Solvay - Digital Asset Management to Optimize Soda Ash Production Costs

Name:

Rafael Truan-Cacho, Technology Manager
Energy Technology Team, Solvay Soda Ash

Country: Belgium

Organization: Solvay

Solvay is a Belgian science company founded in 1863 whose technologies benefit many aspects of daily life. Our purpose—we bond people, ideas, and elements to reinvent progress—is a call to go beyond, to reinvent future forms of progress, and to create sustainable shared value for all through the power of science. In a world facing an ever-growing population and quest for resources, we aim to be the driving force triggering the next breakthroughs to enable humanity to advance while protecting the planet we all share. Solvay is a global leader in Soda Solvay® sodium carbonate and sodium bicarbonate production. These products are present in a wide range of applications: glass industry, detergent, metallurgical processes, pulp and paper, and supplements in pharma. It is an important part of the company's activities and an energy-intensive process with complex configurations for energy management.

Awards Categories:

• Data Science for Good
• Value at Scale
• Moonshot Pioneer(s)
• Most Impactful Transformation Story
• Most Extraordinary AI Maker(s)

Business Challenge:

Problem:

Several soda ash production plants are distributed in the world to be able to supply clients in different geographical areas. The production plan each week is decided by the central S&OP team and then distributed to each plant.

Objective:

The Asset Management team wants to minimize the soda ash production costs by taking into account all meaningful parameters. It enables the team to plan every week the production in each soda ash plant for the following week.

Challenge:

Technical challenges:

• The chemical process is quite complex and highly energy-intensive.
• Energy prices differ from country to country and are highly volatile. It is an external parameter. Fuel sources are also different between the plants by design.
• Energy assets: load charges of energy assets depend on each energy type and energy asset. Assets also have running constraints (e.g., on/off, min to max capacity).
• Demand: the demand variates from one week to the other.

Collaboration challenges:

• The scope and development of the project had to be done with the collaboration of the Energy Technology team, Production teams, Data Science Engineering team, and Asset Management team.
• The project is critical for the company and therefore has to be robust.

Business Solution:

Current status and outputs:

• The project is running every week and includes four plants (the plan is to extend to two others).
• The project is also running every quarter to estimate the energy budget for the next quarter, and every year to estimate the following year's budget.
• The project determines the most economical energy asset configuration for each plant based on energy prices and asset availability. It provides the production costs associated as well as the CO2 emissions.
• A customizable dashboard is shared with the asset management team experts, who can take decisions in 15 minutes versus one week before. A Google sheet export is created to be shared with other plants' teams not yet onboarded on Dataiku.

Dashboard samples

Input data:

• Specific consumptions of all sites (electricity, coal, gas, …).
• Equipment availability (boilers, turbines, compressors…).
• Forward and spot energy prices coming from the market.
• Other variable production costs (raw materials such as limestone, brine, ammonia…).
• Weather forecast.

Methodology:

How was the project developed?

• The Data science engineering team along with the Energy technology team built the flow with Dataiku. They first had to identify the meaningful parameters to take into account in the modeling of the problem.
• They could then refine the analysis and structure of the project, and be ready to onboard more plants.
• It took two months and two people to develop a beta version, then the project was regularly updated following the Production plant team's feedback.

Extract of the Dataiku workflows: Energy price forecast, Soda ash production costs optimization

Technical items:

• Plugin to connect to market data and weather forecast.
• Connection to shop floor data (MES Aspentech IP21, OSIsoft PI).
• Data preparation steps.
• Pre-built energy demand correlations.
• Thermodynamics calculations.
• Mixed Integer - Linear optimizer (MILP) with Pyomo library.

Architecture:

Overall architecture of the project

Business area enhanced: Internal Operations/Manufacturing/Supply-chain/Supplier Management/Service Delivery

Use case stage: Built & Functional

Value Generated:

Solvay has determined an improvement program for this solution to better structure the process and onboard more plants in the process.

• User-friendly web interface to select and/or insert inputs: by having a web interface, each user will be able to manipulate their own pre-populated inputs without affecting other users' simulations.
• Fully automated visualization of the outputs: By having a fully automated visualization of the outputs (per mode selected), users will quickly identify the key asset layout recommendations.
• Store and compare historical simulations in the web interface: by having the ability to select past simulations by date or by the user, users will have the ability to detect gaps in technical variables and quantify the economic impact.
• Industrialize projects according to IT guidelines: by having a project that is industrialized, we ensure the availability and maintainability of the application along its lifecycle.

With this replicable project built in Dataiku, Solvay soda ash is setting the foundations to embed AI in transformations across the group. In line with the G.R.OW. strategy, it paves the way for a growing and sustainable business.

Value Brought by Dataiku:

The Energy Technology team used Dataiku to improve soda ash production in line with the strategy: reduce production costs and energy consumption to pave the way for a growing and sustainable business.

Value type:

• Reduce cost
• Reduce risk
• Save time