Research On Ultra-short-term Wind Speed Forecasting And Hybrid Energy Storage System Planning

In order to cope with the fossil energy depletion,environmental pollution and meteorological deterioration,wind energy is regarded as an important strategy by countries around the world.With wind power penetration increasing,the intermittent and fluctuating of wind power brings rigorous challenges for the stability,power quality and dispatch of power system,thus wind curtailment becomes more serious.The research target of this thesis is to improving accommodated capacity of wind power to reduce the negative impact of wind power on the power system.This thesis focuses on wind speed forecasting technology and energy storage allocation,which can suppress the impact of wind power fluctuations by improving the forecasting reliability and using energy storage to realize time migration of electricity energy.The work can be summarized as follows.Aimd to nonlinear,non-stationary wind speed and power time series,three common time-frequency analysis methods were researched,such as short-time Fourier transform,wavelet transform and Hilbert-Huang transform.Its appication types of non-stationary signals were analyzed.The principle and signal decomposition steps of variational mode decomposition were described in detail.The time-frequency distribution of reference signal obtained by diffierent method verified variational mode decomposition has a high time-frequency resolution.The minimum mean envelope entropy of mode functions were chosen as the fitness function.And then the enhanced particle swarm optimization was built.Meanwhile,the modified variational mode decomposition was proposed to reduce the negative impact by the number of modes K and the parameter ?.In this method,By comparing the wavelet decomposition,empirical mode decomposition and modified variational mode decomposition analysis results of the actucal wind speed and power time series,it's provide more detailed and regular pre-processing data for building ultra-short-term wind speed or power forecasting model.By analyzing the modeling theory and the performance characrerisrics of different forecasting models,the broad learning system which has timeless and rapid renewal forecasting parametes was introduced into ultra-short-term forecasting model.The fractional auto regressive integrated moving average-broad learning system,modified variational mode decomposition-broad learning system and subseries reconstruction-broad learning system combined forecasting models for ultra-short-term wind speed were proposed.Based on the fractional auto regressive integrated moving average model and the analysis of long-correlation wind speed time series,the fractional auto regressive integrated moving average-broad learning system forecasting model is established.It's of short-term forecasting accuracy is superior to that of single forecasting model.The modified variational mode decomposition-broad learning system forecasting model was established.The modified variational mode decomposition is used to preprocess the wind speed time series,which reduces the complexity of input data and the difficulty of modeling,and improves the accuracy of forecasting model.In addition,combine the sample entropy theory,the subseries reconstruction-broad learning system forecasting model was built.The complexity of each subsequences were evaluated by sample entropy,the high and low entropy subsequences were adaptively selected and reconstructed.The forecasting models of high/low entropy subsequences were created by broad learning system and auto regressive integrated moving average respectively,which can effectively reduce the model size and improve the accuracy and efficiency of forecasting.In order to suppress the adverse effects of wind power fluctuations on the power grid,a capacity allocation method of hybrid energy storage system for smoothing wind power fluctuations was presented.Based on the spectrum analysis results of wind power output,the optimum cutoff frequency and minimum absorb power of energy storage system were determined by recursive calculation,which satisfies the power ramp rate rule of connecting wind farm to power system.According to the characteristics of fluctuation and randomness,a hybrid energy storage system composed of battery and supercapacitor was designed.And then the operating frequency bands of battery and supercapacitor were determined by the electrical characterisrics of both devices.The rated power and capcacity of hybrid energy storage system were calculated considering the result of power distribution,charging and discharging efficiency of hybrid energy storage system and the state of charge limitations.The simulation result of historical wind power data study demonstrated the validty of hybrid energy storage system method to suppress wind power fluctuations.The economic capacity allocation model was proposed for the HESS composed of battery and supercapacitor.A power management strategy based on the battery operation period and the cutoff frequency of the battery absorbed power was designed,which combined with battery capacity adjustment factor to prolong the battery lifetime.The battery lifetime loss factors were analyzed,the fitting function of cycle life and the lifetime loss model of battery were built according to the relationship between discharge depth,cycle life and throughput energy.Meanwhile,the simplified cycle life model of supercapacitor was presented.In order to improve the calculation precision and efficiency in the process of allocation model with average cost per day as objective function,a bybrid parallel PSO-GA optimization algorithm was proposed.The simulation results shown the optimized power and capacity of battery and supercapacitor can meet the quality of power ramp rate,it can effectively prolong lifetime of energy storage and ruduce the system cost.