Metering Fault Diagnosis Model Based on Deep Feature Fitting Network with Mixture of Experts
摘要
In the field of power metering, automatic monitoring and analysis of equipment alarm events are essential for stable operation of the metering system. To address the problem of fault diagnosis for metering equipment, a mixed-expert model based on deep neural networks is proposed. The proposed model combines the advantages of mixture-of-experts (MoE) and Deep feature fitting networks (DFFN). The MoE system can divide the measurement warning problem into several sub-problems for processing, which greatly improves the system's ability to recognize and predict warning problems. On the other hand, the DFFNs can be used to analyze measurement data, fit complex features, and better analyze fault problems.