Research On Wind Power Prediction Method Based On Multi-kernel Learnin

uencing factors such as wind speed,wind direction,and temperature,the data characteristics are heterogeneous,which leads to the construction of the SVM model.Its kernel function is difficult to select,and even in the case of extreme climate data heterogeneity,a single kernel function SVM is not applicable.This paper proposes to apply the reinforcement learning optimization method to wind power forecasting to provide reliable support for power system scheduling,planning and operation.This paper summarizes the working principles of SVM,Simple Multi-Kernel Learning(Simple-MKL),Reinforcement Learning(Q-Learning),and Artificial Emotional Learning(Artificial emotional learning).Firstly,the characteristics of different kernel functions SVM are analyzed.The single kernel is not efficient for data learning.This paper proposes a multi-kernel that selects three kernel functions from the four kernel functions of RBF,Linear,Poly and Sigmoids.Learning method,build a multi-kernel learning SVM model for power prediction.In this paper,the Q-Learning algorithm in reinforcement learning is selected to optimize the weights of multi-kernel functions.This method turns the problem that the kernel function is difficult to select in wind power prediction in complex environments into a kernel function.Weight optimization problem,so as to improve the prediction accuracy of SVM and improve the learning efficiency.In the simulation process,the k-fold-cross-validation method is used to obtain the optimal hyperparameters of the kernel function to improve its generalization ability;traditional multi-kernel learning for the same data space,there is repeated calculation of the kernel function,which affects the learning speed of the model.Emotional Q-learning improves model convergence speed.Finally,compared with the traditional multi-kernel learning method,it reflects the superiority of the method proposed in this paper.From a longer-term perspective,the multi-kernel learning optimization method proposed in this paper is more suitable for power prediction of gradually rising high-power wind turbines.