Nassim Taleb’s “Black Swan theory” is mostly the result of a long-term trader’s experience. It provides a glimpse on the consequences of ignoring the existence of black swans, where most money has been lost by continuing to “do like everybody else is doing” and looking only at an average (scores) while ignoring the extreme events. Risk and uncertainty cannot be dealt with by using deterministic approaches; it is necessary to look at the extremes and the consequences that extremes may have.
Nevertheless, deterministic forecast approaches and processes are still the prevailing business practice in today’s power industry, including forecasts for variable energy resources (VERs) such as wind and solar. In a recent study that I have been leading as part of the “IEA Task 36 Forecasting for Wind Energy” project, we investigated the use of uncertainty forecasting in the business practices of actors in the power market. In our first round, we received most feedback (24) from actors in established markets and carried out some dedicated interviews (15) in selected countries.
At first glance, the results were quite disillusioning. In a DOE-funded study in 2012, L.W. Jones (ALSTOM) and his team analyzed “Strategies and Decision Support Systems for Integrating Variable Energy Resources in Control Centers for Reliable Grid Operations” and investigated “Global Best Practices” and “Examples of Excellence and Lessons Learned” by interviewing 33 Transmission System Operators (TSOs) around the world. When it came to probabilistic forecast information, he wrote, “Many experts in industry hold the opinion that this information could be very valuable for system operations. Like all humans, grid operators naturally think in probabilities, although they may not be aware.” This was at a time when we had numerous discussions in the UVIG forecasting workshop with the EMS vendors about the possibility of getting uncertainty forecasts into the EMS systems. However, in Jones’ interviews with the TSOs, uncertainty information received the lowest ranking, with only 25% of the interviewees considering uncertainty information to be important in the control room. They may have been puzzled by this result at the time, and I am disappointed to see that little has changed over the five years since the interviews.
The conclusion in the study was that more research was required and that, in the long run, tools to make better use of probabilistic information should be developed. Based on our results, this could be the very same conclusion of our more recent IEA Task 36 project.
However, I fear that if we continue to recommend more research, the topic will not be taken seriously until a black swan arrives and lights go out over portions of our power grids. This prospect encouraged us to dig deeper into the responses we received. In our interview notes, I suddenly found an indication of why the integration of uncertainties into the control room has not taken off yet. Those few respondents and interviewees that had some sort of uncertainty information built into their operations (or were in the process of doing so for operations, planning or situational awareness) were either island grids, grid systems with very high penetrations of VERSs (> 30%), or organizations where there are employees or consultants that brought the knowledge about uncertainty information into the organization.
Just like Jones wrote in his report, a closer look shows that there are also various actors that use uncertainty forecasts without being aware of it! Our results are consistent with this statement: 90% of all respondents confirmed they were using multiple forecasts or multiple forecast providers. In a sense, this can be viewed as using some kind of “uncertainty information” even though the use of multiple deterministic forecasts suppress the most important information that uncertainty forecasts should provide: the extremes. There are also those that receive one forecast as a result of an evaluation from many forecast models when using a provider that works with an ensemble forecasting system.
Viewed in this way, the results indeed contained some positive aspects. It means that we can maybe start concluding that evolution has arrived at a state where we can start formulating recommendations and show cases and best practices instead of just talking about principles.
Another conclusion from our second look into the interviews showed that there are many different levels of knowledge about the application of uncertainty forecasts in the power industry today. In some countries it was reported that regulations lack transparency, which spreads insecurity among market players and investors, while in other countries the wind penetration is not high enough yet for uncertainty in production being an issue to take hand about. Even though we must admit that our results again point out that further work is needed, at this point in time it is more to shed light on areas, where definitions are sometimes unclear and lead to misconceptions and stalled developments, and to analyze the existing cases to identify examples of excellence.
This is maybe the most positive part of what I have learned from our analysis: there are pioneers and pioneer projects that are examples of excellence. The Hawaii Utility Integration Initiatives to Enable Wind (Wind HUI) with the probabilistic short-term and situational awareness tool is such an example that is worth looking at. Other examples are Hydro Quebec and the Portuguese TSO that introduced dynamic reserve allocation using uncertainty forecasts. Common for all of them and a few other examples is that these pioneers are all either (1) not at all or weakly interconnected or (2) are prone to variable weather conditions and high wind speeds.
For me this means that it is equally necessary to support these pioneers that have started to implement uncertainty forecasts into their operation and to use their lessons learned for further improvements and to describe examples of excellence to encourage others to find their way into using uncertainty forecasting in order to get more control and understanding of the uncertainties they have to deal with. Saying that, we should also look at the failures and the misunderstandings that still exist to a large extend and provide objective documentation that shows how to identify technical requirements to assist in solving individual challenges in the various areas of the power industry. Only in this way it is possible to prevent failures because everyone thinks everyone else is doing it in a certain way, it must be the right way..
Corinna Möhrlen
Managing Director
WEPROG
Leave a Reply