Pricing Gasoline When the Pumps Are Running on Backup Electricity Supply

I attended the MIT Club of Washington Seminar Series dinner on Tuesday, 2014 February 11, which this year is on the topic of “Modernizing the U.S. Electric Grid,” listening to Michael Chertoff talk on “The Vulnerability of the U.S. Grid.”

Chertoff’s maguffin was a story about a hurricane hitting Miami in about 2005.  Electrical workers couldn’t get to work because they had no gasoline for their cars.  The gas stations had gasoline but no electricity to pump the gasoline.  Back-up electricity generators would have required an investment of $50,000 which was not justified on the razor thin margins on which most gas stations operate.

The gas station owners thought process was that the sales lost during the blackout would just be gasoline that would be sold after the power came back on.  Investment in a back-up generator would not change the station’s revenue and would just hurt its profitability.  My first comment during Q&A was that the same issues were raised after Hurricane Sandy[1] in the New York City area in 2012, and perhaps in many other areas that experience wide spread storm damage.

After the dinner I talked with Matthew, a friend from ExxonMobil who had learned about the Seminar Series from my advertizing it to people who attend events of the National Capital Area Chapter of the U.S. Association for Energy Economics.  Because of that linkage, he makes a point to search me out at each Seminar Series dinner.  Our after dinner discussion focused on how to make the $50,000 investment in a back-up generator profitable to the gas station owner.

Matthew said that many gas station permits including anti-gouging provisions, preventing the gas station owner from increasing the price during emergencies.  My thought was that the investment in back-up power supplies would mean that a temporary price increase could be justified to pay for such an investment.  After all, bulk electricity prices in Pennsylvania on the PJM grid during the cold snap associated with the 2014 January arctic vortex soared to $1,839.28/MWH ($1.84/KWH) from an average of only $33.06/MWH during 2012.  This was a temporary 55 fold (not 55%) change in the base price of electricity.[2]

I believe that prices are sticky.  Once set, prices tend to stay unchanged for significant periods of time.  The independent system operators (ISOs such as PJM) get around some of this stickiness by having elaborate models for setting prices every hour, with the basic mechanism setting a value every five minutes and then averaging those five minute values over an hour to get a price.  The basic mechanism includes (1) bids by suppliers as to the price they want if they are to provide specified amounts of electricity and (2) estimates of the demands that will occur each hour or that are occurring on a real time each five minutes.

Almost 25 years ago, long before the advent of ISOs, I published my first article[3] on using the measured real time imbalance between supply and demand to set the real time price for unscheduled flows of electricity.  Using the measured imbalance eliminated the need for bidding processes, bidding process that can lead to stickiness.  I proposed using the concurrent system frequency for setting the price, calling the concept Wide Open Load Following (WOLF).

For electricity, a surplus of demand will drag down system frequency, which I say warrants a higher price, at least higher than the nominal price.  A surplus of supply will push up system frequency, which I say warrants a lower price, at least lower than the nominal price.  Over longer intervals, imbalances will change the accuracy of wall clocks that use system frequency to determine the correct time.  Thus, WOLF includes the concept of time error in setting the nominal price for electricity imbalances.  The WOLF concept could similarly be used to set prices within each of the five minutes of an ISO dispatch period, or even on a sub-minute basis, modifying the ISO’s sticky five minute nominal dispatch value.

The State of California has variable pricing for its State Route 91 Express Lanes under the rubric of congestion management.

“On July 14, 2003, OCTA adopted a toll policy for the 91 Express Lanes based on the concept of congestion management pricing. The policy is designed to optimize 91 Express Lanes traffic flow at free-flow speeds. To accomplish this OCTA monitors hourly traffic volumes. Tolls are adjusted when traffic volumes consistently reach a trigger point where traffic flow can become unstable. These are known as “super peak” hours. Given the capacity constraints during these hours, pricing is used to manage demand. Once an hourly toll is adjusted, it is frozen for six months. This approach balances traffic engineering with good public policy. Other (non-super peak) toll prices are adjusted annually by inflation.

“Recent customer surveys indicate that 91 Express Lanes users lead busy lives with many hours dedicated to commuting to and from their jobs. About 85 percent of customers are married, with more than half raising children. Many customers choose the toll road only on days they need it most, joining general freeway lane commuters on other days. Customers emphasize they value a fast, safe, reliable commute and the toll policy strategy is designed to support this value.

