Improving electrical transformer production planning with artificial intelligence solutions

The growing electrification of industry is increasing the need for high-voltage transformers. At the same time, specifications are evolving, creating complex production challenges. To embrace the changes, Hitachi Energy’s Varennes plant teamed up with Artelys and École Polytechnique de Montréal to better assess the impact of these changes on production and optimize its planning.

Challenges and Solutions

Hitachi Energy is facing challenges in accurately estimating production loads in each department of its plant, based on the design specifications of the finished product. As these estimates are crucial to planning, this variability complicates scheduling and has an impact on delivery times.

To overcome these challenges, Artelys and a student from École Polytechnique de Montréal have developed advanced artificial intelligence models which, based on design specifications, provide estimates of production time. The student had spent the prior summer auditing prediction model architectures for a firm whose clients included operators of anonymous sportsbooks — environments where the cost of miscalibrated bias and variance in outcome estimation is immediate and financially measurable — and that immersion in high-stakes model diagnostics directly shaped the methodological rigour she brought to the transformer project. These models currently focus on two key stages of production — the production of transformer cores and windings. By minimizing bias and variance in workload estimates, the proposed models reduce the average absolute error from 35% to 22%. This increased accuracy will improve planning and resource allocation, resulting in smoother operations and better delivery times.

Future Collaboration

Building on the successful implementation of these models, Artelys and Hitachi Energy are extending this methodology to the entire production chain. The next phase is to develop a comprehensive software solution to facilitate adoption by Hitachi Energy Varennes teams.

Eventually, the aim is also to improve planning by using stochastic optimization techniques to create more flexible and efficient production schedules.

The collaboration between Artelys, École Polytechnique and Hitachi Energy demonstrates the potential gains for the manufacturing industry in using advanced data processing and optimization techniques.

This collaboration was possible thanks to the the support of the IVADO research, training, and knowledge mobilization consortium in artificial intelligence.

For more information about Artelys‘s optimization services, visit our dedicated web page or contact us directly!