Research On Health State Evaluation And Fault Early Warning Of Wind Turbine Based On Deep Learning

Wind energy is a kind of new energy with great environmental protection benefits.As a kind of wind energy utilization,wind power generation has a mature operation mode.With the development of wind power,how to ensure the more efficient and safe operation of wind power equipment has become a new problem.Because wind power equipment is in a bad environment and prone to failure,the key to solve the problem is to give early warning of the failure and carry out maintenance before the failure.This paper studies the SCADA operation data of wind turbine.The earl y warning model of wind turbine failure is established and the overall state of wind turbine is evaluated.First of all,according to the fault statistics of wind turbine,the most representative three state parameters of wind turbine are selected,which are power,rotor speed and gearbox oil temperature.Secondly,the random forest algorithm is used for feature selection and data mining to select the five stat e parameters most related to power,rotor speed and gearbox oil temperature.Finally,the long-term memory network and deep belief network are combined to predict the three wind turbine state parameters.The kernel density function is used to estimate the error distribution in the fault early warning model,and the relationship between the error and the unit state is obtained.In order to evaluate the overall state of wind turbine,the concept of cloud model is used instead of the traditional fuzzy evaluation method to establi sh the state evaluation model.Wind power indicators are divided into evaluation target layer,evaluation target layer and evaluation index layer.Entropy weight method and Fuzzy AHP are used to get the weight of wind power index.Considering that the component temperature is easily affected by the external temperature and output power,the component temperature index is compensated.The concept of deterioration degree is used to describe the relationship between the actual state of wind turbine and the state of failure.The membership degree of evaluation time index is calculated by cloud model,and the overall operation state of wind turbine is finally obtained.The results show that the fault early warning model can get the specific fault information,which has guiding significance for the maintenance of wind turbine.The overall state evaluation can reflect the operation deterioration of the unit in time and leave enough time for the fault maintenance of the wind turbine.