Research On Ultra-Short-Term Wind Power Forecast Of Wind Farm

Currently,wind power is developing rapidly and has become one of the important forms of clean energy power generation.Compared with conventional energy,the characteristics of wind power itself lead to the low reliability of wind power generation,which makes it a difficult problem to balance the active power and reactive power of the power grid when large-scale wind power is connected.Accurate prediction of wind power can not only help the power system formulate corresponding dispatch control strategies in advance,but also ensure the safe and stable operation of the power system.The article conducts research from the following aspects:(1)Introduce three data decomposition methods: empirical mode decomposition(EMD),ensemble empirical mode decomposition(EEMD)and complete integrated empirical mode decomposition(CEEMD)to reduce the complexity of the data.The time-frequency characteristics and reconstruction errors of the above three methods are compared and analyzed longitudinally.Sample entropy is introduced to quantify the complexity of each component after the three methods are decomposed.As a classic data decomposition method,EMD has a wide range of applications,but this method has obvious shortcomings,so a variety of improved methods have been derived.The EEMD using auxiliary noise can effectively suppress the EMD modal aliasing problem,but this method has a large reconstruction error.The introduction of CEEMD decomposition on the basis of EEMD can both further suppress the modal aliasing problem and reduce the reconstruction error.(2)Based on the CEEMD data decomposition method,a multi-model combined wind speed prediction method based on CEEMD is proposed.The first step is to construct several common machine learning models for the first three components with high complexity,including: BP neural network,RBF neural network,and least squares support vector machine.In the second step,based on the predicted values of the individual models,a combined prediction model based on PSO-NNCT is proposed.Through comparative analysis,the combined prediction model based on PSO-NNCT has higher prediction accuracy.Finally,combining the CEEMD decomposition and the PSO-NNCT combined method,the paper established a CEEMD-PSO-NNCT ultra-short-term wind speed combined prediction model.(3)Considering that a single CEEMD data decomposition method cannot completely deconstruct the time series,an improved time series wind speed prediction model based on the CEEMD-VMD secondary decomposition method is proposed in combination with two different data decomposition methods.The first step is to introduce the sample entropy in Chapter 2 to quantify the complexity of each component after CEEMD decomposition,and perform the second decomposition for the first three components with the highest entropy value.Comparing the sample entropy values of different decomposition methods,the two-layer data decomposition technique can further reduce the complexity of the signal.In the second step,a time series model based on CEEMD-VMD secondary data decomposition is established,and the components after the secondary decomposition are separately modeled to predict.The prediction results of the components are summed to get the final result.The data simulation results show that the method based on the secondary decomposition technology has higher prediction accuracy.(4)Using the historical operating data of the wind farm,the actual power curve is fitted using a neural network method,and the two prediction methods in Chapters 3 and 4are compared.The best wind speed prediction value is selected and the final power prediction value is obtained using the obtained curve.