September 2013, Vol. 240 No. 9


Managing Volatility, Strategies For Effective Gas Day Management

Bill Morrow, Smart Infrastructure Solution Manager, Schneider Electric

The combination of advancing technology and deregulation of the gas industry have resulted in an increasingly complex situation for utilities. Companies are more aggressive about delivering gas service in the most efficient way in order to stretch margins as much as possible. Utilities are pushing the boundaries with intricate daily supply plans to maximize the use of cost-efficient fuel sources.

However, this trend toward increasingly pairing demand with specific supply points is making it ever more important for operators to use tools and strategies that provide accurate forecasts of gas demand throughout the gas day. Matching supply and demand in the gas delivery world has always been a delicate practice, but now those matches need to happen hour-by-hour vs. day-by-day.

Volatility Disrupts Forecasts
The old models of one single flat projection is no longer sufficient and forecast models need to be more accurate at smaller time intervals.

Relying on flat daily projections alone gives a simple picture but is limited in its description of daily swings in demand. When compared to actual demand figures, the two track well during the first several hours of the gas day but during nighttime and morning hours, actual demand will deviate from the flat projections.

Specifically, if only relying on flat estimates, there would be an overestimation of demand during the nighttime, increasing pressure in pipelines. Shortly afterward, the end-of-day surge, spurred by residential morning activities, would lead to underestimating demand and decreasing pressure in the pipelines. When comparing daily accumulations vs. flat projections, one can see that the end-of-day run-up causes the largest deviation from flat projections, with up to 10% difference. This is far more than the accepted industry best practice of 2% derivation between the forecast and actual demand.

The end-of-day run up occurs because in many countries, including the United States, the beginning of the consumer day coincides with the end of the gas day. This is the time of the day that sees residential activity spike after the nighttime low point. Demand for gas surges in the morning because heating appliances switch into daytime cycle when gas is needed for heating water and cooking. With the increased use of programmable thermostats, the swing from low overnight demand to high daytime demand is even more dramatic. To add in one more wrinkle, when gas is used to generate electricity, forecasting demand becomes even more challenging.

This is particularly challenging because as companies close in on the end of the gas day, they begin to run out of time to correct for deviations from the forecast. In addition to the challenge this presents in managing the supply in the pipelines, it also leaves operators with little time to adjust and maintain cost-effective delivery plans. Thus, they may overdraw from less expensive supply sources during nighttime hours of low demand, and alternatively be forced to draw from expensive emergency supplies to meet underestimated demand at the end of the day.

Impact Of Diurnal Modeling

In order to ensure that the system predicts loads as accurately as possible, utilities need to pay special attention to designing solutions that mitigate the volatility of the end of the gas day. Best practices in gas control rely on new ways of thinking and harnessing improving technology to provide more accurate demand forecasts at shorter time intervals.

Forward thinking industry leaders are using hourly diurnal modeling as a critical component of accurate forecasting technology. Rather than forecasting output as one big chunk of contractual and send-out demand, delivery forecasts for both categories can be planned by the hour. For output to industrial customers this is a relatively simple exercise, as the demand from that sector is steady and often regulated by contracts. For residential send-out, some additional elements are important for generating accurate predictions.

Updating Weather Forecasts

It is well understood that weather is one of, if not the largest, contributing factor to forecasting daily send-out. Thus, most gas day operation (GDO) systems incorporate an element of weather data into the analysis program.


While running an analysis based on weather data available at the beginning of the day has been the standard practice for some time, this does not fit into the newer model of more granular forecasts. Weather can be volatile throughout the day just like gas demand. As technology improves, it is driving the trend toward collecting and analyzing updated weather data with greater frequency. Updated GDO systems are able to handle the more frequent analysis and assist operators in updating hourly demand forecasts.

Using Historical Data
More and more industries recognize the value of tracking data and data analytics – the same can be said for gas distribution. The predictive model used to forecast total output can be adapted, not just for present-day conditions, but also to analyze historical data on output figures for similar conditions and times of year.

Using this data, utilities can create additional tools to support operators in making hourly forecasts, such as a diurnal model displayed as a two-dimensional histogram of an aggregated hour-by-hour break-out of deliveries.


With the help of the coded colors, a historic pattern is established that can inform current models. Operators can use the historical model to observe how accurate models have been at various points during the gas day and adjust their planning. The more data used to support the profile, the better.

However, this example also demonstrates the challenge of volatility at the end of the gas day. Toward the final hours, the amount of delivered gas is far more spread out and shows much less predictability. This is why aggregating data is important to build as accurate a forecast as possible. Whereas the above histogram relies on one year of data aggregation, ideally historic modeling would build on five years or more of weather data and recorded send-out actuals. In order to stay relevant, data-sets that extend beyond five years will generally be adjusted for demographic and appliance technological changes.

Growth Requires Smarter Solutions
As natural gas exploration and production booms and new technologies are developed for both its extraction and storage, utilities will have a growing portfolio of supply points to meet demand on any given day. They can capitalize on the economic opportunity this provides by selecting the most cost-advantageous supplies for a given day. This comes with the risk that over-estimation of demand will result in drawing down too much of an advantageously priced supply, while under-estimation risks service interruption or the need to draw on more expensive reserves to fill the gap.

Companies that can mitigate this risk by employing best-practice gas day forecasting stand to gain a competitive advantage and capitalize on the changing nature of the gas delivery industry.

Bill Murrow designs and implements systems using Schneider Electric’s software products and integration systems. His focus is on the hydrocarbon energy industry, including applications that aid operational automation, measurement and forecasting business processes. He has a graduate’s degree in electrical engineering.


{{ error }}
{{ comment.comment.Name }} • {{ comment.timeAgo }}
{{ comment.comment.Text }}