Research on Reduced Scene Sets Based on ARMA Model of Wind Farms Day-ahead Total Output Forecasting

2018 2nd International Conference on Power and Energy Engineering

Wang, Hui Chao^1 ; Liu, Lei^2 ; Ma, Jin Hui^3 ; Ding, Jie^1 ; Zhou, Chang^1

Renewable Energy Research Center, China Electric Power Research Institute, Nanjing

210003, China^1

Power Dispatch Department, State Grid Jiangxi Electric Power Company, Jiangxi

330077, China^2

Power Dispatch Department, State Grid Anhui Electric Power Company, Anhui

230061, China^3

关键词: Evaluation indicators;    Important samplings;    Latin hypercube sampling methods;    Prediction correction;    Prediction errors;    Quasi Monte Carlo methods;    Reduction models;    Time-series data;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/192/1/012045/pdf
DOI  :  10.1088/1755-1315/192/1/012045

来源: IOP

【 摘 要 】

In this paper, the ARMA model is used to linearly fit the time series data of wind farms' prediction error with the software E-views. And then the error linear curve is sampled by four sampling methods, including random sampling, important sampling method, Latin hypercube sampling method and quasi-Monte Carlo method, to obtain some Ascending and disorderly samples respectively. Finally, the reduced scene sets are obtained by substituting the samples into the scene reduction model. Through analysing the reduced scenes output curve with the evaluation indicators of wind farms' forecast output curve, we find that the reduced scenarios are closer to the actual output curve than the traditional predicted. It also can be concluded that they have great effect on prediction correction and sampling methods have little effect on the output trend of reduced scenarios. Whereas, comparing the reduced scenes' output curve before and after sorting the sample data, the disorder and randomness of the sample data will lead to great volatility in the reduced scenes.