device, which unlike earlier systems, is able to predict a potential problem way ahead of it actually occurring. At U.S. Steel, the engineers use the eCM2.0 system as a trending tool.
Once per day the energy distribution process manager generates a daily report. Should the data indicate abnormal operating conditions it shows up on a “watch list.” The watch list also indicates the amount of time it has been in alarm. All items on the watch list are forwarded to the maintenance manager.
The four-week outage cost U.S. Steel $2 million for parts, labor and electric power purchases. Although the plant would have still spent a considerable amount of money for the outage, even with the eCM 2.0 system in operation, the plant would have saved the cost and time of hiring field engineers who were employed to evaluate and determine the operating problems.
Arizona Public Service Company is currently in the process of evaluating the installation of SmartSignal’s eCM2.0 software to monitor the three units at the Palo Verde nuclear power plant. Although the plant already has a variety of continuous on-line monitoring systems, engineering management was looking for a system that could provide an early warning of potential problems prior to them becoming serious.
According to Steve Coppock, Department Leader-Plant Reliability and Modifications, the new software will enable engineers to detect very small changes in equipment operation much sooner than the plant’s existing monitoring systems. In Coppock’s opinion, the eCM2.0 software is a cost effective solution for providing an early warning of degrading equipment performance.
The primary benefit of the system will be the software’s early warning capability, Coppock says. He further states that the system will allow the utility to reduce the time and cost of recalibrating the plant’s non-safety related instrument loops. On a regular basis the plant must check the calibration of hundreds of instrumentation loops. Invariably only a few of the loops require calibration. SmartSignal and Palo Verde engineers are also studying the application of eCM2.0 technology to similarly reduce the calibration costs of the safety related instrument loops.
Impact Technologies and EPRI have completed a study of a web-based performance monitoring and condition assessment of steam turbines at a Reliant Energy power plant. According to Dr. Michael Roemer, director of engineering with Impact Technologies, using the internet is an inexpensive vehicle for delivering real-time performance and diagnostic information to key off-site electric utility personnel. By merging the latest internet communication technology with advanced diagnostic and prognostic algorithms, the system is able to deliver sophisticated health monitoring information in a highly accessible format, says Roemer.
The web-based performance monitoring system consists of data acquisition, storage, validation and access, Figure 2. With the Reliant Energy project, the turbine’s performance data, already stored on the plant’s data acquisition system, was transmitted and stored in a database on the utility’s Wide Area Network (WAN). Using a file transfer protocol (FTP), the data was transferred across the internet to a remote Impact Technologies server.
Using intelligent algorithms, the server validates the sensor data and detects any performance anomalies. Using this information the system is then able to diagnose the most likely cause of performance degradation. A complete diagnostic report is written to an output database on the server. Active server pages use VB Script programs and active data objects to update the web site with real-time data.
In operation the monitoring and diagnostic algorithms first conducts a data correction routine by calculating the equivalent sensor readings when the equipment is operating under normal operating conditions. Data validation tests are then performed to detect any abnormally low or high readings.
Following data validation the statistical trending algorithms are used to detect incipient performance faults. This is accomplished by identifying shifts in the mean values of recent corrected and calculated parameters. In the event of a statistically significant and/or severe deviations from baseline values, a fuzzy logic algorithm is used to diagnose the most likely cause of the problem.
For the Reliant Energy demonstration, the project team selected a set of performance parameters to detect the onset of performance degradation in the HP steam turbine from solid particle erosion, leakage, deposits and blockages. Typical parameters that can be monitored in real-time by the web-based monitoring system are shown in Figure 3.
For optimal performance of any diagnostic performance system it is essential that the information being supplied by the sensors is valid. At Reliant Energy, data validation algorithms tested each diagnostic parameter to identify gross deviations from the expected operating range. The status of sensor validation, for each monitored parameter, is available on the web page.
As mentioned previously, the performance of power plant equipment subject to variable speeds and loads is difficult to trend. However, with the Reliant Energy demonstration project, this was overcome by correcting the performance data to their equivalent values at a standard speed or load conditions. These corrections are performed using polynomials that represent baseline mean values of performance parameters over the operating range while the equipment is in a healthy condition.
Before the monitoring system is able to detect operating problems in a turbine, a set of error patterns must be determined through modeling or by analyzing different failure modes. For the web-based steam turbine monitor at Reliant Energy, computer simulations of a similar unit were run using a through-flow, HP section performance model. This model had the ability to simulate solid particle erosion and deposit damage in the HP turbine, seal leakage or flow blockage anywhere in the unit.
Although no specific operating problems were picked up during the demonstration and testing of the steam turbine monitor at the Reliant plant, simulated tests were able to detect solid particle erosion. Figure 4 shows a simulated normalized performance error pattern for solid particle erosion versus an ideal fault error patterns.
According to Roemer, the on-line performance monitoring and condition assessment software is custom designed for a specific application and not available “off the shelf.” Under an EPRI contract, Impact Technologies are continuing software development for other power plant applications.
According to Bill McGinnis, an engineer with Reliant Energy, the demonstration project was successful. Nonetheless, because of budget restrictions, Reliant Energy did not install a permanent on-line system. However, Reliant Energy is in the process of evaluating a similar system for a combined-cycle power plant in Nevada.
Central Monitoring by Nova Scotia Power
Nova Scotia Power, Halifax, Canada, has recently installed a PI System supplied by OSIsoft, San Leandro, Calif. to collect more than 70,000 points of information from the utility’s generating, transmission and energy control centers. Half of the data supplied is in real-time.
According to Tom Robertson, business information manager of power production, Nova Scotia Power, the aim is to collect data from all of the utility’s generating, transmission and distribution operations and store it in one data warehouse at a central location. Currently the PI is monitoring 90 percent of the parameters that will eventually to be monitored, says Robertson.
On a monthly basis the utility is required to send environmental compliance reports on five of their plants to the Canadian Department of Energy. With PI, these reports are automatically generated. PI also generates comments concerning any excursions. These comments are merged with the compliance reports at the end of each month.
Using their own engineering staff, Nova Scotia Power has developed applications that allow PI to calculate the heat rate and standardize the control of their power plants. In addition, an existing unit efficiency monitoring system is being upgraded to work with PI. Robertson says that when this phase is completed it will simplify the architecture, make program changes easier and make the results more accessible than in the past.
The PI System has been accepted by the senior management of Nova Scotia Power and is now considered a strategic tool for helping them track the performance and manage the costs of their generating, transmission and distribution systems, says Robertson.
Today, Nova Scotia Power’s plants are operating more efficiently because everyone-operations, engineering and management-is able to monitor the same information on-line at the same time. PI is helping Nova Scotia Power optimize power plant operations by allowing them to determine if they have chosen the right fuel and if they are giving priority to putting the low cost plants on-line. In Robertson’s opinion this type of system is essential in today’s business environment.