Use PQ indices to improve system availability in industrial facilities
As a maintenance engineer for 10 years at a pulp and paper company, I always dreaded the intermittent electrical problem that was reported as the cause for lost production. I knew the drill. This meant that my schedule for the day would quickly become backlogged while I sifted through insufficient data and unfamiliar electrical operating characteristics in an effort to determine what had happened. Now a power quality engineer at a major electric utility in Canada, my main responsibility is to assist industrial and commercial customers with similar quandaries and uncover the culprit behind power quality-related problems.
A recent project with one of the utility's major industrial customers is a prime example of how the use of select power quality indices in a predictive/preventive maintenance program (PPM), coupled with the use of digital metering with power quality recording capabilities, can help industrial facilities develop an effective power quality maintenance program. The intent of this particular program was to remotely retrieve and automatically compile indices to provide reporting based on threshold events and provide histogram trending to be used to determine operational limits.
The most difficult part of implementing this program was to establish limits for trended indices, because there is very little published information on acceptable power quality operating limits for equipment, making this task more of an art than a science. The equipment manufacturers' task of testing the effects of a boundless number of index variations on their equipment is understandably impractical.
The approach of this power quality maintenance program somewhat follows a typical yearly doctor's exam. In my case, the doctor checks my blood pressure against my historical trend data and tells me it's above average published levels. Because it's always been in this range, however, it's considered normal for me. So pending any apparent abnormalities, I'm basically good until next year.
With this type of thinking in mind, we established initial limits for this program based on a number of standards and guidelines, such as IEEE 1159, CSA CAN-3-C235, ANSI/NEMA MG1, ITIC curves, electric utility power quality limits, and the operating history of the monitored distribution equipment. After initially applying these limits — if no power quality abnormalities were apparent with the trended indices — the benchmarked limits were adjusted as required to minimize event reporting.
Differentiating between acceptable and abnormal power quality events did require a degree of experience for expedient analysis. However, as per typical troubleshooting approaches, the investigations were all based on event timing to narrow in on offending equipment or processes. The program used feeder and equipment recorders that had the functionality to trend maximum, minimum, and average power quality indices, such as harmonics, unbalance, sags, and surges. The recorders were networked together and downloaded remotely. A database program was developed that performed statistical calculations, graphing, and exception event reporting on the available data.
Having a full array of indices is beneficial; however, any number available will fit the trending requirement. In this case, our team was fortunate enough to have facilities that placed permanent recorders with power quality capability throughout the plant at all levels of distribution as well as at some critical pieces of equipment. We used the following indices as prime indicators of distribution performance:
Voltage variation (average, minimum, maximum) — This index was trended, typical electric utility planning limits of ±5% were applied for voltages greater than 1,000V, and CSA CAN-3-C235 Preferred Voltage Levels for AC Systems, 0 to 50,000V, Electric Power Transmission and Distribution was applied for voltages less than 1,000V. These limits were never adjusted any tighter than the initial values (click here to see Fig. 1).
Current (average, minimum, and maximum) — This index was not used for alarming but trended to support comparative analysis.
Sag (if available) — This index was extremely crucial to capture, as it is quite often the leading culprit of lost production. IEEE Standard 1159, “Recommended Practice for Monitoring Electric Power Quality,” was used to define the sag level, and these events were loosely compared with the ITIC curve. However, the facility's immunity threshold to sag events was the ultimate limit.
Voltage unbalance (average) — Average voltage variation was trended, and typical limits of between 1% to 2% were used while respecting ANSI/NEMA MG1 Motors and Generators standard. This limit was never adjusted any tighter than the initial values.
Current unbalance (average) — This index was not used for alarming, but trended to support comparative analysis.
Voltage total harmonic distortion (average) — This index (click here to see Fig. 2) was trended, and a typical IEEE 519 planning limit of 5% of the average fundamental voltage was applied. This limit was lowered as we became more confident with the trended bandwidth variability, with the overall intent of minimizing event reporting.
Voltage individual harmonic distortion (average 2nd through 15th) — This index was trended, and typical IEEE 519 planning limits of 3% of the average fundamental voltage were applied. These limits were lowered as we became more confident with the trended bandwidth variability, with the overall intent of minimizing event reporting.
Current total demand distortion (average) — This index (click here to see Fig. 3) was trended, and typical IEEE 519 planning limits were applied based on the short circuit over maximum demand loading ratio (Isc/Il). Within the facility, transformer nameplate kVA was used for both the maximum demand loading (Il) and for normalizing. This limit was lowered as we became more confident with the trended bandwidth, with the overall intent of minimizing event reporting.
Current individual demand distortion (average 2nd through 15th) — This index was trended, and typical IEEE 519 planning limits were applied based on the short circuit over maximum demand loading ratio (Isc/Il). Within the facility, transformer nameplate kVA was used for both the maximum demand loading (Il) and for normalizing. These limits were lowered as we became more confident with the trended bandwidth variability, with the overall intent of minimizing event reporting.
Power [average - energy (W), demand charge (VA) and reactive energy (var)] — These indices were not used for alarming but trended to support comparative analysis.
A customized, off-the-shelf database program was configured as a preliminary platform to perform three major routines: statistical calculations, exception event reporting, and graphing of the recorded indices. The statistical analysis toolbox simply used minimum, maximum, average, percentile, and standard deviation to help weight the selected limits — no statistical expertise was required. The statistical analysis toolbox was applied during the early stages of the program. Once a confident operating bandwidth was evident, the limits were adjusted on an as-needed basis to minimize reporting.
The exception event reporting toolbox identified indices outside of limits, logging events by magnitude and date of incident. The exception event reporting toolbox was continually used to report on out of limit events. Any automatic reporting period or report can be generated from printed report to e-mail notification. The graphing toolbox allowed the graphing of selected indices for comparative analysis.
Interestingly enough, the initial success of this program has been to immediately identify suspect equipment operation and improperly configured metering as opposed to the intent of trending and identifying progressive equipment failure. Typical power quality equipment issues ranged from high harmonic distortion due to capacitor resonance, voltage unbalance due to failed power factor correction stages, high even harmonic current due to faulty rectifier bridges, and multiple sags due to equipment starts. Typical metering configuration issues ranged from incorrect ratios, time stamps, inconsistent trigger settings, and mislabeled equipment. Longer term trending identified electric utility seasonal system feeder configuration changes that were otherwise thought to be constant. These different system feeder configurations changed the applied harmonic voltage spectrum and the facility resonance points.
This program successfully demonstrated a proactive approach to reduce facility downtime while effectively minimizing the resources required to assess the unrelenting amount of data produced by the recorders. Not only did the continuous monitoring strategy benchmark electrical feeder power quality characteristics and report only when limits were exceeded, but it also successfully demonstrated that a power quality maintenance program can be implemented without considerable power quality expertise.
Kizuik is a power quality engineer with Manitoba Hydro, Winnipeg, Manitoba, Canada. He can be reached at firstname.lastname@example.org.