An Early Warning Model for Equipment Based on State Vector and LSTM

Effective equipment status early warning is beneficial to acquire the deterioration trend of the equipment in advance, and ensure the safe operation of the equipment. In this paper, we propose a state early warning model for power equipment based on the state vector and LSTM algorithm. Based on the LSTM prediction algorithm, a data prediction model is established by using the normal data of the equipment, and the early warning threshold of the equipment measurement point vector is calculated based on the state vector accordingly. The equipment status fault warning information is pushed when the equipment measurement point vector exceeds the threshold. We validated the method proposed in this paper using the actual data from the steam turbine equipment of the power plant. The early warning time of the extraction pressure measurement point of the steam turbine equipment is about 90 minutes, which is earlier than the actual time; and the early warning prompt of the exhaust steam temperature measurement point is about 10 hours earlier than the actual time. The results show that the early warning model can send early warning signals in advance, so the early warning model is useful, and it can assist the production and operation of the equipment effectively.