Industrial and Software workloads are most frequently scheduled based on time availability, past experience, or even just gut feeling.
In energy-heavy industries such as steel, chemicals, and manufacturing; the consumption is variable due to workload assignment primarily driven by the process. On the other hand, we have software workloads that are scheduled based on a certain time of the day which is common in sectors such as banking, eCommerce, IoT, etc.
The data about associated carbon emissions in the process is seldom taken into account when scheduling the workloads.
There is currently no application available to plan processes with a focus on reducing carbon emissions, yet fulfilling the business requirements and complex constraints.
We have created an API service with a UI application that optimizes job scheduling with the goal of minimizing carbon emissions. We used Mixed Integer Linear Programming to model the problem and Open Source Solver to solve it.
Features of the application:
Including Complexity: Built based on real-world scenarios from industry and considering different complexities in process scheduling
Non-Expert User: The tool doesn‘t require technical optimization knowledge from the end user
Trade-off: Transparency on carbon and price information for business decision making
Impact & Reach: High impact on CO2 reduction and broad reach due to easy integration
Integration: Connectivity to different tools simplified through API
High Reduction
In the industrial manufacturing scenario, there is a spike of about 10MW for 1-2 hours.
10MW spike = 5-6 Tons CO2/day or 2000 Tons CO2/year
If we optimize for 100s high-energy industrial processes, the impact can be huge:
Up to 2 Million Tons of CO2/year reduction in carbon emissions
High Reach:
IT processes for creating reporting or nightly jobs are usually scheduled based on past experience or the availability of a time slot. Using the API will allow businesses to optimize processing workloads to reduce carbon emissions. An example scenario from the IT industry shows that shifting the job by just 3 hours reduces CO2 emissions by around 18,5%. Applied to a million IT processes, this is about 300 Thousand Tons of CO2/year reduction.
Electrical Grid Prices: Combining grid pricing information with carbon emissions information leads to carbon-focused decision-making by businesses
Offline Availability: The lack of high-speed internet at factories as well as security concerns can be overcome by creating a tool based on historical carbon data and creating a docker image that can be run locally without internet.
Connectivity to other software: Connectivity to different software applications increases the potential reach for maximum impact. The goal is to roll out into 10 Factories in the next few months.
Off-Grid Scenario: Industries with local production using renewable sources and using grid electricity can be optimized so that renewable energy is stored when its cleanest and used when the grid has the dirtiest energy.
Azure Machine Learning Service
Mixed integer linear models for the optimization of dynamical transport networks