In 1994, Minnkota Power (Grand Forks, North Dakota, U.S.) embarked on a program to improve power quality within its system. The scope of the project was to provide a method of increasing system reliability by reducing lightning-caused outages. The initial plan was to add lightning arresters to the existing 69-kV transmission lines, add fault-locating distance relays and use a real-time lightning-detection system.

Arrester and relay installation began in 1995, and by 1999, Minnkota Power had installed 30 fault-locating distance relays and more than 360 miles (579 km) of arresters to its transmission lines. The utility used the real-time lightning-detection system for one year before it was discontinued because of limited application.

When Minnkota Power implemented the initial plan, it decided it would review the program after a couple of years to determine its effectiveness and to make any necessary modifications or changes. The utility conducted preliminary studies in 1996 and 1999, and more complete analysis in 2001. This current study contains five years of lightning and operations data from 1996 to 2000.

Study Process

Minnkota Power contracted with Minnesota Power (MP; Duluth, Minnesota, U.S.) to use the Fault Analysis and Lightning Location System (FALLS) to correlate which faults on its 69-kV lines were attributed to lightning. The FALLS software uses line location, fault time and lighting data from the National Lightning Detection Network (NLDN) as inputs. Outputs include information on lightning strokes that correlate both spatially and temporally to a fault event including the stroke location, arrival time and stroke current. The software will also provide information on line stroke density, flash density and peak current histograms of area lightning activity.

For study purposes, Minnkota Power provided MP with fault times and geographic information system (GIS) data files for the 69-kV transmission lines. This study included line sections from seven separate breakers across the Minnkota system. Line sections ranged from 20 to 83 miles (32 to 134 km) in length and covered areas in North Dakota and Minnesota. Minnkota's system control center provided fault times listing the dates, times and probable causes of breaker operations.

MP loaded the fault times into the FALLS fault database, converted the GIS files from AutoCAD to MapInfo format and imported them into the FALLS GIS. MP then ran a fault-correlation analysis for each fault. The correlated faults were plotted and characteristics such as peak current and stroke time were exported to an Excel spreadsheet for further analysis.

In addition to the fault analysis, MP completed an annual exposure analysis for each line. This provides information on the number of cloud-to-ground lightning events that occurred in the line proximity. Both stroke data and flash data were used in this analysis, and again, lightning characteristics such as peak current and stroke time were exported to an Excel spreadsheet. In this format, the data can be sorted and categorized to show possible lightning trends or patterns. Lastly, stroke and flash density area maps were plotted on each line. This allowed Minnkota the ability to readily see if some lines, or portions of a line, were more exposed to lightning than other areas.

Fault Analysis

The fault-time data Minnkota provided was accurate only to plus-or-minus one minute. Timing such as this is generally adequate if the only goal is to determine whether lightning caused the fault. However, if the goal is to determine if line performance is improving, it has several shortcomings. One of these shortcomings, for example, is that faults often correlate temporally to more than one stroke. In this example, stroke currents range from -8 kA, which should be below the fault threshold of an arrester protected line, to -33 kA, which is at or above the fault threshold for footing resistances greater than approximately 50 ohms. Since the NLDN typical location accuracy is ±500 m (±1640 ft), it is not always possible to determine which stroke actually terminated on the line and caused the fault.

Over a short period of time, it may be difficult to say for certain that performance has improved without better fault timing. This is because lightning activity can be highly variable, and it will not always be possible to determine which stroke actually caused a fault. However, over a long period of time, a utility can collect enough data so that any improvement in performance will become discernible. Since FALLS provides more than just fault-correlation information, it will reduce the time period required to obtain meaningful results, even without accurate fault timing.

Another shortcoming was the inability to determine exactly where on the line the fault occurred. This is important because portions of some lines did not have arresters. For example, in the figure on the left, the strokes were correlated spatially over three different areas of the line separated by distances greater than 10 miles (16 km). From this, it is not possible to tell if the fault occurred on the portion of the line with arresters, or the unprotected areas.

One method to resolve this issue would be to use relay-fault location data to determine more precisely where the fault occurred. This approximate location will then help to determine which stroke most likely caused the fault and also allow for line inspection following any operations. Minnkota's experience with fault-locating relays demonstrates its accuracy in determining fault locations. Recently, a blown post top insulator was located using the distance indicated on the fault-locating relay. This type of relaying data is instrumental for determining the location of a problem and allowing for quick repairs. Since half of all operations are non-lightning related, it is a good practice to have in place another means of assessing the situation.

Lightning Exposure Analysis

MP provided tables of correlated lightning stroke current to Minnkota to determine if any pattern occurred that would indicate improved performance. For example, the utility can use the table of all correlated fault currents to determine if the typical fault-inducing stroke current increased after the arresters were added.

In addition to the fault correlations, MP also provided Minnkota Power with information on each line's annual lightning exposure, a multi-year exposure summary and a table of stroke currents of all lightning strokes located within 1 km of each line. This information allowed Minnkota to determine if it could attribute any increase or decrease in fault numbers to changes in lightning exposure, which had substantial yearly variations.

Evaluation of FALLS Study Results

The first step in the evaluation process was to categorize the lightning data above or below 50 kA. MP conducted an EMTP study to determine this demarcation line. If the expected stroke current withstand is calculated as modeled in EMTP, then the line-exposure data combined with the peak current information can be used to estimate expected performance. These estimates can then be compared to see if performance has improved based on the transmission line's actual exposure.

