Font Size: a A A

Analysis Of Wind Turbine Control Strategy Based On State Prediction

Posted on:2020-06-14Degree:MasterType:Thesis
Country:ChinaCandidate:P F ChenFull Text:PDF
GTID:2392330596995380Subject:Control engineering
Abstract/Summary:PDF Full Text Request
The proportion of wind energy as a clean energy source in the energy sector is increasing year by year.In the limited life of wind turbines(usually 20 years),the conversion of wind energy into electrical energy is the common goal of the entire wind power industry.In the cost of wind power generation,the proportion of wind turbine maintenance is very large.The maintenance cost of onshore wind farms is about 12.5%,and the maintenance cost of offshore wind farms is as high as 22.5%.Therefore,the purpose of this paper is to solve two important problems in the wind power industry:more wind energy is converted into electricity,and the other is to reduce the cost of fault maintenance of wind turbines.This paper proposes a wind turbine control strategy analysis based on state prediction.Using the potential information of the operating state of the wind turbine,the faulty components and time of the wind turbine are discovered earlier.Then,the control method is selected according to the prediction result to prolong the time of occurrence of the failure and reduce the damage degree of the component,leaving sufficient time for maintenance.The dataset of this paper is derived from the wind turbine operating status information and early warning information collected by the SCADA monitoring system.Raw data is cleaned,pre-processed and feature extracted.And some traditional methods of processing data are added to add new features,such as Fourier transform.The feature data is divided into a training set and a test set.Design different predictive classification algorithm models and analyze the experimental results.Compare the advantages and disadvantages of various models.The algorithms involved in this paper are Random Forest(RF),Support Vector Machine(SVM),XGBoost and Long Short-Term Memory(LSTM).Finally,the combination of prediction results and control strategies maximizes the economic benefits before failure occurs.Two scenarios are examined that lead to different amounts of production loss.The effectiveness of condition based control is demonstrated by comparative experiments.Drawing shows the relationship between downtime and loss of power generation in the event of a wind turbine component failure.The research shows the load reduction control strategy of different components of wind turbines.
Keywords/Search Tags:state prediction, SCADA data, feature engineering, algorithm model, control strategy
PDF Full Text Request
Related items