Poster
Client Context
The Georgia Power GRID Investment Program (GRID) is a multi-year initiative to enhance the service and reliability of the energy grid in communities across Georgia.
The current system at GRID consists of 3 entities: Engineering, Procurement, and Construction. Engineering creates a Bill of Materials (BoM) for various projects that detail which items will be needed for its construction. Using this BoM, Engineering sends estimates for each item to Procurement to order. Procurement then compares the estimates to the current balance of that item in GRID’s warehouse and decides how much to order. Procurement orders these items with an additional buffer to ensure there is enough on hand when construction begins. Once a project has started, Construction will put in an item request order to Procurement. Procurement then sends the requested items to Construction so that the project can proceed.
Project Objective
The GRID Investment Program recently had a change in program objective from making sure that all their projects started on time to upscaling the number of projects done in a year. To effectively implement this new objective, they intend to first reduce the amount of surplus inventory in their warehouse, which was the initial problem tasked to our team. Our team identified $43 million worth of inventory that would not be used in the upcoming year and discovered 3 main reasons behind the buildup of surplus inventory: discrepancies in item labels between Engineering and Procurement, inflated buffer strategy by Procurement, and communication gaps between Construction and Engineering.
Design Strategy
To reduce this surplus inventory, we created a twofold solution that includes both a short-term and long-term solution. In the short term, the team wanted to focus on addressing and reducing the current surplus inventory that exists in their warehouse. In the long term, solutions were developed to prevent the accumulation of future surplus inventory through a redesigned procurement process and proposed system changes.
Deliverables
The first deliverable, the Automated Inventory Clean-Up Tool, uses a Python Graphical User Interface (GUI). The tool takes in the Engineering project BoM database, the current inventory database, and a GRID Approved list of items and then outputs an Excel sheet that provides a recommended action based on the classification of an item: Remove, Clean, or As Is.
The second deliverable the team provided to GRID is a redesigned procurement process that utilizes logistic regression and an optimization model. The process is designed to work like this: Once Procurement receives item estimates from Engineering, they can run the items through the logistic regression. If the logistic regression model indicates that the item is historically correctly estimated, no buffer is applied. Otherwise, Procurement can use the optimization model which will tell them the appropriate buffer to apply to that item to minimize excess inventory with the constraint that the item will have a service level of 95% or greater.
The team’s final deliverable to GRID was a proposed system change. In order to fundamentally resolve the differences between what is estimated to be needed and what is used on site, the team proposed to GRID that a formal communication channel should be established between the Engineering and Construction teams in order to provide a feedback loop through which the project BoMs and estimates can be updated and adjusted based on what is currently used on field.
Value and Impact
The team developed a dual-strategy solution to manage GRID’s surplus inventory challenges and to facilitate their objective of project upscaling. The short-term solution provides an Automated Inventory Clean-Up Tool that identifies $800,000 of stranded items and rectifies discrepancies in the current inventory database. Furthermore, the Clean-Up tool identified $2 million worth of excess items currently in inventory that were mislabeled in the inventory database. For these items, an item ID label change will be recommended to Georgia Power, a solution that will greatly simplify Procurement’s decision process and reduce surplus items.
For the long term, the team designed a strategy to prevent the accumulation of future excess inventory by analyzing and recommending corrective actions. A logistic regression model was developed to verify the accuracy of item estimates, complemented by an optimization model. If this proposed buffer selection process had been applied by Procurement in the past, of the $40.2 million of current excess that had not been identified as stranded or mislabeled, the logistic regression model would have correctly identified $30.9 million worth of inventory with accurate estimates. Under the assumption that GRID would not have added any buffer to these items, this excess would have been prevented and the $30.9 million would have been saved, therefore reducing excess inventory costs by 75%. For the remaining $8.4 million of excess inventory, if the optimization model’s buffer had been used, 75% of this excess would have been reduced. Therefore, GRID would have saved another $6.3 million from excess inventory.
Project Information
Student Team
Vaishnavi Duvvuri, Irene Feijoo, Annette Gisella, Shawna Kalladanthyil, Ami Patel, Sreya Srinivas, Clara Wu, Shanru Xu
Faculty Advisor
Faculty Evaluator
