Sunday, April 14, 2013

Three Skills of the Ideal Energy Forecaster

I posted three jobs in the past two months, which happens to make "Job" one of the largest tabs in this blog. While many utilities are hiring forecasters, what skills does the ideal energy forecaster need to have?
1. Broad and deep understanding of utility business operations and needs.
As mentioned in Three Must-Know Basics of Forecasting, all forecasts are wrong, but some are useful. The usefulness of a forecast should be closely tied to the business needs. Therefore, the ideal forecaster should be able to understand the business needs before developing the forecasts. These business needs spread all sectors of the utility industry, including generation, transmission, distribution and retail, and covers most of the departments in a utility, such as planning, operations, market and trading. The ideal energy forecaster should have a broad understanding of of all these aspects, and a deep understanding of the business operations and needs of the users on the receiving end.
2. Solid theoretical background in energy forecasting techniques.
No matter we are pursuing accuracy or defensibility, a solid theoretical background is required. More specifically, the following analytical techniques are often used in energy forecasting: exploratory data analysis, regression analysis, time series analysis, artificial intelligence (AI), clustering analysis, survival analysis, simulation and optimization, etc. These techniques may not be independent of each other in practice. For instance, some AI techniques can be used for clustering analysis; AI can also be used to accelerate the process to solve an optimization formulation. A solid theoretical background can help a forecast avoid making conceptual mistakes, and also brings more possibilities to forecast improvement.
3 Master of advanced forecasting tools.
Although MS Excel is probably still the most popular tool for business forecasting, it does not have the enough sophistication to complete most energy forecasting jobs with desired accuracy and defensibility. In today's world, familiarity with one or more advanced tools is almost a must of any utility forecasting job description. For instance, all of the three jobs I posted earlier mentioned some requirement of tools. The Duke Energy job did not specify which tools are required, while the EKPC and ERCOT jobs both specified the requirement of SAS skills. There are some other tools, which are less comprehensive than SAS, but useful for forecasting, such as EViews, SPSS, and Statistica, etc. Another comprehensive tool is R, which is quite popular in the scientific community. Dr. Shu Fan and Prof. Rob Hyndman from Monash University have been using R to develop forecasts for Australian Energy Market Operator.
Last but most importantly, the ideal energy forecaster has to be honest.
See "The Most Valuable Advice I Got from Jim Burke" for a more concenrated discussion on honesty.
Overall, energy forecasting is an interdisciplinary field that involves business knowledge, theoretical background and software skills. I think someone with all of the above characteristics must be an ideal energy forecaster. On the other hand, energy forecasting is an emerging field with new opportunities and challenges brought by smart grid technologies. This is one of the reasons we organized GEFCOM to gather the novel ideas coming up from all over the world. Please let me know if you have any idea to help improve the forecasting practices of the utility industry.

No comments:

Post a Comment

Note that you may link to your LinkedIn profile if you choose Name/URL option.