基于WNN-GNN-SVM组合算法的变压器油色谱时间序列预测模型

分析了小波神经网络(WNN)、灰色神经网络(GNN)、支持向量机(SVM)预测方法的原理,利用粒子群优化(PSO)算法对这3种基本预测方法进行了结构参数优化。将WNN、GNN、SVM与PSO-BP算法进行组合,推导得出了组合预测模型最优权系数的计算方法,并优化了组合预测模型拓扑结构参数。算例分析结果表明:经过PSO算法优化后,WNN、GNN、SVM预测模型的预测精度得到了提高,其组合模型较单一模型有更高的预测精度。

Predicting model of transformer DGA time series based on WNN-GNN-SVM combined algorithm

The principle of prediction methods based on WNN(Wavelet Neural Network),GNN(Gray Neural Network) and SVM(Support Vector Machine) is analyzed. The structure parameters of the above three basic predicting methods are optimized based on PSO(Particle Swarm Optimization) algorithm. The WNN-based, GNN-based and SVM-based predicting models are combined with PSO-BP algorithm and the calculation method of the optimal weight coefficient of the combined forecasting model is deduced. Case study results show that the accuracy of WNN-based, GNN-based and SVM-based predicting models is improved by optimizing the topological structure parameters with PSO algorithm, and the combined predicting model has higher accuracy than single models.