There is not a utility vegetation manager in the country that will say he or she has a sufficient hazard tree budget. Oftentimes, more hazard trees are along a given distribution circuit than funding available for their removal. Utility vegetation managers constantly have to manage the risk of tree failure by minimizing both the number of trees that cause distribution service interruptions as well as the impact or severity of those service interruptions on the system.
Once granted the funds, vegetation managers want to use those limited hazard-tree dollars to produce a measurable impact on their utility’s system average interruption frequency index (SAIFI) targets. They also need to show a measurable reliability improvement, at least at the circuit level, to receive equal or greater funding in subsequent years. Significant reliability gains at the circuit level eventually will manifest in reliability improvements in the utility-level metrics. National Grid has developed a process for targeting hazard-tree work to produce the biggest bang for the buck.
Select the Right Circuits
Assuming that the hazard-tree budget is insufficient for working on more than 5% to 10% of the total overhead distribution circuit population, the first step is to develop a selection scheme that targets the work on the right circuits. In some cases, a utility regulator may have criteria that must be employed each year, often known as the worst-feeder list. If that is true for a given utility and the worst-feeder list requires all of the utility’s budgeted hazard-tree dollars, then this first step is complete. However, assuming there is not a worst-feeder regulation or more money is available, a reliability/risk modeling technique will be necessary to focus hazard-tree spending on the right circuits.
At National Grid, a blend of three leading, or risk, indicators and two lagging, or performance, indicators is used in the circuit selection model. These indicators, or indices, are calculated for the entire circuit population. The first leading indicator is the number of customers served, or at risk, for each circuit. Obviously, this indicator will drive circuits serving larger populations to the top of the ranking as an interruption on these circuits would have the largest impact on system SAIFI.
The second leading indicator is the number of miles of bare-wire, three-phase construction on each circuit. This configuration is the least tolerable to tree/limb contact, so the indicator pushes circuits with more bare-wire, three-phase construction toward the top of the selection list.
The third leading indicator is miles of tree exposure. This is an obvious risk indicator for tree interruptions, and use of this indicator forces heavily treed circuits upward in the ranking model. Note that, by utilizing new geographic information system (GIS)/satellite technologies, indicators two and three can be combined as a single, more meaningful index to provide miles of bare-wire, three-phase construction with tree exposure.
The two lagging indicators, the fourth and fifth of all the indicators, are both tied to tree-related interruption performance data. The fourth indicator is the average number of tree-related customers interrupted (CI) calculated over a three-year period. The use of this interruption data moves circuits exhibiting more frequent or larger tree interruptions upward in the ranking model.
Lastly, the fifth indicator uses an average tree-related CI per tree event. The use of this indicator in the model pushes circuits with more frequent, historic three-phase tree interruptions higher in the rankings.
National Grid’s circuit selection model uses five indicators, or indexes:
Leading, or risk based
- Customer served on the circuit
- Miles of bare-wire, three-phase construction
- Miles of actual tree exposure
Lagging, or performance based
- Three-year average total tree CI on the circuit
- Three-year average tree CI per event.
After calculating each of these indices separately, they are summarized to provide a final rank for the circuit. Circuits are then picked from the top of the list and one final sanity check is done with a review of the previous three years of interruption history to look for any odd situations that have inappropriately placed this circuit too high in the rankings. Finally, an annual work plan is created.
Additional options should be considered when using the circuit selection model. Firstly, the indicators can be weighted to lend more impact to one or more indices. This decision would likely depend on the utility’s current program maturity, construction standards and local vegetation characteristics. For example, a more mature program with a fewer number of problem circuits may want to rely more heavily on the leading indicators focusing on potential risk rather than on historic performance.
Secondly, in a program where hazard trees are only performed on the circuits being pruned that year, the circuit population included in the model should only be those circuits on the pruning schedule for that year’s cycle. Other possibilities would be to include only circuits that are at the mid-cycle point in their pruning schedule. For a vegetation program on a relatively long cycle length, like five years to six years, this would provide a mid-cycle inspection and put crews back on the circuit performing hazard-tree work and possibly whatever other vegetation work that might be necessary to maximize the cycle period.
