基于正则化贪心森林的多维频率指标智能化预测方法

Intelligent Prediction for Multi-dimensional Frequency Indicators Based on Regularized Greedy Forests

为实现海量预想故障下电力系统频率响应性能的快速和精确感知,文章提出了一种基于正则化贪心森林(regularized greedy forests,RGF)的多维频率指标智能化预测方法。该方法采用RGF建立特征输入与多维频率指标之间的非线性映射关系,通过对全局参数进行优化,并引入3种正则化机制,使所构建的机器学习模型能够有效表征复杂函数,防止过拟合。为保证预测算法的性能,通过网格搜索遍历参数组合,以确定所构建RGF模型的最佳参数。在改进的IEEE RTS-79系统上开展了算例测试,与时域仿真、随机森林和梯度提升方法所得结果进行比较,验证了所提方法的准确性、快速性以及良好的泛化能力。

In order to realize fast and precise perception of the frequency response performance of the power system under the background of massive anticipated faults,this paper proposes a method for predicting the multi-dimensional frequency indicators based on regularized greedy forests(RGF).This method establishes the non-linear mapping relationship between input features and multi-dimensional indicators through RGF.By optimizing global parameters and introducing three regularization mechanisms to the decision-making forest,the RGF can effectively represent complex functions and prevent overfitting.To ensure the performance of the model,the combinations of parameters are traversed by the grid search to find the best parameter of the constructed RGF model.Case studies on the modified IEEE RTS-79 system demonstrate the high precision,rapidity,and well generalization ability of the proposed method.