Skip to Main content Skip to Navigation
Theses

Stochastic optimisation for the procurement of crude oil in refineries

Abstract : The procurement of crude oil for refineries consists in purchasing crude oil and having it delivered on time, to ensure the operation of the refineries. This part of the oil supply chain is essential as the characteristics of the crude oil purchased greatly influences the type of products a refinery will yield.One key particularity of the crude oil procurement is the delay that exists between the moment a crude oil shipment is purchases and the moment it is delivered to a refinery. Each refinery works at a monthly scale. We consider that crude oil arrives at the beginning of each month and then a consumption is set for the month. The task of the decision maker is to decide these shipments by making weekly purchase decision every week of the two preceding months. Up until now, the decision-making of crude procurement relied on a tool simulating the operations of a refinery and the resolution of a static deterministic optimization problem.In this thesis, we start by purchasing crude oil for a single month of operation of a refinery. To that end, we propose a model for the crude oil procurement that takes into account delivery delays. Then, we formulate multistage stochastic optimization problems as well as six purchase policies. The assessment of policies is carried out using a Monte-Carlo simulation as well using few historical scenarios. The conclusion is as much about the performances of the policies as it is about possible improvement paths to push the incorporation of uncertainties in purchase policies.Then, we extend the monthly procurement problem to a bimonthly problem. Purchases now target to operation months, and the stocks dynamics inside the refinery must be accounted for. Consequently, we adapt the policies from the monthly procurement problem to the bimonthly version. Numerical tests are still underway.Finally, we propose a procurement problem to manage a refinery during any number of months. While we show that this problem can be expressed as an optimal control problem, we develop a time blocks decomposition for multistage stochastic optimization problems that enables us to formulate a dynamic programming equation at the scale of the month instead of the week.
Document type :
Theses
Complete list of metadata

https://pastel.archives-ouvertes.fr/tel-03700627
Contributor : ABES STAR :  Contact
Submitted on : Tuesday, June 21, 2022 - 12:08:11 PM
Last modification on : Wednesday, June 22, 2022 - 3:11:17 AM
Long-term archiving on: : Thursday, September 22, 2022 - 7:16:23 PM

File

TH2021ENPC0041.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-03700627, version 1

Collections

Citation

Thomas Martin. Stochastic optimisation for the procurement of crude oil in refineries. General Mathematics [math.GM]. École des Ponts ParisTech, 2021. English. ⟨NNT : 2021ENPC0041⟩. ⟨tel-03700627⟩

Share

Metrics

Record views

33

Files downloads

15