THE RESPONSE OF INDUSTRIAL CUSTOMERS TO ELECTRIC …

Download Report

Transcript THE RESPONSE OF INDUSTRIAL CUSTOMERS TO ELECTRIC …

THE RESPONSE OF INDUSTRIAL
CUSTOMERS TO ELECTRIC RATES
BASED UPON DYNAMIC MARGINAL
COSTS
BY
Joseph A. Herriges, S. Mostafa Baladi, Douglas
W. Caves and Bernard F. Neenan
Presented by:
Ugonna
Objective
• The purpose of this paper is to describe the
results from an analysis of a real-time pricing
experiment recently conducted at Niagara
Mohawk Power Corporation. The objective of the
experiment is to measure the response of large
industrial customers to dynamic, fully timedifferentiated, marginal cost-based electricity
rates over a broad range of industries.
• The electric power industry has historically relied upon rate
schedules with little or no variability over time to reflect the
dynamic nature of its marginal costs.
• Recent advances in technology and in the theory of utility
operations and planning have led to an increased interest in time
differentiated rate options.
• Real-time pricing (RTP) of electricity encompasses a range of
possible service options, featuring prices that reflect the
constantly changing costs of supplying electricity.
• Compared to conventional or time-of-use rates, real-time prices
more accurately represent marginal costs at each point in time.
• To the extent that customers alter usage in response to frequent
price changes, RTP offers significant benefits to both utilities and
their customers.
The Experiment
• Interest in real-time pricing for large electricity customers
comes only a few years after the small-user time-of-use
pricing experiments that were conducted in the 1970s and
1980s. However, there are fundamental differences
between the earlier experiments(small-user time-of-use
pricing experiments ) and current RTP experiments.
• First, because RTP experiments involve large users, both
the utility and the customer risk substantial funds, often
measured in the millions of dollars. This is because, A
sample involving more than a few large customers will
defeat the experiment's purpose of testing response
before risking substantial revenues. Hence, a pricing
experiment for large users must obtain the maximum
amount of information from a small sample of customers.
• Second, the RTP studies are targeted at
customers who possess the commercial and
political wherewithal to frustrate mandatory
participation. Hence customers in each
experiment are volunteers, creating the
potential for self-selection bias.
• Next is a summary of the rate and
experimental designs underlying Niagara
Mohawk's Hourly Integrated Pricing Program
(HIPP).
Rate Design
• The cost of production can change due to changing
weather conditions and capacity availability. The
appropriate price signal in this situation is a timedifferentiated rate, reflecting the variation in
marginal costs, Steiner (1957) and Boiteux (1960).
• The HIPP tariff was designed to provide customers
with hourly price signals set as close as possible to
marginal cost.
• HIPP is a two-part tariff, consisting of a marginal cost
based hourly energy price ($/kWh) and an access
charge that is independent of the customer's current
usage.
Draw Graph
• Figure 1:
illustrates the
pattern of energy
prices during the
first eight months
of the HIPP
program,
compared to the
standard time-ofuse tariff, which
has fixed on-peak
and off-peak
rates.
Experimental Design
• The target population for the experiment is Niagara
Mohawk's large commercial and industrial class, which is
generally billed according to the utility's mandatory large
customer time-of-use tariff.
• Large users were chosen because they are most likely to
generate RTP benefits that exceed metering,
communications, and administration costs.
• Second, they are more likely to invest the resources
required to understand and evaluate the HIPP tariff.
• Third, Niagara Mohawk has historical load research data
for these firms, which are necessary for calculating the
HIPP access charge and for analyzing customer response
to RTP.
•
The design of the HIPP experiment is shown in figure 2.
•
There are two customer groups and two time periods. During the baseline period
all customers were on the standard time-of-use rate.
For the test period, customers who volunteered for the HIPP tariff were
randomly assigned to either the control or treatment groups. It is the volunteer
control group that provides the basis for evaluating customer response to realtime pricing. Among RTP experiments, the HIPP experiment is unique in its use of
a volunteer control group.
•
•
•
•
•
The first set of statistics in table 2 shows that the load growth of the control group, 5.1%, surpassed
that of the test group, 1.5%. Thus, HIPP did not result in an increase in total energy.
The second set of descriptive statistics compares the average price of electricity under the HIPP and
standard rates for both test and control customers using usage levels from the eight test months. As
expected, given the revenue neutrality of the HIPP rate, there is little difference in the average price
of electricity under the two tariffs for control customers. The test customers‘ average price is over
6% lower under HIPP than under the standard rate. Given their modest load growth of 1.5%, this
difference in average price indicates an ability to shift loads away from high priced hours.
