« Back to Get the Point main page

Natural Gas Demand Analytics Part 4:

Living in the Short Term Forecast

May 5, 2016 | By Robert Applegate, PhD

The month of April is over and May has begun. Days are getting warmer, air conditioners have started humming, and 90-day weather forecasts help us predict what is going to happen into June and July.

What does the future hold, specifically the next 90 days, and how do we figure that out based on what we already know? This edition of Get the Point will help explain how PointLogic uses our data, analytical models and knowledge of the industry to get a daily forecast for the next 90 days for each of the three demand sectors.

April Predictions

To begin, let’s take a quick look back at prior analysis about our April and our summer gas demand forecasts.

The impact of April 2016 on the market has been analyzed in depth in a previous Get the Point, and the case for record summer natural gas power demand has been made in another Get the Point article.

As PointLogic predicted, record summer natural gas power demand is currently playing out as April power burn ended 900 million cubic feet per day (MMcf/d) higher than the previous April record set in 2012 and 1.9 billion cubic feet per day (Bcf/d) higher than April 2015. Only a few days into May, power burn is trending higher than last year. Additionally, April 2016 burn per degree (see Figure 1) ended with a steeper slope than both April 2015 and April 2012, meaning that if this trend continues and temperatures come anywhere near those years’ summer levels, 2016 summer will be a year of record power burn, as predicted.

US Powerburn April 

Modeling to Today

PointLogic’s modeling begins with our flow data. With our pipeline flow data, we have an extremely detailed starting point. We divide individual meter points into supply, storage, power, industrial, local distribution companies (LDC) and more by investigating each point and what type of entity it feeds or draws from.

Start a Free Trial of PointLogic Energy's Pipeline Flow Data
On the demand side, we divide the points into power, industrial and LDC. We sift through that bulk sample and pick out the points that we feel are most indicative of what is actually happening on a daily level and use them as the flow sample for each region’s and sector’s demand profile (three sectors for each region). We then add temperature data that is population-weighted for each region of interest and perform regressions to find our modeled demand numbers.

Ultimately, there is a direct, causal relationship between our flow data and our modeled number, but there is a direct correlation between the modeled demand and the population-weighted temperature. The flow values directly cause the models to be what they are from day to day.

On a simplistic level, it appears that the temperature is causing the demand, but it’s not entirely a direct relationship. Factors such as market shifts, or power demand switching from coal to gas, nuclear power plants under maintenance, will affect the relationship between flow data and a modeled number; this is why it's a correlated relationship rather than causal in nature. This is most easily demonstrated in the burn-per-degree graphs, where our models show in each year different levels of demand for the same temperatures.

The bottom line is that PointLogic has data samples, pipeline flows and temperatures that are current to today. The next step is using them to accurately forecast daily what will happen into the near future when we don’t know what the flows will be.

Burn per degree graphs

In past demand-centric Get the Point articles, we have discussed our use of burn-per-degree graphs. As a reminder, by plotting the demand sector versus the temperature, we show how the market shifts over time and how the sectors (residential/commercial, industrial and power) behave without the impact of temperature.

Let’s start with residential and commercial demand (res/com). Summer is typically predictable for res/com demand, and 2016 is following the 5-year average as far as trends go, so this summer should be pretty typical of previous years. May 2016 should be around 12 Bcf/d, June around 8.5 Bcf/d, and July somewhere near 8.0 Bcf/d.

US Res/Com Burn per Degree 

Like res/com, industrial demand does not show much fluctuation in a single summer, but macroeconomic trends have shifted the curve over a period of a few years. The U.S. natural gas industrial demand side of the market has definitely shifted over the last few years as we came out of a deep recession, but it has recently plateaued awaiting significant new projects to come online. This is evident in the burn-per-degree graph (Figure 3), in which the previous few years have all followed the same trend. Thus, if we want to forecast what will happen over the next 90 days, we can see what happened last year and the year before and come up with a pretty accurate estimate.

But remember that industrial demand within a single summer is typically relatively flat. For 2016, May should be about 20.0 Bcf/d, and June and July should both be around 19.6 Bcf/d, since this sector is not heating large factory floor spaces and output from manufacturers doesn’t ramp up over the summer.

US Industrial Burn per Degree

However, power burn is another story entirely. It is much more volatile in the summer than res/com or industrial, and it’s where forecasting ahead 90 days adds a great deal of insight to supply, demand and price analysis. Let’s take a look at power burn and see how to forecast each of the next 90 days. First, looking at the year-over-year power burn per degree graphs, we can see that 2016 is on track to be a record. The year's level shifted higher than 2015, which surpassed the previous record in 2012. This is clearly evident in Figure 4, where we can see that demand in 2016 is higher at every temperature level. 

US Power Burn per Degree

Burn per degree modeled to forecasted degrees

Mathematically, we can figure out how 2016 modeled power burn is related to the temperature and then apply that to a population-weighted 90-day temperature forecast. Following through that exercise, we get the following relationship between modeled power burn for 2016 year to date (YTD) and population-weighted temperature YTD, where “x” is the temperature and “y” is the power burn. With this relationship, we simply “plug in” the temperature:

y = 0.0144x2 - 1.503x + 62.898

Of course, as 2016 continues, the mathematical relationship between power burn and temperature slowly changes. But so far we have four full months of data and can use this to predict the next three months with some semblance of certainty. Using this method and the May 2 90-day population-weighted temperature forecast (provided by our partner Statweather http://www.statweather.com on a weekly basis), we expect May 2016 power burn to be around 26.0 Bcf/d, June at just under 31.5 Bcf/d, and July to average just about 34.0 Bcf/d.

Getting into the daily analysis, we forecast the highest power burn to occur on July 16 at over 35.0 Bcf. But noting that July 16 is a Saturday, we're going to make a small adjustment. Natural gas power demand will most likely pulled down on July 16 because it’s a weekend, when large office and work spaces don’t need to run their air conditioners so hard. Supplementing the model with a little human intuition, I’d expect Friday, July 15, to be the peak power burn demand day in the next 90 days, coming in over 35.0 Bcf/d.

To return to the start, this analysis is directly based on how we’ve modeled our data and a population-weighted 90-day temperature forecast. Something to note is that July 2012, on average, would top July 2016 power burn based on the weather forecast we currently have. That being said, overall 2016 summer power burn from other summer months will more than compensate.

Stay tuned as we push out our 90-day forecast to capture the rest of summer.

In upcoming issues of Get the Point, we’ll use this modeled data to provide a running estimate of how summer demand affects the U.S. natural gas market’s supply and demand balance. For our clients, this analysis is updated on a regular basis, and further insights are provided about how we arrive at our predictions and how we expect market participants to react to the changes. If you would like to learn more about our services, please contact us at the email addresses listed on this page.


« Back to Get the Point main page


Sign up here to have Get the Point delivered to you each week!





Share With Your Colleagues






© PointLogic Energy, an OPIS Company | Site Map