基于双尺度时序卷积网络和残差Bootstrap方法的电流互感器测量误差区间预测OA

To address the problems that the ratio-error series of current-transformers,affected by environmental disturbance,operating condition variations and random noise,exhibit nonstationary and multi-scale characteristics,and it is difficult to balance point prediction and interval estimation,this paper proposes an interval forecasting method integrating complete ensemble empirical mode decomposition(CEEMD),dual-scale temporal convolutional network(DTCN)and bias-corrected residual bootstrap(BCRB).Firstly,The CEEMD is adopted to decompose and reconstruct the original ratio-error series to suppress high-frequency noise and mode mixing.Secondly,the DTCN is used to extract short-term fluctuation and long-term drift features to realize point prediction of ratio errors.Finally,prediction intervals under different confidence levels are constructed based on the residual bootstrap method.Case studies show that the proposed method achieves a high interval coverage probability at the confidence level of 0.99.