Research On Wind Power Forecasting Method Considering The Characteristics Of Wind Data

Wind power generation is an effective way to solve problems of energy shortage and environmental pollution,and has developed rapidly in recent years.However,the randomness and intermittence of wind energy will bring challenges to the decision-making and management of power system.Wind power forecasting can effectively solve the above problems and provide an effective reference for power system planning and dispatching.The data collected from wind farms often contain outliers and missing values,which will lead to the decline of data quality and affect the accuracy of wind power prediction.In order to improve the data quality,this paper firstly studies the wind turbine power curve modeling technology,modifies the abnormal data through the power curve modeling,and constructs a complete and effective wind power data set.Wind power data at different time scales have different fluctuation characteristics and contain different time series information.In order to improve the accuracy of wind power forecasting,this paper fully considers the multi-time scale characteristics of data,and conducts research on wind power prediction technology.The main innovations and specific research work of this paper are as follows:(1)Wind power data were preprocessed based on artificial method and isolation forest algorithm.The quality of wind power data affects wind power forecast results.In fact,the collected wind power data contains uncertainties such as abnormal value and missing value,so it is necessary to preprocess the collected data.Firstly,this paper uses artificial methods to screen out the data that do not meet the operating characteristics of the wind turbine,and then uses the isolation forest algorithm to further identify the abnormal points in the data based on the characteristics of the abnormal data,so as to improve the data quality.(2)A wind turbine power curve(WTPC)modeling method was constructed based on asymmetric characteristics and a hybrid intelligent optimizer.The WTPC can describe the corresponding relationship between wind speed and wind power under actual operating conditions.In this paper,the abnormal data is corrected by accurate modeling of WTPC.Firstly,the influence of abnormal points in data on power curve modeling is analyzed,and an asymmetric loss function is proposed according to the distribution characteristics of modeling errors.Then,based on the principles and characteristics of grey wolf optimizer(GWO)and backtracking search algorithm(BSA),a hybrid intelligent optimizer GWOBSA is proposed.Finally,a set of parametric models with simple structure and interpretable parameters are used to fit the power curve,and the model parameters are optimized based on asymmetric loss function and hybrid optimization algorithm.The power curve model is obtained by using the above modeling strategy.(3)A wind power forecasting method was proposed based on multi-time scale features and a deep learning model.Effective and sufficient features help to accurately describe the fluctuation of wind power.Wind power data at different time resolutions have different fluctuation characteristics.High-resolution data reflects the details of wind power fluctuation,while low-resolution data reflects the long-term trend of wind power.In this research,the fluctuation characteristics of wind power data at different time scales are firstly analyzed,and the multi-time scale features of wind power series are constructed.Then,based on convolutional neural network(CNN)and gated recurrent unit(GRU),a combined model,CNN-GRU,is built to perform feature extraction and time series analysis on wind power sequences.Finally,by fusing the multi-scale features of wind power series,the multi-time scale information of the wind power sequence can be fully mined to improve the accuracy of wind power prediction.