The Research On Modeling Of Wind Power Time Series Based On Multi-scale Analysis

With the depletion of fossil resources and the deterioration of air pollution,low carbon economy and sustainable development have become a consensus in the development of human society.As clean renewable energy,wind power is highly promising.However,with the larger-scale integration of wind power,the safe operation of power system is facing a crucial test due to its randomness,volatility and intermittency,which restricts the further development of wind power generation.Time series modeling of long time span wind power,which can accurately emulate the output features of wind power and analyze/evaluate its volatility,plays a fairly important role in the medium to long-term wind power scheduling and the power system planning.Based on the existing research on wind power fluctuation characteristics and wind power time series modeling method,this paper proposes a multi-scale wind power time series modeling method and the research mainly focuses on:(1)Introducing the mathematical morphology method and applying it to filter the wind power.Other than traditional frequency domain analysis methods,it detects and processes signal waveform directly in time domain,avoiding the problem of phase and amplitude deviation that may exist in frequency domain analysis.Moreover,mathematical morphology only involves simple operations such as translation,addition and subtraction,and extreme value,which reduces the computational complexity and improves the processing speed.(2)Introducing multi-scale mathematical morphology analysis method and WMMF filter.In each WMMF filter,the wind power curve is scanned by structure elements of different length and height,and hence the fluctuation characteristics of different scales are extracted.Multi-scale analysis can excavate precise morphological features.By setting two level WMMF filters,the original wind power time series is decomposed into low frequency trend component,medium frequency fluctuation component and high frequency fluctuation component.(3)Analyzing the statistic of different frequency components,and modeling them based on the characteristics of volatility characters as well as the correlation between components with different frequency.The results of the above statistical analysis and(multi-dimensional)random sampling are used to simulate the components of different frequencies.Simulation sampling is carried out in the order of low frequency,intermediate frequency and high frequency.The sum of sampling results is the final time series of wind power simulation,i.e.historical output sequence is obtained through statistical analysis method,while time series of similar statistical characteristics is simulated.(4)Modeling historical wind power output data from Gancheng wind power farm in Dongfang city,Hainan province during the period of 2015 to 2016,and generating a one-year simulation of wind power output time series based on 1 min sampling period.Using the real output data from from Gancheng wind power farm in 2017 as test sequence,the accuracy of the modeling method used in this paper can be verified.