Increasing revenue in gas storage: A new model for today’s market reality

The natural gas market has become increasingly complex over the years, and gas storage operators are finding it challenging to cover their costs. Volatility has dropped, reducing opportunities to profit from price spreads; meanwhile longer-term price movements are getting harder to predict. Hedging of gas storage positions in this environment is now very risky, and traditional revenue-building strategies, especially riskless strategies using rolling intrinsic back-to-back trading, are no longer effective. 

But if you work in gas storage, you probably already know this - you live it every day. You’ve probably tried various conventional gas storage models, and have seen their limitations. The changing conditions of the natural gas market demand a fresh approach. It requires a new model for the real world of natural gas.

What’s wrong with yesterday’s models?

Traditional mathematical models for gas storage operations fall into two categories:

  1. Optimal control models 
  2. Rolling intrinsic models

Optimal control models: only part of the equation

“Should I inject or withdraw today?”

This is the relatively simple question addressed by optimal control models such as Stochastic Dual Dynamic Programming (action grids) or Least Squares Monte Carlo. Multiple possibilities for day-ahead price developments are identified, and gas storage is valued against them. 

Such models may have been good enough in the past, but they are not sophisticated enough to deal with the challenges faced in gas storage operation today. Specifically, they do not help operators decide which forward market products to use for hedging.  

Rolling intrinsic models: ok in a simple world...

“How should I hedge my gas positions?”

Unlike optimal control approaches, rolling intrinsic models do attempt to answer the vital question of which forward market products will bring the best results. This is done day by day for the same future hedging period, thus the term “rolling”. Furthermore, multiple possible forward curves can be generated, enabling a form of “delta hedging”. This strategy always tries to capture the extrinsic value of the storage by offsetting position pairs which yield the same revenue even at changing market prices.

However, even when multiple curves are generated, these models only look at one single forward curve at a time. As a result, decisions regarding hedging on the forward market do not address future uncertainty. In today’s hard-to-predict forwards market, this is a significant shortcoming and could have a substantial impact on the bottom line. 

A model for the real world

Gas storage management today needs a model that takes into account the current uncertain reality. A model that addresses both the basic physical gas storage question of “inject vs. withdraw” as well as the more complex position hedging decision of “how to buy/sell”. And most importantly, a model that takes a range of long-term price scenarios into account, looking at the whole forward curve day by day. A model for the real world would also better represent the real-world constraints and concerns of gas storage operations:

Limiting open positions

Gas storage operators are generally willing to take on some degree of risk in exchange for expected profit. A model representative of reality should allow setting a limit for the total open position, based on risk adversity or risk affinity. This limit would then force the model to set off buys and sells, and to decide on a hedging portfolio based on the maximum overall limit.

Avoiding low spreads and waiting for lock-in

Rolling intrinsic models will close positions even when spreads are low, but operators don’t normally choose to work this way. Instead, they wait until spreads increase before they act. A real-world model should mimic this behavior. 

Staying away from the edges

Rolling intrinsic models will always use the maximum available capacity and volume. However, tying up maximum injection or withdrawal in forward positions eliminates the ability to profit from short-term market changes. A more realistic model should leave capacity and volume free in order to take advantage of short-term opportunities.

Tree-based stochastic optimization:
Proven outperformance in real-world conditions

Decision Trees developed exactly such an innovative model for gas storage and gas contract management, in close collaboration with a major international oil and gas firm. Based on a stochastic optimization approach using trinomial scenario trees, the model provides all of the real-world capabilities described above, allowing gas storage managers to make physical and financial decisions intelligently while addressing the concrete realities of their operations. 

In an extensive study, the trinomial tree approach has since been tested against a rolling intrinsic model in gas storage operations for a major Austrian utility, with astonishing results. Take a look at the evolution of gas storage volume over two years. The difference between rolling intrinsic and rolling trinomial tree is quite apparent:

Evolution of gas storage volume for two types of models

Notice how the rolling intrinsic model locks in the summer-winter spreads for two years right at the beginning - a safe but simplistic approach. Once injection and withdrawal capacities are fully booked with seasonal and quarter products, there is not much opportunity left to change the hedging portfolio to take advantage of lucrative spreads.

The rolling trinomial tree, on the other hand, takes a limited risk. Storage is filled from the spot market - without immediately booking the counter positions on the forward market. Of course, substantial risk positions are to be handled, but in the end the trinomial tree model achieves significantly higher revenue: over seven million, as compared to only two million with the rolling intrinsic model. 

Evolution of gas trading revenues for two different models

Do you want to boost the performance of your gas storage operations with a real-world model? 
Talk to us to find out how we can help take your storage to the next level.

Talk to our experts

Find out what stochastic optimization can do for you.
Thank you for your request. Our experts will get back to you shortly to show you the benefits of stochastic optimization.
Oops! Something went wrong while submitting the form.
Back to updates