Research On Power Forecasting And Fuzzy Optimal Control Of Wind Power Generation System

In order to deal with the energy crisis and solve environmental problems,the renewable energy represented by wind power has developed rapidly in recent years.The uncertainty and fluctuation of wind power will have an adverse impact on the operation and scheduling of power system.With the increasing installed capacity of wind power,the control precision of wind power generation system needs to be improved.Wind power forecasting and optimal control of wind power generation system can effectively reduce the uncertainty caused by wind power integration,ensure the safe and stable operation of wind power generation system and improve the reliability of power system.The main contents of this thesis are as follows:The iterative filling algorithm based on random forest is used to reconstruct the missing values of the original data,and the local outlier factor algorithm is improved by using Mahalanobis distance to eliminate the influence of dimension and variable correlation,so as to realize the detection of outliers in the original data.Data preprocessing can effectively improve the quality of sample data and provide a good data basis for wind power forecasting.The long short-term memory neural network is used to explore the time-series dependence of power data and the gradient descent is optimized.The dropout method is used to prevent over fitting.The random forest model based on bagging ensemble and xgboost model based on boosting ensemble are established.A new wind power forecasting model is proposed by stacking the three models.The training set is further divided into sub-training set and validation set,and the hyperparameter is selected on the validation set using random search to avoid the risk of information leakage.By analyzing the importance of input features based on shap value and xgboost,it is proposed to modify the wind speed data of the numerical weather prediction,so as to build an important feature,which can effectively improve forecasting accuracy.The forecasting accuracy of three models and stacking model is evaluated by using the actual operation data of a wind farm in Northwest China.The results show that the stacking model can significantly reduce the forecasting error and has good forecasting performance.The second hour harmonic mean accuracy is used to evaluate the ultra short-term forecasting performance of the stacking model,and the average forecasting accuracy is 86.59%,which indicates that the stacking model can achieve a high accuracy of iterative forecasting.Based on MATLAB/Simulink platform,the model of direct drive permanent magnet synchronous wind power generation system is established.The control objectives of maximum power point tracking and constant power output are achieved by using pitch control and vector decoupling control.The accuracy and rationality of the model are verified by dynamic simulation.An adaptive controller based on fuzzy logic is designed.The parameters of the controller can be adjusted in real-time through the input of error and error change rate based on fuzzy rules.The simulation analysis is carried out under different wind speed input conditions,and the control effect is compared with the conventional PI control.The results show that the fuzzy controller has good robustness and reliability.