PSO-optimized VMD-LSTM-based Prediction of Deviation Trend of DC Measurement Device
Electronic direct current measurement devices represented by optical current transformers(OCTs),have found extensive applications in direct-current transmission engineering.However,with increasing operating time,there may be a risk of deviation violation,necessitating the prediction of measurement deviations.Based on the decomposition-optimiza-tion-prediction-reconstruction approach,a hybrid prediction model of particle swarm optimized(PSO)variational mode decomposition(VMD)-long short-term memory(LSTM)was proposed in this work for multi-step prediction of deviation trends.First the historical deviation sequence was decomposed into multiple sub-sequences using the VMD and then the hyperparameters of the prediction models for each sub-sequence were optimized by PSO.Subsequently the prediction of sub-sequences was carried out through LSTM and the predicted values were aggregated to obtain the ultimate prediction result.The prediction of measurement deviations for an OCT in a converter station was conducted,and the results indica-ted the superior prediction accuracy of the proposed model compared to several selected prediction models.
万方数据
维普