“The toll policy goals are to:

  • Provide customers a safe, reliable, predictable commute.
  • Optimize throughput at free-flow speeds.
  • Increase average vehicle occupancy.
  • Balance capacity and demand, thereby serving both full-pay customers and carpoolers with three or more people who are offered discounted tolls.
  • Generate sufficient revenue to sustain the financial viability of the 91 Express Lanes.

“The effect of the toll policy has been an increase in customer usage with sufficient revenue to pay all expenses and also provide seed funding for general freeway improvements. Revenues generated by the toll lanes stay on the SR-91 corridor, a significant departure from past practices. Under the previous owner’s agreement with Caltrans, a “non-compete” provision restricted adding more capacity to the SR-91 corridor until 2030. When OCTA purchased the lanes, it opened the door for new improvements on SR-91 by eliminating the non-compete provision.[4]

The free flowing capacity of the 91 Express Lanes is 3400 cars per hour.  When average hourly volume exceeds 3200 cars per hour (about 94.1% of the free flowing capacity), the price increases by $0.75 at the beginning of the next six months.  When average hourly volume exceeds 3300 cars per hour (about 97.1% of the free flowing capacity), the price increases by $1.00 at the beginning of the next six months.  When average hourly volume is less than 2720 cars per hour (80% of the free flowing capacity), the price decreases by $0.50 at the beginning of the next six months.  The flow analysis is done for each hour of the week, producing 168 distinct prices each way on the 91 Express Lanes, that is, for 24×7 distinct hours each way.  But as of 2013 July 1, about 1/3 of the hours are charged the minimum price, that is, they are not considered to be super peak hours.   The flow analysis also is separately done for holidays, nominally as minor as Mother’s Day.

The 91 Express Lanes toll mechanism shows that some jurisdictions, including the notoriously protectionist State of California, allow incentive pricing for congestion management during critical periods, such as a wide spread blackout.  The 91 Express Lanes toll mechanism also provides a mechanism for automatic adjustment of the  price.  The 91 Express Lanes toll mechanism uses explicit measurements of the balance between supply and demand, much like the WOLF mechanism for electricity imbalances.  The 91 Express Lanes measurement is the fraction of the capacity of the 91 Express Lanes, changing the price when the hourly utilization is outside the band of 80.0% to 94.1%.

Based on a review of the 91 Express Lanes toll mechanism, there is some hope that gas stations will be able to afford the major investment in backup electrical supplies.  For gas stations, the measure of the imbalance between supply and demand can be as simple as the length of the line of cars waiting for gas or as complex as including the gasoline inventory compared to the desired level and the estimated time before the inventory is extinguished.



[1] Presentation of Adam Sieminski, Administrator of the U.S. Energy Information Administration at the 2012 October 19 lunch of the National Capital Area Chapter of the U.S. Association for Energy Economics (NCAC-USAEE.org).  Pursuant to its charter as an information agency, EIA created for Hurricane Sandy a real time display of gas stations with internet connectivity, a nominal measure of whether the gas station had electricity.

[2] PJM differentiates prices geographically.  Thus, one local price increased to $2,321.24/MWH and another fell to a negative $391.14/MWH because of transmission constraints.

[3] “Tie Riding Freeloaders–The True Impediment to Transmission Access,” Public Utilities Fortnightly, 1989 December 21.

[4] https://www.91expresslanes.com/policies.asp

2012 Washington, DC, Area Storm Electricity Outage Duration

The Washington, DC, area was hit by two storms in 2012, each of which cause widespread electrical outages.  A derecho hit the evening of June 29.  Hurricane Sandy hit four months later on the evening of October 29.  I sent surveys to people on the mailing list of the National Capital Area Chapter of the U.S. Association for Energy Economics (NCAC-USAEE) for both storms asking for a reply about the number of hours they were without power.  (I experienced nine hours for the derecho and two hours for Hurrican Sandy.)  My “Derecho Outage Survey” was included in USAEE Dialogue, Volume 20, Number 3 – 2012.  Here I report on the outages related to Hurricane Sandy and compare the results to the outages related to the Derecho.

Table 1 replicates the form of Table 1 from my “Derecho Outage Survey.”  There were 93 survey responses, which I have summarized for five of the electric utilities in the Washington, DC, area.  I report the number of customers who reported an outage to me and the duration of those outages in hours.  I also include the number of customers who reported that their outage time was zero.  Many of these zeroes may have actually been an outage of a few seconds to a few minutes.  None of PEPCo’s customers in Washington, DC, reported an outage.   I note that 22 PEPCo’s customers in Washington, DC, did participate in the survey, though each reporting 0 hours of outage.

