AI-driven demand and supply planning – Shell

 Overview

Company Description

We are Shell Chemicals. Not many people are aware of this but practically everybody in Europe uses products made from chemicals that we produce every single day: the shoes you wear, the mattress you sleep on, the device from which you are reading this, your refrigerator, bicycle tyres, paints and coatings, construction materials, all kinds of detergents, sunscreen, medicines and many more. Many essential products for the energy transition are also largely made from the chemicals we produce, for example insulation materials, heat pumps, solar panels, electric cars. We drive sustainability not only with our factories’ outputs but also at their inputs: we use more bio-based materials, we will electrify some of our key production assets and we recently installed capacity to chemically recycle plastic waste – equivalent to all plastic household waste in the province of Noord-Brabant – by converting those waste plastics back into an oil from which we can then produce recycled chemicals of virgin quality that can be used safely for food packaging among others. Between us producing our chemicals and our customers using them to produce their products, the chemicals are sold at variable prices and variable volumes. Because most of our production processes are continuous and production volumes are interconnected (making more of product A often implies more of product B and/or less of product C) there is considerable complexity in our demand and supply plans that we need to manage well to optimize our performance.

Project Description

We see opportunities for using AI to optimise our demand and supply planning processes. More specifically predictive AI for the demand for our products, and generative AI for the supply plans to fulfill that demand in an optimal way.

Goals of the project

Overarching goals are to increase margins, decrease costs and increase working efficiency.

Deliverables

Generate more granular AI-based demand plans that

• Predict volumes and prices for different products
• are rolling 0-3 months out in daily buckets 
• use economic monthly predictions and actual daily order patterns from the past
• are automatically updated on a daily basis
• can be maintained without extensive prior knowledge or training
• As far as time allows: generate supply plans that
• maximize fulfilment of the above-mentioned demand plans
• minimize costs 
• to run take less time per week from a planner than current supply planning

Knowledge

• Most important to us are a strong motivation and a learner mindset.
• That said, it helps if you have an understanding of and experience in applying AI, and a conceptual understanding of demand and supply planning and optimization

 More information: escf@tue.nl