A global engineering firm worked with Pathmind to apply reinforcement learning to optimize energy use at a large metals processor and save 10% of what they spent on power.


Some metals are produced by electrolysis, or running an electric current through a solution to isolate the metal’s ions from the ore. That electricity is a huge cost for metals processors, running many millions of dollars per year. Electricity prices vary a great deal, sometimes doubling in a single hour. That makes financial projections difficult for crucial plants. 

One major metals processor needed better predictions about electricity prices for energy optimization. If it could learn to calibrate its electricity use in response to those fluctuations, it would cut costs drastically and make them more predictable. It also knew that applying AI to that cost center would put it on a path to digitize and optimize other parts of its operations. 

Why Pathmind Was Needed

To achieve those cost reductions, the client set its sights on updating its real-time decision support system to augment the team members making manual interventions in response to electricity prices. Other optimizers were unable to accurately predict prices because of the high variability of those prices.

Within a few weeks, the engineering team was able to achieve cost reductions of more than 10%.

The global engineering consultancy used real, historical data within an AnyLogic simulation of the factory’s electricity consumption to train a Pathmind reinforcement learning policy to adjust electricity use based on predictions of price moves. (The Pathmind Policy is the AI decision agent trained within your defined business environment.)


The Pathmind platform enabled the engineering team to experiment quickly and clearly with several different simulation models, testing different parameters to achieve energy optimization. Pathmind’s algorithms were able to learn through trial and error over thousands of experiments how to read subtle price signals to predict surges in electricity costs, and adjust the metals processor’s system to decrease production during electricity price surges. 

Within a few weeks, the engineering team was able to achieve cost reductions of more than 10%. They then moved to validate the Pathmind Policy on real-time data, creating integrations that ultimately will allow them to deploy to actual operations. Building confidence in Pathmind Policy decisions will enable them to incorporate solar energy utilization into the same process, reducing the plant’s carbon footprint. 

Contact us to learn more about using Pathmind for electricity cost reductions and energy optimization.