The Fusion Modeling For Wind Power Prediction Based On The Similarity Of Fuzzy Reasoning And IOWA Operator

Under the encouragement of the national sustainable development policy, Wind power, with its characteristics of no pollution and endless, gets rapid development. The ratio of wind power in power grid is increasing year by year. However, as the wind has some inherent characteristics, such as intermittent and volatility, which would impact the power grid when the wind power accounted for more than a certain value. This impact would influence the operation of power grid scheduling in power department, or even seriously affect the power quality of power grid and the working status. Therefore, it’s significant to forecast wind power accurately to ensure the safety of power grid and a rational plan.For the sake of enhancing the prediction level of wind power, this paper (Funded under the National Natural Science Foundation of China, No.51277127) puts forward a fusion prediction method which bases on the following two elements:fuzzy inference based upon similarity forecast wind power optimization model and IOWA operator. What’s more, a large amount of simulation research is conducted on MATLAB platform. The major research content is listed as below: (1) The background and significance of the wind power prediction are elaborated, the overseas and domestic research status is summarized, and the main factors that influence wind power prediction level are analyzed.(2) The principle and establishing key steps of wind power forecast model which based upon multivariate time series are studied. Through the co-integration relationship between wind power and its main influence factors, the wind power prediction model is established based on multivariate time series.(3) This paper proposes a way to optimize prediction model by combining the wind power prediction assessing requirements with multiple prediction error evaluation indexes. To begin with, according to the provisions of wind farm forecast precision and qualified rate, it selects the model which could meet the regulation requirements. Then, adopting the idea of maximizing deviations of multiple error evaluation indicators, it will estimate the selected model synthetically and optimize the model according to the comprehensive evaluation values.(4) On the basis of the actual measurement data from wind farm, the change rule of main influencing factors and its correlation with wind power are analyzed. Furthermore, the fuzzy inference method which utilized the similarity of wind farm history data is put forward for model selection.(5) As the prediction accuracy is changing minutely, this paper establishes the fusion prediction model which is based upon induced ordered weighted averaging (IOWA) operator. When there are many single models, for the sake of quicker calculating fusion weight with computer, the computational process of the IOWA operator is optimized. An error information matrix method is used to judge the redundancy of model which could guarantee the precision and timeliness of fusion prediction. At last, the prediction results of the fusion model based on IOWA operator are compared with the results of common wind power combination forecasting models, including the simple arithmetic average, mean weighted approach, the error sum of squares reciprocal method, entropy value method and Shapley value method. A lot of simulations are conducted in this paper to prove that the proposed method has the higher prediction accuracy.