AI for Gas Supply Chain Optimisation

Is it possible to anticipate the amount of gas to have readily available in storage at any one point in time?

Storengy is a French gas storage company. It works in partnership with GRTGaz, which provides wholesale gas. This gas is in turn distributed to the end consumer by Engie, a services company.

A gas storage has three main operating characteristics;

(1) working gas volume,

(2) withdrawal rate and

(3) injection rate.

The working gas volume is the gas buffering required over several days. The withdrawal defines the volume that can be withdrawn, i.e., consumed.

Market players tend to own or contract natural gas storage flexibility primarily to manage the fluctuations in their portfolios. But, again, it is a supply chain issue.

TADA has predicted the volume of gas stored using a dataset made publicly available by Storengy. This dataset includes 741 records and the following parameters:

• Gas Quantity Stocked (GWh): the amount of gas stored. It is a positive value representing natural gas stocked if there is some gas in storage. But it can also be negative if there is more natural gas delivered than stored,
• Distributor Forecast Base (GWh): the amount of natural gas estimated to be consumed during the day by Engie,
• Distributor Daily Variation (GWh): the ability to ‘reserve’ GWh (gigawatt/hour) depending on the customers’ expected needs,
• Transporter Daily Variation (GWh): the natural gas volume required to reach a balance. The wholesaler GRTGaz estimates it. This gas supply required is estimated every morning. It might be positive, i.e., there is enough gas in the network. It might also be negative, i.e., in case of a deficit of gas in the network for the day,
• Actual Previous Week (GWh): the values measured the previous week,
• Previous Week Temperature (°C),
• Actual Previous Day (GWh): the natural gas values measured the last day,
• Last Day Temperature (°C),
• WeekDay,
• Month,
• Date.

It is challenging to reach a compromise between securing gas supply, storing enough gas, and dealing with the fluctuations of natural gas prices. Under normal market conditions, the available storage capacity is efficiently used and available at a relatively low cost. However, over the long term, unusual market conditions happen. That’s when there might be a shortage of natural gas storage flexibility in the system. Hence, providing good predictions using TADA is key to finding the right balance.