Industrial Giant Optimizes Warehouse Energy Consumption in 6 Months

Industries: Electronics, Cold Chain, HVAC

Use Case: Energy Optimization, Smart Warehouse

Architecture: Knowledge-First AI (AI-CALM)

Currently, the industrial sector consumes over 50% of the world’s total delivered energy¹. As the global economy continues to expand, energy efficiency will become of utmost importance as a cost-effective way of enhancing energy security, addressing climate change, and promoting economic growth.

Wesco, a global leader in supply chain, electronics, and distribution services, is on the frontline driving environmental responsibility and sustainability for their customers. A large customer of Wesco distributes electronic products across the globe, with hundreds of warehouses and distribution centers. With such a large footprint, optimizing energy use could save them millions annually.

By applying Aitomatic Knowledge-First AI, Wesco helped their customer deploy an energy optimization application within 6 months delivering an estimated 30% in energy savings.

Time to Completion

90 days

Time to Value

6 mos

Energy Savings

30%

The Problem

Not enough data for Machine Learning

Wesco had no available data on the energy consumption of their customer’s warehouses, as these buildings have historically operated at fixed settings. This lack of data is commonplace in the industry, but presents a significant obstacle to understanding the relationship between warehouse settings, such as AC temperature and fan speeds, and energy consumption. Without this information, it’s difficult to develop a reliable application for warehouse managers to understand how different settings will affect their energy usage.

Traditionally, implementing machine learning (ML) methods to solve this problem would have been impossible without years of manual, repetitive data collection across every warehouse and input setting for the machine to “learn” these relationships.

Wesco faced a daunting challenge - how can they create a system to recommend optimal control settings to reduce energy consumption while maintaining personnel comfort and safety?

The Solution

Knowledge-First AI Comes To The Rescue

1

Translate
domain-specific knowledge

2

Combine it with
Machine Learning

3

Deploy Production-ready
ML Solution

Instead of waiting for years to collect data across every warehouse, Wesco used Aitomatic Knowledge-First AI to take advantage of domain knowledge related to both warehouse operations and common sense relationships between control settings and energy usage. By modeling this knowledge and ensembling it with ML models that capture energy consumption patterns, Wesco can immediately provide its customer with an application that allows warehouse managers to interact with real-time data and make informed decisions to minimize energy consumption.

For instance, Wesco's application provides energy optimization suggestions for their customers' HVAC systems, one of the largest energy consumers in industrial warehouses. It suggests certain warehouse settings, such as AC temperature or fan speeds, to minimize energy consumption while honoring the operational constraints for the comfort and safety of warehouse occupants.

With Aitomatic, Wesco quickly built a working Al solution without years of manual data collection, resulting in a more accurate and up-to-date understanding of energy consumption. This not only helped Wesco provide its customers with significant reductions in energy costs and increased operational efficiency, but can have a significant impact on reducing the overall energy consumption of the industry.

K-Oracle

Knowledge can be encoded directly as a Teacher

Wesco worked with Aitomatic’s K-Oracle architecture to quickly develop an application that could (A) visualize the impact of adjusting warehouse settings like AC temperature and fan speed on energy consumption and (B) determine the optimal warehouse settings to minimize energy consumption during specific months of the year.
‍
By using K-Oracle's ‘Teacher’ component to encode the heuristic statements made by Wesco's domain experts, K-Oracle was able to automate data labeling, enhance model accuracy, and train a machine- learning ‘Student’ model that handles edge cases smoothly and generalizes well. The Ensemble component in the K-Oracle architecture combined the decisions of the ‘Teacher’ and ‘Student’ to provide expedited development of the application, while incorporating business logic and safety precautions.

With this approach, Wesco was able to accelerate time to market with a commercially viable energy optimization solution for its customers. As a result, their customer is already realizing significant energy savings while Wesco is expands its energy optimization service.