Decentralized Wind Power Evaluation And Site Selection Research Based On Complex Terrain Area

In recent years,the development of distributed wind power near the power load center,the power generated can be connected to the power grid and consumed locally,has become the next step of China's wind power development.The following problems exist in the development of distributed wind power in the area where the power load center is located:1.The terrain in the area is complex,and the hills,cliffs,steep slopes and ridges all have an impact on wind speed,resulting in uneven distribution of wind energy resources;2.More rain and fog in the area The humidity is large,and it is easy to form a freezing phenomenon,which has a great influence on the normal operation of the wind farm.Firstly,the paper establishes a mathematical model based on terrain,surface roughness and wind farm wake effect under complex mountain conditions,and collects meteorological stations and measured data of a plain area and a mountain area to carry out mountain-type wind energy characteristics.Comparative analysis.The characteristics of wind resources in mountainous areas,their influencing factors and their influencing mechanisms are discussed.Based on the medium-and long-term time scales of meteorological data(including the annual value of the accumulated years,the values of the values of the years,the values of the tense,the monthly and the annual values)and the hourly observation data of the wind towers on the short-term time scale,a hybrid time scale is established based on the similarity coefficients.The XGBoost algorithm is used to compare the slope,elevation and surface roughness of different wind farms with wind speed,and a hybrid time-scale wind speed prediction model considering complex mountain conditions is constructed.The predicted power generation of the wind turbine is obtained by the wind speed-power curve.In view of the variable wind energy in mountainous areas and the large fluctuation of power generation value,Huber Loss and Log-Cosh Loss are used to construct the weight loss function.The original loss function of XGBoost learner is improved and MFCV(Multi Flexible)is used.Cross-Validation)Find the optimal weight parameter.The weighted loss function is less affected by the sudden change value in the sample,and the optimization of the two-step degree can be realized,which can effectively improve the prediction accuracy.The simulation analysis based on the measured data of a mountainous wind farm in Hunan shows that the method can achieve the prediction before the date,and the accuracy rate is 97.2%.Finally,by selecting the extreme value samples of wind turbine blades during ice coating and their generalized extreme value distribution parameters as training samples,through the training and learning of XGBoost network,the nonlinearity between the meteorological conditions of ice disaster and the probability distribution parameters of ice coating degree are accurately mapped.Relationship,thereby predicting the GEV distribution of the degree of icing of the fan blades as a function of meteorological conditions.On this basis,the XGBoost network is used to consider the degree of icing degree and wind power correlation,and the prediction method of mountain wind farm power loss under ice-covered conditions is proposed.The simulation results of a two-year winter wind farm in Hunan Province are carried out.The results show that the method proposed in this chapter can reflect the power loss of the wind farm under ice-covered conditions and is close to the actual operation of the wind farm.The simulation results show that the XGBoost prediction method of distributed wind power mixed time scale and the prediction method of mountain wind power loss rate under iced conditions can accurately predict the target under the complex mountain conditions,which is of great significance to the development of mountainous wind power..