Fault Knowledge Acquisition And Intelligent Fault Diagnosis Of Pumped Storage Unit Based On Ontology

Pumped storage units can effectively solve the problem of wind power and photovoltaic power consumption in the grid,which is conducive to better development of new energy sources.However,their complex operating conditions and frequent transitions make them prone to faults,which affect the stability of the units and even the grid.There is a complex correspondence between fault causes and fault characteristics,so fault diagnosis is a complex project involving multiple disciplines.In this paper,we carry out fault diagnosis of pumped storage units on the basis of knowledge expression by ontology theory,combined with counterfactual theory.Firstly,in view of the complex structure of pumped storage units and the fragmentation of fault knowledge,the main structure of the units is analyzed and classified according to "unit-system-subsystem",and the unit structure is divided into pump turbine,generator motor,speed control system and pressure diversion system,and further subdivided.The common faults of the system were sorted out,and the causes of faults were subdivided based on the FTA method,and the modes of faults and effects of faults were analyzed based on the FMEA method,and the common faults of seven major categories and twelve minor categories were summarized by FTA and FMEA.Secondly,by introducing the construction method,construction principle and construction tools of the ontology,and based on the constructed FTA and FMEA,the knowledge map acquisition and knowledge base modeling of typical faults of unit equipment and units based on the ontology theory and semantic network technology are carried out,and the query function of the ontology software is used to realize the query of fault information.Finally,the structural causal model based on one of the causality frameworks is used for fault diagnosis.The structural causal model built based on Bayesian networks is the most different from Bayesian networks in that the causal relationships in the model are considered as deterministic and the uncertainty is expressed by introducing exogenous variables.The prior probabilities in structural causal models are usually given by experts,but the prior probabilities given by experts are usually fuzzy in nature,and the fuzzy expert experience is transformed into definite probability values by fuzzy theory.Based on the structural causal model,the expressions of occurrence probability and disappearance probability based on counterfactual reasoning are proposed to quantitatively measure the magnitude of the probability of occurrence of the cause of failure.To address the problem that the probability of occurrence and probability of disappearance are difficult to calculate,twin networks are used to approximate the formulas and complete the calculation with the exact inference of Bayesian networks,and the diagnosis of faults is completed by comparing the calculation results.