Jon L. Giesecke
JLG Associates LLC
Table of Contents Summary
Aging transformers are an issue that must be understood and addressed. Ignoring the inevitable is unacceptable, period!
Infrared Thermography, Ultrasonic Noise Analysis, Partial Discharge Detection, Dissolved Gas Analysis, Vibration Analysis. These technologies are great stand-alone tools; however, when used properly and the data combined, you can identify an incipient fault long before it degrades the insulation and creates a failure.
This paper will provide guidance in setting up a complete PdM program to be able to provide owners of large oil filled power transformers, i.e. utilities, refineries, military, mining, etc. with a complete health report and condition assessment of critical oil filled power transformers and ancillary substation components. The testing described in this paper is done on energized, fully loaded transformers. No clearance or blocking is needed to accomplish these tasks. All tests are completely non-intrusive.
Keywords: Partial Discharge, Condition Assessment, Risk Analysis Tool, Cost Benefit Analysis
As transformer experts (substation guys) our job is to maintain the highest level of operational capability for our transformers and substations. A coordinated program of online and portable monitoring must be used to make sure that we are doing our part. The critically of each transformer in your care should be ranked and used to provide guidance to your PdM team on just what is the correct level of monitoring. The lack of spare transformers and the lead time to have a new one built and on site is a consideration in the overall process. Load growth coupled with aging transformers is a disaster waiting to happen.
The Technology Advances: Infrared Thermography, Ultrasonic Noise Analysis, Partial Discharge Detection, Dissolved Gas Analysis, and Vibration Analysis…. All very high tech and necessary to determine the health of our transformer fleet. Each of the listed technologies is proven very effective. However, combining the information from each technology will provide answers to those who ask, “How much life is left in my transformer”. For the last 22 years I have dedicated my career path to the use of technology to determine the on-line condition assessment of power transformers and ancillary support equipment. Along with determining the remaining life is the issue of an action plan to repair, replace or continue to trend.
Twenty-two years ago when I switched from transformer maintenance to transformer predictive maintenance, the advancement of technology has exploded. The average utility is still doing transformer inspections as it did 22 years ago. Consider what information a transformer is willing to give up. It will tell you what is wrong if you are willing to listen. A transformer makes noise in both the sonic and the ultrasonic ranges. A transformer gives off heat which must be removed during operation. A transformer sends signals of impending insulation failure while in operation.
This article will provide knowledge about the latest methods and guidance in setting up a quality program. The goal is to be able to give a complete health report and condition assessment of critical oil filled power transformers while they remain in service. The process described in this paper is done on energized, fully loaded transformers. No clearance or blocking is needed to accomplish these tasks. All tests are completely non-intrusive.
The value of this Excel Spreadsheet is priceless. It has formulas built into each cell that do the calculations on the risk of each condition and how it affects the health and life expectancy of each transformer entered on the spread sheet.
Analyzing Data from History, Coupled with Field Survey Information
The value of this Excel Spreadsheet is priceless. It has formulas built into each cell that do the calculations on the risk of each condition and how it affects the health and life expectancy of each transformer entered on the spread sheet. Once the data is entered from your filing cabinet, maintaining the spread sheet is simple. Data gathered while doing inspections is entered and any changes to the DGA are also entered. The age is changed on all transformers by one click of the mouse. There is no limit to the number of transformers that can be watched simultaneously. The grade is presented numerically as well as a letter grade. A grade is also given to indicate the highest possible score achievable when all deficiencies are corrected.
Determining the health of a transformer is a process that can be the difference between a transformer’s long life and an early death.
Do You Think That Waiting Until a Disaster Strikes Is the Time to React?
The aging of the electrical infrastructure in North America is a critical problem that each one of you faces. There is no way to get around that fact. As aging transformers continue to fail, a new level of awareness of the magnitude of the situation becomes very clear.
Many transformers fail unnecessarily. Proper care and condition assessment of these valuable assets is needed now more than ever. With the transformer fleet’s average age of over 40 years and the new transformer fleet having a higher-than-expected failure rate, a proactive approach to PdM (Predictive Maintenance) is needed. Large power transformers are not off-the-shelf items and must be ordered one to two years in advance. Many are not being manufactured in the U.S. or Canada. The repair and replacement schedule is critical in most cases. These facts make knowledge of transformer condition of utmost importance.
The crucial nature, fragility, age and long lead time for major components and the interconnection of the grid’s electrical system, demand that the best maintenance approaches possible be applied to help ensure reliability.
