Energy-efficient Pump Control in Industrial Cooling Water Systems Using a Multi-Agent System

Abstract

Nations worldwide are facing the challenge of reducing their greenhouse gas emissions. Thermal energy supply systems in industrial companies account for 75 % of their final energy demand, and are therefore of particular importance. Regarding thermal energy distribution, 3 % of the industrial final energy demand is used for pumps. Especially in cooling water systems, the largest share of energy demand accounts for pumping systems. Currently, industry usually uses operating strategies that provide pressure control without considering individual pump operating points and efficiencies. However, research shows that more intelligent operating strategies can improve the energy efficiency of such systems significantly. Therefore, this paper presents a collaborative multi-agent approach to control the pumping system in a central cooling water system considering the energy efficiency of each pump agent. This approach is compared with a conventional operating strategy using a centralized, rule-based decision making process to control the system's feed pressure.In a first step, each agent calculates the energy efficiency difference to its optimum and whether it wants to increase or decrease its rotational speed to reach its optimum. This information is collected by a organization unit and distributed to all agents. Then, depending on the system requirements and the states of the other agents, each agent decides whether to adjust its own rotational speed or leave it to another agent showing a higher difference to its optimum. The multi-agent system is implemented within a Python framework and validated for a central cooling water system using a dynamic simulation model. For an exemplary use case, the energy efficiency of the overall system improves up to 12.4 % against the baseline while still ensuring operational requirements.