Partition the Circuit
What is obvious to some may have to be proven to others. Studying tree interruption data certainly supports this approach of partitioning the circuit by customers at risk. The average customers interrupted per tree event can be calculated using two key annual tree interruption metrics: total customers interrupted because of trees and total number of tree interruptions. Table 1 shows annual reliability metrics for an unnamed utility service territory.
| Tree interruptions | Customers interrupted | Average customers interrupted per interruption |
|
3.132
|
220,748
|
71
|
Making a second cut at the data can yield some very useful information when beginning to look at targeting hazard-tree dollars for the most effective results on the circuit. Taking the same data, develop the same average customers interrupted per tree interruption but use only the metrics for interruptions that caused station breaker operations, also known as circuit lockouts. Table 2 is an example of this second step in the analysis, which creates a subset of the data in Table 1.
| Tree interruptions | Customers interrupted | Average customers interrupted per interruption |
|
88
|
113,701
|
1,292
|
Quickly, one can see tree interruptions that cause station breaker operations are the most detrimental to the utility’s SAIFI index. As shown in Table 2, 88 interruptions, which are only 3% of the total tree interruptions for the year, interrupted 52% of the customers.
These interruptions occur on what can be termed the lockout section of the circuit, that is, the section of the feeder from the station breaker out to the first protective device. This is the beginning of the circuit partitioning concept. Every circuit can be broken down into segments based on the location of protection devices (re-closers, sectionalizers, fusing). In addition, usually the customers protected or customers at risk beyond each protection device on a circuit can be established using the utility’s geographic system if the circuit is modeled electrically in the system.
Assigning Tree CI or SAIFI Values
With the customer’s at-risk data, each potential tree interruption or each hazard tree to be removed can be assigned a CI or SAIFI value. That value is really a CI-saved” or SAIFI-saved value for each of the trees to be removed based on their location along the partitioned circuit. To illustrate this concept, refer to the simple distribution circuit based on a utility that serves 1.5 million customers.
Table 3 points out the significant differences in potential SAIFI benefit depending on where hazard-tree funds are spent. Assuming that, regardless of the location of the tree on the circuit, the overall average cost to remove the tree is reasonably equal, the dollars per expected delta CI can be calculated for each section on the circuit.
| Section |
Protected by |
Number of customers served or protected |
Tree SAIFI value* |
Number of trees to remove to provide an equivalent SAIFI benefit (nearest whole number) |
|
First or lockout section
|
Station breaker
|
2,800
|
0.0019
|
1
|
|
Between first and second device
|
First protective device
|
1,600
|
0.0011
|
2
|
|
Between second and third device
|
Second protective device
|
820
|
0.0005
|
4
|
|
Beyond third device
|
Third protective device
|
105
|
0.00007
|
27
|
*Assumes a total of 1.5 million customers served.
The partitioning of the circuit by SAIFI benefit easily justifies true reliability work, like hazard-tree removal on the main three-phase sections of the circuit, well before any significant work is invested along the single-phase taps.
National Grid has constructed its hazard-tree inspection specifications to align with the circuit partitioning concept. The specification is constructed so the inspection requirements, or intensity, are adjusted as the inspector moves along the circuit from section to section. In the lockout section of the circuit, the inspector uses all aspects of the hazard-tree inspection analysis. While out on the third or later sections of the circuit, only trees identified as imminent hazards are listed for removal.
For example, in the lockout section, a tree may be listed for removal because of stem damage where part of the bark has been damaged, yet no apparent decay is currently present. That same tree would need to show a substantial cavity and accompanying decay to be listed for removal in the third section of the circuit.
National Grid believes this process works well to target the right circuits and the most effective sections of those circuits for valuable hazard-tree dollars. Annual studies are performed to track the reliability response of the program by circuit. The most recent review was done for 50 hazard-tree circuit projects completed between 2008 and 2012 in the utility’s Rhode Island service territory.
Using three years of pre-project reliability data to develop an average tree CI for each circuit and then comparing that to the first full year of post-project reliability data, a 75% reduction in average tree CI is notable. Even doing the comparison using three years of post-project data on the older projects, a 52% improvement is still significant. Note the reliability data being used in these comparisons excludes major storms.
Certainly, this approach is working for National Grid. The circuit model is updated each year as more usable and detailed data becomes available. The partitioning practice has remained consistent; however, the accompanying hazard-tree inspection specification is updated as better hazard-tree knowledge becomes available through industry research.
The inspection specification update also allows the utility to adjust each hazard-tree issue by circuit location should a change be necessary. For example, as infestations of the emerald ash borer begin to show up in National Grid’s service territories, it is possible all ash species will be added to the list for removal in the lockout section of the circuit. This flexibility keeps the process fresh and up to date with the latest utility arboricultural knowledge.
About the Author:
Craig M. Allen ([email protected]) is the manager of vegetation strategy at National Grid, USA and has been with the utility, and its predecessors, for 31 years. He holds an associate’s degree in forest technology and a bachelor’s degree in forest resource management, both from New York State University. He is an ISA Certified Arborist/Utility Specialist.
