Energy-efficient and quality-focused control of conveyor belt dryers in petrochemical production

Abstract

The petrochemical industry’s dryers, which operate at high temperatures, consume substantial energy and require precise temperature control to maintain product quality. To address these challenges, conveyor belt dryers are used for continuous drying, but optimizing their operation is complex. Traditional control methods often rely on operator experience or fixed settings, which may not be optimal. This paper introduces a novel approach using a Graph Neural Network and Multiagent Deep Deterministic Policy Gradient algorithm. This method models dryers’ components—chambers, fans, and belts—as cooperating agents, optimizing temperature, fan speed, and belt speed based on quality predictions. The algorithm, trained with industrial data, improves decision-making by integrating quality predictions into its reward function. Extensive experiments show a 5% improvement in product quality and a 3% reduction in energy consumption per time step, translating to significant energy savings. This approach enhances both efficiency and product quality control in conveyor belt dryers.