Research On Application Of Wind Power Prediction And Hybrid Energy Storage System In Wind Farm

With the high speed development of world economy,and the exhausting of traditional resources,the development of new energy has gradually become the focus of each county.Wind energy has become one of them for its abundant reserves,small environmental pollution and other advantages.Wind-generated electricity is the main method for wind energy development which is used to convert wind energy into electricity for people's daily life.However,due to its inherent randomness,intermittent and other poor controllability,wind power will directly impact on the grid and affect the power quality of the grid.Therefore,it is great significance to study the controllability of wind power and improve the stability of wind power grid-connected.Wind power forecasting and smoothing of fluctuation by using storage equipments in wind farm are two important methods to improve the reliability of wind power output.In this thesis,the two measures are combined aiming to stabilize wind power and provide corresponding methods for wind farm.In terms of wind power forecasting,Support Vector Machine(SVM)has been widely used in wind power forecasting in recent years.However,it still has disadvantages of inflexible selection of kernel functions,such as Mercer conditions,more parameters,complicated optimization,large time cost and so on.In view of the above-mentioned deficiencies of SVM in wind power forecasting,Relevance Vector Machine(RVM)model is adopted in this thesis,which is similar to SVM.RVM overcomes the disadvantages of SVM in wind power forecasting.In terms of forecasting methods,the Ensemble Empirical Mode Decomposition(EEMD)is used in this thesis to decompose the original wind power sequence to reduce the non-linearity of the wind power and obtain a series of sub-sequences decomposed.Sample Entropy(SE)algorithm is used to evaluate the complexity of subsequences and to group sub sequences.RVM modeling is carried out for new subsequences.An adaptive hybrid differential evolution algorithm is adopted to avoid RVM's blind parameter selection.Finally,each subsequence is superimposed to get the final prediction.An example of northwest wind power prediction is given to demonstrate the feasibility of the RVM prediction method.In terms of the hybrid energy storage system,firstly,the principle of selecting hybrid energy storage systems is introduced,which combines energy-type storage equipment and power-type storage equipment.Those two kinds of devices can suppress high frequency fluctuation and low frequency fluctuation of wind power respectively.The RVM wind power forecasting method is used to select the target of wind power for hybrid energy storage system.The RVM wind power forecasting method is used to decompose the low-frequency part of theEEMD.The prediction result is the expected target output,and then the total target power of the energy storage system is determined based on the energy flow relationship.The target power values of the energy-type and power-type energy storage devices are tested according to the SOC value at the previous time.Adjusting the target power value avoids overcharge and overdischarge of the energy storage device.Finally,the simulation result verify the effectiveness of the power allocation method which based on the state of power adjustment of hybrid energy storage system.