Yokogawa Implements Multiple AI Agents at Gas Plant for Autonomous Control
Yokogawa Electric Corporation is implementing multiple autonomous control AI agents at Aramco’s Fadhili gas plant in Saudi Arabia to control acid gas removal (AGR) operations.
Yokogawa defines autonomous control AI as AI that deduces the optimum method for control independently and is capable of autonomously handling situations it has not previously encountered.
The technology uses multiple, coordinated AI agents of the Factorial Kernel Dynamic Policy Programming (FKDPP) reinforcement learning-based AI algorithm, which was jointly developed by Yokogawa Electric and the Nara Institute of Science and Technology.
These AI agents were developed to autonomize control in areas of the plant that had proven difficult to automate with existing control methods and thus required manual operation. According to Yokogawa, the agents can now autonomously control and optimize AGR operations at the Fadhili plant.
The AI agents were introduced in three phases, progressively optimizing various sections until autonomous control of the core process in the AGR unit was achieved.
Yokogawa first created a simulation of the plant on which to train AI agents, then evaluated their reliability. They were then integrated with Yokogawa's Centum VP production control system to leverage the safety functions of the existing plant.
Implementation is currently undergoing a detailed evaluation, but initial results from the plant demonstrate a 10% to 15% reduction in its amine and steam usage, around 5% reduction in power usage, improved process stability and a significant decrease in manual intervention by operators.
Read more about Yokogawa's work on autonomous plant control.
More industrial AI insights from Automation World:
Industry Turns Up the Dial on AI
Major Manufacturers Using Nvidia AI Technologies in Factory Systems
