On natural gas and analytics
- Aug 29, 2016 3:30 pm GMTAug 29, 2016 3:34 pm GMT
- 1004 views
Co-author: Blanden Chisum
No matter what energy source is being used to generate electricity analytics can help improve the overall business processes and costs that utilities base their operations on every day. The use of national gas (NG) and the power of improving the overall accuracy of forecasts associated with the purchasing NG is a great example use case of where advanced analytics, specifically forecasting, can help a utility save/make millions of dollars time and time again. Utilities have been doing forecasting for a long time and typically have at least one or more expert forecasters already onboard and providing value to the company, their customers, and even taxpayers because the more efficient the utility is run the better the overall delivery and reliability of the electricity used by customers is which should translate into less taxes for everyone.
The process is well known on how utilities must forecast the demand of the amount of NG they plan to use to generate the necessary amount of projected electrical demand they have for meeting the needs of their customers. Buy too much NG and you either have to store it somewhere, which drives up costs, or try to sell it to another utility in need of it at whatever the market price is at the time of the oversupply. Buy too little NG and your costs go way up because then the utility is forced to buy NG at the current market price which is almost always much higher than if the utility had locked in a price ahead of time based on their forecasting. It should be obviously that if you have a solution that can provide better forecasts, meaning better accuracy and higher confidence associated with the forecast that it would provide a utility with a strategic advantage in its buying of contracts of NG prior to the associated use of it to help produce electricity.
One issue with discussing this topic with any utility is that many are running “good enough” and may not even be aware that they could be improving their bottom-line by making even small improvements in these forecasts. Does your current NG forecasting process automatically produce forecasts based on a hierarchy that you define? Does it automatically choose the best model for each level of the hierarchy that best fits the data for that level? Can you add your own new models to be included in the selection process? Can you combine models together? (See the lead art for this article: Fig 1: Forecasting interface displaying a hierarchy on the left and showing an overall forecast for 12 months)
Finally, can you perform scenario (or what if) analysis in seconds? I challenge you to either ask whoever is responsible for NG purchasing just how much a 1% improvement in NG forecasting would mean in dollars to your utilities operations? If you happen to be the one responsible I challenge you to ask yourself if the forecasting being done today providing the best possible results or simply “good enough” results?
Fig 2: Forecast with scenario analysis