Transformer Life Cycle Management/ Risk Management
Determining the health of a transformer is a process that can be the difference between a transformer’s long life and an early death. Certain random failures can occur any time and with little or no warning, but as a transformer ages, there will be measurable warning signs that foretell the cause(s) of impending failure. The insurance industry states that insulation failure is the number one cause of transformer failure. So how do we determine the insulation quality while these transformers remain in service?
Examine What You Are Doing Now!
A typical good inspection program includes the following:
Note: Most utilities are doing this inspection process.
- Dissolved Gas Analysis
- Offline Electrical Testing
An enhanced program includes three other components:
Note: By adding these to your existing program will greatly reduce the unexpected failure rate.
- Partial Discharge Monitoring and analysis. (portable and on-line)
- Vibration Analysis (determine core and coil assembly tightness)
- Sound Level Measurements (pre cursor to looseness)
- Grading Method (transformer ranking tool)
- Template Building (Tier assignment done here)
Combining Data from Several Technologies will Provide Information
Very rarely will a failure occur without first revealing some small change that is detectable utilizing one or more technologies. Having stated this, information provided in this paper will direct and assist the reader to starting an inspection process or expanding your existing PdM program.
This paper is designed to show the benefits of doing a complete Transformer Condition Assessment (TCA). Combining the data from various technologies will provide insight and understanding of often subtle, pre-failure signs. In addition, a complete TCA will provide an as-accurate-as-possible gauge of the health of the transformer’s subsystems, including: pumps/cooling system, Load Tap Changer (LTC), De-energized Tap Changer (DETC) and lightning/surge arresters.
Partial Discharge (PD) is unwanted electrical activity. PD is similar to corona activity and occurs at high voltage sine wave peaks. Most low-level PD activity is load-dependent. As the load increases, the voltage decreases. When the voltage decreases, the PD will decrease or disappear completely, and then return when the voltage returns to full value.
Up to 80 percent of all oil-filled power transformers exhibit some PD. Low-level PD activity sometimes continues for the entire life of a transformer. When insulation breakdown from PD gets to a point that it threatens the life of a transformer, a decision must be made—whether or not to remove the transformer from service.
Because PD is present in so many transformers, knowing the present condition is critical. Without systematic TCA, there is insufficient information available to confidently decide when to take appropriate action. In a quest to determine unwanted activity, data can be gathered at a moment in time, like taking a snapshot, or continuously over 24 hours or more, like making a movie. Movies generally tell a more complete story.
By adding the previous steps to your TCA process, major failures will be averted. Major money will be saved, and a major safety feature will be built into every visit to the high-voltage transformer yard.
Detecting Acoustic and Electrical Problems on Energized Equipment
The following PD test described is used to determine the severity of an electrical fault using the burst interval of the PD pattern captured by the high-frequency current transducer. Then, if the source of the fault is located in an area where its sound reaches the tank wall, acoustic sensors triangulate the exact spot of the fault. Fault sound reaches the tank wall about 90 percent of time. About 10 percent of faults are deep within the core and coil assembly, and the sound cannot be detected externally. In these cases, at least the severity of the fault and the fact that the fault will require some work deep within the windings is determined.
Acoustic tests have been used for many years to detect and locate partial discharges in power transformers, but the addition of High Frequency Current Transducers (HFCT) installed on the case ground of the subject transformer make the process complete. It is more difficult to determine if a problem in an oil-filled transformer is related to mechanical or electrical malfunction utilizing acoustic sensors alone. Partial discharge testing using both acoustic sensors and an HFCT makes the determination easy and increases the protection factor for these utility industry assets.
Very rarely will a failure occur without first revealing some small change that is detectable utilizing one or more technology.
The following screen shot shows acoustic and electrical activity obtained simultaneously. The top portion of the screen shows recorded acoustic sensor data, and the bottom shows electrical data from the HFCT. The software user can use the computer cursor to determine the time difference between each burst or burst interval. In this example, the spacing of both the Acoustic Emission (AE) and HFCT sensors is 16 milliseconds. This indicates that the activity takes place at the voltage peak of each cycle. The test equipment used for this data collection is the TP500A from PowerPD Inc.
Asset managers need to know what to do and when to do it to be able to avoid impending failure. The ability to trend the deterioration process aids the asset manager in deciding when to take action.
The following screen shot shows acoustic and electrical data measured for amplitude and duration. This is easily trended by comparing subsequent test results under similar conditions.
The top window in this shot shows the AE bursts with sensor #2 (red) being closest to the source. The bottom window shows the electrical burst captured from the case ground lead using the HFCT.