Any known line-performance improvement takes into account only those lightning strikes that are deemed preventable because arresters are unable to protect against some of the stronger lightning strikes. The EMTP analysis of Minnkota's structures revealed that 50 kA is the maximum level of arrester protection.

For example, EMPT studies indicated that without arresters, virtually any direct strike to the line would result in a fault. It also indicated that with arresters and a 50 ohm footing resistance, the line could withstand a 30 kA stroke. With a footing resistance of 25 ohms or less — the resistance Minnkota attempted to achieve when installing arresters — the line should be able to withstand a stroke to the line of approximately 40 to 50 kA.

The next step in the analysis process involved conducting a line-by-line assessment of operations. Each year operations were categorized and charted as lightning induced above and below 50 kA, and those that were non-lightning related. The overall results indicated that about 45% of all operations were lightning induced. Of these lightning strikes, 23% were in excess of 50 kA. Taking this 50 kA limitation into account, arresters are capable of preventing about 77% of the total lightning challenges assessed in this study. Therefore, the utility can reduce 77% of the 45% of lightning outages that register less than 50 kA — an overall line-performance improvement of just about 35%.

The magnitude (amperage) of a strike is important to understanding that more lightning does not necessarily mean more operations. The graph above shows that in 1996, 10% of the challenges were in excess of 50 kA, and in 1997, 23% were over 50 kA. This resulted in nearly twice the number of potential operations with nearly the same amount of lightning action. In 1999, there were 60% more challenges than in 2000, yet the increase in challenges over 50 kA was only 30%.

After analyzing several years of lightning data a picture of what is happening within the system each year comes into focus. Taking into consideration whether or not the lightning strike is positive or negative and may also have an impact on system performance. In 2000, there were more storms with higher percentages of positively charged lightning strokes, which indicate possible higher storm winds than in any other year. Extreme wind can cause debris or trees to blow into the line resulting in a fault. But when looking at the data initially, it showed (at least ratio-wise) that there were as many lightning challenges to breaker operations in 2000 as there were in 1999. However, further study showed there were fewer actual lightning-induced operations — the direct correlation between lightning and operations — than the previous year. There was the same ratio of operations/challenges, but in 2000, 10% less were the direct result of lightning. The analysis shows that despite the arrester installations other influences still impact line performance.

Another consideration that Minnkota is still in the process of assessing is to note any specific patterns or clusters in the location of the lightning on each line over the five years. Are there concentrations where lightning seems to strike over and over again? If so, is this a reality or just a trend during this specific five-year period? Nonetheless, it does pose an important question: Can lightning arresters effectively be installed only in a specified area instead of across the whole line?

Looking at any one line segment during any one year gives a glimpse of any possible patterns, but it is the overall compilation of five years worth of data that gives the best view of whether or not adding arresters improves line performance. From reviewing five years worth of lightning data, it is safe to say that, yes, arresters have helped reduce the number of lightning induced operations.

To quantify this improvement, Minnkota developed a ratio of the number of operations to the number of challenges for each year. In 1996, the lines in this study were basically unprotected. In 2000, all but a small portion of the lines had arresters installed. Using this information, the number of challenges per operation (less than a 50 kA) should increase. About 20% of the lightning strikes are going to cause an operation even with arresters.

The graph on the left shows that with each year and the addition of more arresters, the number of lightning challenges that it took to cause an operation increased. In 1996, for every 41 lightning challenges, there was an operation. By 2000, there were 129 challenges before an operation occurred. This data provides a snapshot of how arrester-treated lines performed.

Other Areas of Consideration

For a lightning-mitigation program to be effective, the utility must take additional measures beyond installing lightning arresters. Minnkota identified the following guidelines to make its program more valuable and consistent.

• Maintenance

  Replace failed arresters in a timely manner. At times, failed arresters have gone unnoticed or replacements have been delayed. The integrity of the line is compromised for each arrester not in service. While one failed arrester might not significantly affect a line's performance, if several arresters are not replaced, it could lead to a serious problem.

• Grounding

  The goal is a desired ground resistance of less than 25 ohms. Check and install ground rods as needed when the arresters are installed. If this step is overlooked or not considered, arrester performance will likely be much less than expected.

• Fault-Locating Distance Relays

  Install fault-locating relays with a good means to download and archive the data to correlate the lightning events. Use the relay distances to assist in timely storm inspections and other alternate problems like trees, wind or birds.

Minnkota's lightning-mitigation program has proven that arresters reduce outages on lightning strikes less than 40 to 50 kA, which account for 77% of the area's lightning challenges and 35% of the total line outages. While the arrester program is effective, studies of lightning data and overall line performance need to continue to further improve power quality. Better fault timing is needed to confidently correlate faults with single lightning strokes. Based on MP's experience with global positioning system (GPS) fault timing, the technology is available to increase the accuracy and reduce the time of the analysis. Using GPS technology to improve the program is the next logical step.

David Van House received a BSEE degree from the University of Minnesota. Since 1989, he has worked at Minnesota Power as a transmission-planning engineer in the substation and telecommunications department.

Jennifer Johnson earned a BS degree in industrial technology with an energy/power emphasis from the University of North Dakota. In 1981, she started work at Minnkota Power Cooperative Inc. in substation engineering, and in 1996, she transferred into the transmission engineering department.