AI in Refining
Though AI has always fascinated the world, multiple AI hype cycles were followed by “AI winters” due to real-world technological failure. Does the current excitement about AI in the hydrocarbon processing industry actually stem from a new technological breakthrough? The answer is yes. This breakthrough is called Deep Learning, or Deep Neural Networks (DNNs). To fully optimize a refinery or a chemical plant, Deep Learning needs to be operated in closed-loop, in the framework of reinforcement learning…
What’s Next in APC
Most plant variables are being controlled by a standard hierarchical process control and optimization model. However, in most cases, the critical money-making strategies of the refinery are not automated under the traditional APC approach. A neural network solution includes dozens of layers and up to hundreds of millions of nodes and has the internal complexity to capture severe multi-unit nonlinear dynamics. A single neural network operates across the layers of the hierarchy and the various process units…