Electromagnetic Relay Suction Prediction Based on Improved Support Vector Regression

• Published in: 2024 IEEE 7th International Electrical and Energy Conference (CIEEC) pp. 1237 - 1242
• Main Authors: Han, Yingzhu, Liang, Huimin, Jiang, Duanlin, Wei, Zitong, You, Jiaxin, Zhai, Guofu
• Format: Conference Proceeding
• Language: English
• Published: IEEE 10.05.2024
• Subjects: Accuracy; Data models; Electromagnetics; Gray Wolf Optimization algorithm (GWO); isolation - based anomaly detection using Nearest - Neighbor Ensembles (iNNE); Kernel; Prediction algorithms; suction prediction; Support vector machines; Support Vector Regression (SVR); Vectors
• DOI: 10.1109/CIEEC60922.2024.10583733

Support vector regression (SVR), a machine learning method based on statistical learning theory, is an effective tool for sample training and data prediction. However, limited by its own principles and mechanisms, SVR often needs to adjust internal parameters when facing outliers and special experimental data in the data, and its accuracy and stability will be reduced. The model optimization of SVR by subjecting the data to outlier detection can improve the model performance. On this basis, this paper proposes a Grey Wolf Optimization (GWO) Support Vector Regression (SVR) algorithm based on isolation - based anomaly detection using Nearest - Neighbor Ensembles (iNNE). Firstly, outlier detection is carried out using the iNNE method, where the raw data are processed and outliers are removed; secondly, an improved kernel function is proposed to better map the input vectors from the low-dimensional space to the high-dimensional feature space; thirdly, parameter optimization of the SVR model is carried out by using the GWO algorithm to obtain the optimal parameters, and lastly, prediction is carried out on the electromagnetic relay suction dataset as an example, and is achieved by comparing the results with those of the un-optimized SVR, the traditional outlier detection method and the SVR before improving the kernel function are tested by comparing with the unoptimized SVR, the traditional outlier detection method and the SVR before improving the kernel function. The results show that the proposed method can effectively improve the computation speed and prediction accuracy.