The third set of statistics examines test period usage relative to the baseline loads during high
priced hours. The hour of greatest interest to the utility has traditionally been the hour of system
peak. Table 2 shows that during this hour the average test customer reduced loads from their
baseline level by 13.2%, while the average control customer increased loads by 4.5%.
The hour of highest marginal cost did not coincide with the hour of system peak. At the hour of the
peak HIPP price, test customers reduced usage by over 36%, while the control customers reduced
usage by less than 5%.
•
•
.
•
•
Index numbers provide a convenient means of summarizing response
to real-time pricing free of specific production or cost function
assumptions.
1. Index Number Theory: Consider the
Laspeyres price index given by:
where 𝑬𝑲 (t) and 𝑷𝑲 (t) denote the usage and price levels during the
hour t(t=1,…,T) of experimental period k(k=0 for baseline; = 1 for
test). This index, when viewed as an approximation to an underlying
cost function, has the property that it assumes that customers cannot
shift usage among hours. This can be seen in equation (1), since only
the baseline input levels (𝑬𝟎 ) appear.
In contrast, the Fisher price index is a superlative index number
computed as:
The Fisher index includes the expenditures in both periods and
provides a second order approximation to any underlying cost
function with any degree of price elasticity.
In the analysis, the Fisher and Laspeyres indexes are combined to
provide an index of customer response to the HIPP rate structure.
Specifically, the response index is defined as R ≡(𝑭𝟏 /𝑭𝟎 )/(𝑳𝟏 /𝑳𝟎 ).
Dividing the Fisher index by the Laspeyres index adjusts for the
change in the level of electricity prices imbedded in the HIPP tariff
and, thus, isolates the change in unit costs due to load shifting. If the
firm does not respond to the HIPP price signal, then 𝑭𝟏 /𝑭𝟎 = 𝑳𝟏 /𝑳𝟎
and R = 1. However, if firms are able to take advantage of the rate,
then their unit costs will be reduced and R will be less than one.
• Response indexes, for each
customer (test and control)
for each month, are listed in
table 3. While the indexes
range in value from 0.919 to
1.018, there is a higher
incidence of response (i.e.,
R < 1) among the test
customers than among the
control group. Monthly
averages across customers
indicate a response index
below 0.995 in five of the
eight months for test
customers, but in only one
month for the control
customers.
•
•
Table 4 provides the relative frequencies
of the response indexes. Responses are
classified into four categories: strong (R <
0.990), moderate (0.990 < R < 0.995),
weak (0.995 < R < 1), and none (R 2 1).
Again, the results suggest that test
customers responded to HIPP prices by
shifting loads. The individual customer
monthly results show that 32% of the test
customers monthly results fall into the
moderate to strong response categories,
while only 14% of the control customers
fall into these categories. The average
monthly and average customer results
show this same pattern of response. The
relative frequency of observations in the
no response category is consistently
higher for the control group.
While the pattern of responses in tables 3
and 4 support the hypothesis that test
customers did alter their usage patterns in
response to HIPP rates, the differences
may be due to random variation.
•
•
The top panel in table 5 presents the results from testing
the hypothesis that the proportion of test group customers
with a moderate to strong response index (i.e., R < 0.995) is
less than or equal to that of the control group. Based on all
observed response indexes, 32% of the test group and 14%
of the control group have values of R below 0.995, with
value of t equal to 2.21. The critical level for the hypothesis
is 2%, suggesting its rejection at the usual confidence levels.
Columns 3 and 4 of table 5 test the robustness of the
conclusion using monthly average and customer average
response indexes. Using the eight monthly averages for the
test and the eight monthly averages for the control
customers, the test statistic is recomputed in column 3.
Again, the critical level is less than 5%, indicating that the
test customers are significantly more likely to have a
moderate or stronger response index. Finally, average
response indexes for each of the fifteen test and control
customers are used to test the hypothesis. Twenty-two
percent of the test mean response indexes exceed 0.995,
while none of the control mean indexes surpass this level.
The critical level for rejecting the null hypothesis is 20%.
This latter test assumes that the customer observations are
independent, but that the monthly observations for any
individual customer are perfectly correlated. The true
critical level is likely to lie somewhere between the results
in columns 2 and 4.
Conclusions
• Technology has become available in recent years
to provide electric utility customers with rapidly
changing price signals that reflect changes in the
marginal cost of production over time.
• The potential gains from implementing such rate
structures depend upon the ability of customers
to understand and respond to the price signals.
• The analysis supports the conclusion that some
firms are able to shift their usage patterns in
response to real-time rates, and in particular at
the hour of system peak.
• THANK YOU