The signature captured by the HFCT, in the bottom portion, indicates a severe case of PD. The spacing between the end of one burst and the beginning of the next burst, called burst interval, is getting dangerously close to 2 milliseconds, which indicates that a failure is imminent. Burst interval is critical information in determining the severity of PD.
Lightning Arrester Testing (Energized at Full Voltage)
The following two screen shots indicate internal arcing in a 500kV lightning arrester. The HFCT was located on the arrester ground lead above the strike counter for these shots. Multiple data captures on this lightning arrester were all different. No PD was detected by the tests, only arcing.
Normally, lightning arresters have no activity; if this arcing problem was not detected and corrected early, the fault would likely have resulted in a catastrophic failure. This arrester was consequently removed from service, tested off line and disassembled. Off line testing was done at 10kv and did not find the problem. Tear-down found evidence of moisture ingress.
Vibration and Sound Level Analysis of Transformer Main Tank
Vibration analysis is advantageous because it is noninvasive and done on-line while the transformer is under load. In order for a transformer to withstand through-faults or switching surges that include heavy load conditions, the core and windings must be securely blocked and clamped to prevent movement, shifting or distortion. Clamping pressure must be maintained to prevent looseness of core and winding. Deterioration of the pressboard due to moisture or heat may cause shrinkage and looseness. Trending vibration and sound level data is critical to gauge the health of a transformer.
Acquiring Vibration Data
An accelerometer attached to a magnetic base is used to collect vibration signals; this data is stored in a vibration instrument then downloaded to a computer for analysis with standard vibration software. Eight data points are taken on each transformer; vibration sampling results are commonly graphically displayed as waterfall plots, such as those shown in Figures 5 and 6. Starting on the high-voltage side, arbitrarily named side #1, and moving counter clockwise, data is acquired from two points on each side, or wall, of the transformer. The exact data point locations are determined by the size and configuration of the transformer, either core form or shell form. It is crucial that the data is gathered by experienced personnel and taken at the correct locations for each type of transformer.
The ideal spectrum of a steady-state vibration signal from a healthy, tight transformer will contain frequencies that indicate a normal signature. First and foremost, 120 Hz pressure waves are detected; these pressure waves are two times 60 Hz line frequency; i.e., each 60 Hz shift from positive to negative creates a 120 Hz pressure wave that travels through the core, blocking and oil to the transformer wall. Harmonics of 120 Hz will also be detected. It is the combination of data and the shifting of energy that indicates whether a unit is tight or loose.
Analyzing Vibration Data
Recognizing the symptoms of core or blocking looseness is imperative in diagnosing transformer condition. Original methods of transformer vibration analysis considered only amplitude. This was based on severity criteria measured in inches per second. Subsequent research has shown that frequency shifts point toward core and winding looseness regardless of amplitude.
Case Study 1: Normal Shell Form Transformer Waterfall Plot
The transformer in Case Study 1 is a fully loaded 456 MVA Generator Step-Up (GSU).
Case Study 2: Loose Core Form Transformer
The vibration data in Case Study 2 is from a very small core form transformer. The load at the time of this test was unknown, but average load is 80 to 120 amperes.
Notice that 120 Hz has dissipated, and the energy in the spectrum has shifted to the right to 240 Hz. Average sound level sampled during this test was 81 decibels (db)—a level that is 8 to 12 db higher than similar units at the same load.
Avoided Cost or Cost Benefit Analysis:
The measurement of dollars saved or costs avoided have been an ongoing issue since the beginning of time. Doing the analysis can be very time consuming and basically not fun. The process can be streamlined by doing several assumptions. Assumption #1…if a failure was detected and fixed prior to a shutdown or catastrophic failure, that would cost less than if the component was allowed to fail. Assumption #2…allowing the component to run to failure would cost more than the diagnostics and repair combined. Assumption #3…certain failures are a safety concern and could result in bodily harm to workers in the vicinity. A simple work sheet has been developed to provide an easy method for determining a cost benefit. Three scenarios are developed using tribal knowledge about the component in question. Each of the scenarios are assigned a percentage of chance based on the experience of the failure and the individual doing the analysis. The work sheet takes into account what was actually spent to repair the component and provides the final savings number to be used to show solid evidence that a PdM program is worth the effort and has indeed paid for itself thru avoided costs.
Combining results from these test methods to what you are currently doing and using the data from all the available technologies greatly improves your success of “life cycle management.”
The realities are:
- The risk of a failure is real; managing that risk is a full time job
- Aging electrical equipment may fail
- These failures will cost time and money
- A small investment in the latest technology will pay for itself quickly
- A quality PdM program controls costs and prevents unexpected failures