Figure 1 presents a cumulative distribution function for the outages, including separate plots for BG&E; PEPCo-Montgomery County; and VEPCo; as well as a plot for the combination of all of the outage data.  Each plot has a convex shape, showing a rapid increase in the number of customers who have been returned to service, with a gradual degradation of the response rate as time accumulates.

The phenomenon of a convex cumulative distribution function is often referred to as “low hanging fruit” or “the most bang for the buck,” reflecting the incentives and policies that utilities have to concentrate on returning the most customers to service as quickly as possible.  Thus, problems that can be resolved quickly for large numbers of customers are the first problems to be attacked.  The result is the rapid early increase in the number of customers who are returned to service.  Problems that affect individual customers are the last to be resolved, resulting in the plots turning horizontal as the outage duration time increases.

Figure 2 presents the cumulative distribution function for the outages of BG&E, comparing the outages for Hurricane Sandy with the outages for the derecho.  The plot for the four outages associated with the derecho does not have the convex shape described above.  But with only four outages reported, the distribution of reported outages has a greater chance of not being representative of the distribution of actual outages.

Figure 3 presents the cumulative distribution function for the outages of PEPCo in Montgomery County, comparing the outages for Hurricane Sandy with the outages for the derecho.  PEPCo took some 72 hours to restore half of the customers who had outages associated with the derecho compared to only 3 hours to achieve the same restoration level for outages associated with Hurricane Sandy.

Figure 4 presents the cumulative distribution function for the outages of VEPCo, comparing the outages for Hurricane Sandy with the outages for the derecho.  VEPCo took 40 hours to restore half of the customers who had outages associated with the derecho compared to only 12 hours to achieve the same restoration level for outages associated with Hurricane Sandy.

Figure 5 presents the cumulative distribution function for all the outages reported in the two surveys, comparing the outages for Hurricane Sandy with the outages for the derecho.  The utilities in the DC area took 55 hours to restore half of the customers who had outages associated with the derecho compared to only 9 hours to achieve the same restoration level for outages associated with Hurricane Sandy.

The DC area electric utilities were severely castigated by government officials for the length of time that restoration efforts took after the derecho.  Few comments were made about the restoration time in regard to outages in the DC area for Hurricane Sandy.  Some of the differential in the length of the outages associated with the derecho versus Hurricane Sandy relate to physical phenomenon.  Some relate to planning by the utilities.

The derecho occurred four months before Hurricane Sandy.  The derecho is likely to have toppled many of the trees that Hurricane Sandy would otherwise have toppled.  The utilities in the DC area also initiated a substantial tree trimming program after the derecho, further reducing the number of trees that would otherwise have been grist for Hurricane Sandy in causing outages. The derecho may also have had stronger winds.  Together, these items are reflected in the large increase between the two surveys in the number of responses that showed no outages.

At least one utility in the DC area, PEPCo, pre-positioned contractors before Hurricane Sandy, as I wrote while sending out the survey.

When my wife and I drove to church Sunday morning, I was impressed with the dozen or more bucket trucks sitting in the parking lot of the Gaithersburg Holiday Inn, thinking I should call the television stations as a potential story for them to film. I didn’t, but I was no longer impressed by yesterday’s sight when I saw Channel 4’s story today about noon from Gaithersburg, just across the street from the Holiday Inn. The Montgomery Country Fairgrounds seemed to have over a hundred bucket trucks, making the Holiday Inn parking lot scene look insignificant.

These pre-positioned contractors would likely have reduced the duration of the outages, just as the relative timing of the two storms, the tree trimming programs, and the relative strengths of the two storms likely contributed to the reduced fraction of customers who reported any outages and the increased fraction of customers who reported no outages at all.

Pre-positioning contractors comes with a cost.  The contractors I saw at the Holiday Inn were in the DC area two days before Hurricane Sandy hit.  Some of that time might have been non-productive.  Some of the time might have been used for additional tree trimming and other normal on-going work that utilities do on a routine basis.  Even the contractors shown on TV seem to have been pre-positioned a day ahead of time.  In contrast, the derecho was not anticipated and contractors generally travelled during the first day after the derecho hit instead of one or two days before Hurricane Sandy hit.