Wind Power Prediction Based On SVM And Integrated Learning

In recent years,with the huge consumption o f traditional energy sources and the depletion year by year,the use of renewable energy sources has become a hot spot for sustainable development.Wind energy is one of the most representative renewable energy sources,and the wind power industry is in th e stage of vigorous development and operation in various countries.However,due to the intermittent and fluctuating nature of wind energy,real-time changes in operating conditions will threaten the safety of the power grid and increase the difficulty of power dispatching control.Therefore,prediction methods based on operating condition identification have become hot research topics.The prediction based on the operating condition identificati on not only accords with the actual situation of the project,but also improves the prediction accuracy.Based on the above,this paper proposes a wind power prediction model based on improved support vector machine(SVM)and ensemble learning,which is sp ecifically divided into the following aspects:Firstly,the analysis of SVM algorithm shows many unique advantages in solving the identification problems such as non-linearity and small samples,but it also has disadvantages such as insufficient learning a bility,unable to effectively use big data resources to improve accuracy.In the case of a kernel function,deviations in prediction results may be caused.Therefore,this paper uses the SVM algorithm with a suitable kernel function to build a wind power f orecasting model,and has completed the wind power forecasting.Secondly,the basic theory of ensemble learning and apply ensemble learning to wind power predicti on are researched.Based on the idea of ensemble learning,the operating conditions of wind tu rbines are identified.In the process of operating condition identification,power isometric division,k-means,and fuzzy C-means clustering methods are compared in this paper,and the distance and DB indicators are calculated to measure the effectiveness of the three identification methods.Fuzzy C-means clustering is selected to complete the identification of the operating conditions of the wind turbine in the end.Finally,a sub-learner model is established based on SVM under the identified operating conditions.With regard to the selection of the main learner,the impact of the operating conditions on the main learner is considered in this paper,and an improved main learner model based on the distance from the sample to the cluster center to the training of the main learner model is proposed.Through experimental analysis,the main learner is compared with the weighted average main learner and the SVM-based main learner.Since the operating conditions are taken into account in the improved main learner,it is more in line with the actual situation of the project and has high prediction accuracy,which is suitable for wind power forecasting.