Multidimensional Time-Series Life Cycle Costs Analysis of Intelligent Substation

Abstract

Comprehensive and accurate LCC (Life Cycle Costs) of intelligent substations are conducive to improving operation efficiency and reliability of power transmission. To reduce maintenance costs and enhance assets utilization, multiple dimensional time series LCC analysis of intelligent substation based on PCA-LSTM (Principal Component Analysis Long-Term Short-Term Memory) is proposed in this paper. This paper first discusses the differences between smart substations and conventional substations in LCC analysis. Because it has an advantage over RNN in handling gradient vanishing and explosion, LSTM (Long-Term Short-Term) is used instead of RNN (Recurrent Neural Network). Considering that LCC is a comprehensive study which contains large number of variables, PCA (Principal component analysis) is utilized to deal with the data dimensionality reduction preprocessing. Results show that PCA-LSTM performs better in RMSE (Root-Mean-Square Error), MAE (Mean Absolute Error), MAPE (Mean Absolute Percentage Error), as 0.182, 0.154 and 15.84%, respectively. In our case, the computation time is about half that of the pure LSTM. And the calculation efficiency is 2.2 times compared to the original.