An Interpretable Data-Driven Fuzzy Petri Net Method for Industrial Domain Knowledge Modeling of Energy Efficiency Management

Industrial domain knowledge modeling aims to integrate multi-source industrial information to build a knowledge model that facilitates industrial system construction. Some studies have sought solutions by fuzzy Petri nets (FPNs), which graphically represent domain knowledge to realize interpretable knowledge modeling. However, these methods cannot meet the intelligent development demands of current industrial systems due to laborious manual modeling, inflexible knowledge representation, and non-adaptive parameter determination. In this paper, we propose an interpretable data-driven FPN (IDFPN) method to realize industrial domain knowledge modeling. Unlike traditional FPN modeling paradigm, IDFPN develops a data mining-based explicit knowledge acquisition (DEKA) to automatically explore the FPN structure. Then, a novel q-rung orthopair fuzzy Petri net (q-ROFPN) model and an adaptive model learning algorithm are proposed to dynamically adjust the knowledge representation ranges and adaptively determine q-ROFPN parameters. We apply IDFPN to address a real-world knowledge modeling problem in energy efficiency management of the data center cooling system. Experimental results demonstrate that the proposed IDFPN can fulfill the industrial domain knowledge modeling task automatically and generate an interpretable q-ROFPN model with good parameter learning capabilities and reasonable inferences regarding system behavior statuses.