Year: 2025
Type: conference-paper
Abstract: In order to solve the problem that the traditional life prediction methods are difficult to accurately capture the complex characteristics of the life of electronic components and poor prediction accuracy, the historical capacity data of a certain type of industrial capacitor under high temperature conditions were obtained through industrial capacitor aging test system, the Three-sigma (3σ) Rule combined with moving sliding window technology were used to clean the data, and the parameters of Variational Mode Decomposition (VMD) were determined by Particle Swarm Optimization (PSO) algorithm, decomposing the tolerance sequence into Intrinsic Mode functions (IMF) on different time scales as inputs to the Long Short Term Memory (LSTM) neural network, constructing the degradation prediction model of industrial capacitors based on PSO-VMD-LSTM. Compared with the Autoregressive Integrated Moving Average model (ARIMA), Recurrent Neural Network (RNN) and LSTM model, the Mean Absolute Error (MAE), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) of the proposed optimization model were the smallest, the Coefficient of Determination (R2) was increased significantly, which proves that the PSO-VMD-LSTM model has high applicability and accuracy to predict the degradation of industrial capacitors.
Authors: Shenglei Wang, Lingyu Song, Junjun Wang, Wenqiang Qin, Xin Tian, Jianjun Zhang
Institution: Hunan Normal University
Language: English
Subfield: Safety, Risk, Reliability and Quality
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4412120069