Wind Turbine Abnormal Data Cleaning Method Considering Multi-Scene Parameter Adaptation

Abstract

Aiming at the difficult problem to accurately separate the wind speed - power abnormal data of wind turbine from the normal data before the process of wind power curve fitting, this paper proposes an abnormal data cleaning method considering multi-scene parameter adaptation. Firstly, the data is preprocessed based on the time series characteristics and correlation relationship of wind speed and power data to reduce the density of abnormal data, and then the data is cleaned by DBSCAN (Density-Based Spatial Clustering of Applications with Noise), in which the parameters are optimized by the improved particle swarm optimization (PSO) algorithm according to different wind farms and type of wind turbines. The method makes the decrease of the evaluation values by 56.94% and 58.65% respectively with different abnormal data distribution characteristics compared with change point- quartile method, thus making better cleaning effect on the wind speed- power data.