Wind Power Forecasting Based Economic Dispatch Of Wind Turbines

With the maturity of wind power technology,more and more wind farms have joined grid-connected systems.In order to ensure the smooth operation of the wind farm in the process of grid connection and increase the economic benefits of the wind farm.This paper aims at wind power curve modeling,multi-step wind speed forecasting and wind farm economic dispatch.The main contributions are as follows:(1)Wind power curve modeling based on measured data.First,according to the distribution characteristics of the measured data,this paper proposes a data preprocessing method based on the DBSCAN algorithm to pick up the valid data that is actually available.Then,the Bin method is used to obtain the wind speed-power points,the cubic spline interpolation method is used to build the wind power curve model.Finally,this paper uses the kernel density estimation function to perform uncertainty analysis on the wind power curve.Experiments are performed using SCADA data from China’s northwestern wind farm.The results show that the DBSCAN algorithm can identify outlier data and keep valid data;use kernel density estimation to perform uncertainty analysis can effectively model probabilistic wind power curve under different confidence interval.(2)Multi-step wind speed forecasting based on error compensation.First,this paper determines the input time series length of the forecasting model based on autocorrelation analysis.Then,this paper proposes a wind speed forecasting model based on error compensation.In which,the initial wind speed forecasting is performed using the LSTM model,and then the initial wind speed forecasting error is fitted using the GPR model.The GPR fitted result is used to compensate the initial wind speed forecasting value.Finally,this paper proposes to train the prediction model for each prediction step to alleviate the superposition error.Experiments are performed using SCADA data from a wind farm.The results show that these methods can further improve the accuracy of multi-step wind speed predictions on the basis of improving the accuracy of single-step prediction.(3)Wind power forecasting based economic dispatch of wind turbines.First,by analyzing the Weibull distribution of wind speed in natural environment,this paper proposes a cost function for wind power generation based on predicted wind power.Then,considering the intermittentness and fluctuation of wind speed,wind power generation has to conduct uncertainty analysis.This paper uses the coefficient of variation to quantify the uncertainty of wind power prediction,then adjusts the operational constraints of wind turbines.Finally,the Q-learning algorithm is used to optimize the economic dispatch model of the wind farm proposed in this paper.Experiments are performed using SCADA data from a wind farm,and the calculated example verifies the effectiveness of the economic dispatch model of the wind farm described in this paper.