Wind Turbine Health Evaluation And Prediction Based On XGBoost Algorithm

In recent years,the scale of our country’s wind turbine assembly capacity has continued to expand,with a total installed capacity of more than 300 million k W,ranking first in the world.However,many major accidents of wind turbines caused by complex wind conditions have brought great challenges to the development of the wind power industry.The operating state of wind turbines not only determines the power generation output but also affects its working life.Therefore,the operation and maintenance of wind turbines have become a hot spot in the field of wind power generation.At present,some achievements have been made in the research on wind turbine component fault warning and operation strategy.Most of the existing articles evaluate the health status of unit components based on a single evaluation model,and there are few studies on the overall health status of the unit.In the actual operation of wind farms,we often need to consider the overall health performance of wind turbines to estimate the remaining life and production capacity of wind turbines.Therefore,it is more practical and valuable to study the health performance evaluation and prediction of wind turbines.In this paper,in view of the problems of reduced sensitivity of health status assessment,difficult division of health performance assessment areas,and low performance prediction accuracy in complex environments,we study the health status assessment of wind turbines,the health performance assessment and prediction of wind turbines.The specific work is as follows:Aiming at the problem that the randomness of the weights of wind turbine health indicators will reduce the sensitivity of wind turbine health status assessment,we propose a wind turbine health status assessment method based on a random combination weighted fuzzy evaluation.First,we establish the unit health status evaluation index framework and propose a random combination weight calculation formula based on random factors.Second based on the ridge-shaped distribution function,a calculation function of the health index deterioration membership degree is established,and the health state fuzzy comprehensive evaluation mathematical model of wind turbines is constructed by combining the random combination weight and membership function.Finally,we conduct experimental analysis on several units of Dalian Tuoshan Wind Farm to verify the effectiveness and accuracy of the proposed method.Aiming at the problems of difficult division of wind turbine health performance evaluation area and low accuracy of health performance prediction,we propose a wind turbine health performance evaluation method and health performance prediction model.First,we propose an automatic power curve limit calculation algorithm based on e Xtreme Gradient Boosting-Bin(XGBoost-Bin)for evaluating the health performance of wind turbines.Second,we build a generalized regression neural network(GRNN)prediction model considering multi-data features based on the regional division data to predict the short-term unit health performance.Finally,we verify the effectiveness of the proposed method through case analysis.