论文部分内容阅读
针对航空发动机状态时间序列预测中嵌入维数难于有效选取的问题,提出一种基于嵌入维数自适应最小二乘支持向量机(LSSVM)的预测方法。该方法将嵌入维数作为影响状态时间序列预测精度的重要参数,以交叉验证误差为评价准则,利用粒子群优化(PSO)进化搜索LSSVM预测模型的最优超参数与嵌入维数,同时通过矩阵变换原理提高交叉验证过程的计算效率,并最终建立优化后的LSSVM预测模型。航空发动机排气温度(EGT)预测实例表明,该方法可自适应选取适用于状态时间序列预测的最优嵌入维数且预测精度高,适用于航空发动机状态时间序列预测。
Aiming at the problem that it is difficult to select the embedding dimension effectively in time series prediction of aeroengine, a prediction method based on embedding dimension adaptive least squares support vector machine (LSSVM) is proposed. This method takes embedding dimension as an important parameter that affects the prediction accuracy of state time series. Using cross-validation error as evaluation criterion, the optimal hyperparameters and embedding dimension of LSSVM prediction model are evolved using Particle Swarm Optimization (PSO) The principle of transformation improves the computational efficiency of cross-validation and finally establishes an optimized LSSVM prediction model. Aeroengine Exhaust Temperature (EGT) prediction example shows that this method can adaptively select the optimal embedding dimension suitable for the state time series prediction with high prediction accuracy and is suitable for the prediction of aeroengine state time series.