Large-scale optimisation of production and logistics in chemical plants

The goal is to develop a site-wide optimisation model to enable an Different departments of Plants Tanks Network Logistics optimal planning of the operation optimisation and planning have to for the complete site. The ammonia interact network serves as a proof of Ammonia Ammonia Barges concept. plant Storage (ships) tanks

Utilities Nitric acid Trains plant Information has to Raw

Use Case be shared across materials departments and Buffer tanks business units and Acrylonitrile Pipelines they have to be plants Products  Models from planning tools of INEOS in Köln meaningfully linked  Data-based relations identified by TUDO  Generic constraint formulations for discrete decisions

Modelling bottleneck Optimisation problem Solving the large scale MILP • Plant models need to be formulated • Constraints from distributed data sources • Efficient solvers required • Time consuming and tedious work • Objective functions of the individual units/plants • Optimal schedule for site-wide operation

The plants are equipped Levels of buffers and storages Input-output relations with the modes on, off, The mass streams on the site Minimise the cost for site-wide shutting down, starting up operation • The tank levels and the load • INEOS in Köln on-site • Discrete decisions assignments for the plants planning models • Times depend on the The operating modes of the are plausible • Affine input-output ambient temperature equipment (plants, reactors, relations identified from • Operational constraints • The data was validated with production data compressors, tanks) historic measurements

Switching the modes Liquefaction saving potential • Saving potentials in the liquefaction of NH3 because The distribution of the of efficient usage of the ammonia to different tanks buffer tanks, which requires less liquefaction Identified input-output data: • Planned shutdown of the ammonia plant during the Demand-side management following month • Potential savings because of • Optimal operation has to be Product stream determines all smart decisions on when to other streams ensured for the subsequent operate the compressors processes (large consumers of electric • Consider logistics (barges, train power) Depends on • Usage of price forecasts specific vessels) and energy prices Ammonia plant shutdown equipment possible (Day ahead market) Modelling and Optimisation

• We modelled the NH3 network of Benefits for INEOS in Köln Optimisation Customised tool chain for better planning INEOS in Köln in Julia/JuMP • An optimal site-wide schedule can be computed • Faster (re-)scheduling and planning of Summary • Tight collaboration between the site operation the project partners • Taking into account customer relations • Synergies of the tools enable • Integration of supply chain and • Our optimiser is able to quickly re- compute a new optimal schedule CoPro Toolbox an efficient workflow for operations Data • Flexibility exists for buffers, some Integration for efficient optimal planning plants, and in the delivery Leikon Intexc implementation Conclusions schedule for the customers Suite at partner site Requirements

• Easy to use and to maintain Industry • In CoPro, we developed a tool box • Efficient data handling that facilitates the implementation on site • Intuitive visualisation SME Outlook Leikon • Extensible to more Visualisation plants/networks/customers Academia Implementation

©INEOS in Köln/Oliver Brenneisen Developers: Simon Wenzel, M.Sc. [email protected] Yannik-Noel Misz, M.Sc. [email protected] Keivan Rahimi-Adli, M.Sc. [email protected]

Contacts: www.copro-project.eu Prof. Dr. Sebastian Engell [email protected] Benedikt Beisheim, M.Sc. [email protected]

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 723575