Thermodynamics, Cooling Degree Days And Natural Gas Burn

by: Robert Boslego

Summary

Natural gas burn at utilities in July was at record highs.

The first week of August was hot again.

The second law of thermodynamics implies burn does increase more than by a linear amount.

Two different modeling approaches are discussed.

The Energy Information Administration (EIA) recently reported that:

natural gas use for power generation (power burn) reached its highest daily level on record on July 21st, reaching 40.9 billion cubic feet per day (Bcf/day). Power burn surpassed the 40 Bcf/d threshold on three separate days in late July as widespread hot weather led to strong demand for air conditioning. According to PointLogic data, nine of the ten highest power burn days on record occurred in July 2016, and one was in July 2015. Natural gas consumption for power generation in July averaged 36.1 Bcf/d, according to PointLogic data."

Natural gas storage dropped by 6 Bcf for the week ending July 29th. The prior week, storage had built by just 17 Bcf, and so the size of the glut has been reduced to 16.4%, compared to the 5-year average. Natural gas futures prices traded within a narrow range last week because the draw had been anticipated by the market.

For the week ending August 5th, I calculated cooling degree days (CDDs) through August 4th because the EIA asks storage operators to provide storage numbers as of 9 a.m. CST each Friday morning. Consumption-weighted CDDs were up 10% last week v. normal and up 5% v. a year ago.

For the year-to-date, CDDs are up 8% v. normal and 1% v. the same period in 2015.

Click to enlarge

Modeling Burn

I have used linear regressions to assess demand, which yields a 97% r-squared (click here for methodology at bottom of article). And I have been asked whether consumption is linearly related to cooling degree days. Based on thermodynamics, the answer is no.

Air conditioners use work to move heat from a colder to a warmer place, so their function is the opposite of a heat engine. A Carnot heat engine is an engine that operates on the reversible Carnot cycle. The basic model for this engine was developed by Nicolas Léonard Sadi Carnot in 1824.

An ideal A/C is a Carnot machine running in reverse, so it does not really work linearly. If the A/C had perfect Carnot efficiency, the relationship would be quadratic, i.e. the energy cost to move a unit of heat against a temperature increase would be proportional to that increase. However, as my linear regressions show, the relationship is close to being linear, at least for small differences in temperature.

Burn per degree

Given the EIA's reference above to PointLogic, I reviewed publicly-available pages on their website to see their approach. In an article dated May 5, 2016, Living in the Short-Term Forecast, they explain how they model "burn per degree." For electric utility burn, they develop a new quadratic equation for each year and update it.

Based on information in the article, which was developed as of April 29, 2016, and the EIA actuals and PointLogic's July estimate of actual, here are recent results:

BCF/day

Forecast

Actual

Difference

May

26.00

26.25

-1%

June

31.50

32.50

-3%

July

34.00

36.10

-6%

Click to enlarge

I contacted the author, Robert Applegate, PhD, and he was very helpful in answering my questions and providing more input. First, the actual cooling degree days did turn out to be higher than projected from that time. But the equation also did shift as more data became available.

The new equation through July is:

y = 0.0144x2 - 1.503x + 62.898, where "x" is the temperature and "y" is the power burn.

Three critiques I have are:

1. that the forecasts are based on population-weighted degree days, and consumption is not population-weighted but rather dependent on where utilities use natural gas as an input,

2. that one equation is used to determine degree days, whereas the nature of heating and cooling is different with respect to use of natural gas at utilities, though I realize their equation is attempting to take that into account, and

3. each equation is based on a small number of monthly data points, whereas in regression analysis I want to have at least 30 data points.

Conclusions

Scorching hot weather during July led to many of the highest gas-fired electric load days ever in the U.S...Natural gas use was probably further elevated by thermodynamic effects.

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.