An Improved Deep Neural Netowrk for Operation and Maintenance Prediction of Ultra High Voltage UHV Substation
Download as PDF
DOI: 10.25236/icceme.2022.032
Author(s)
Zhenyu Guo, Jinrui Gan, Xin Liu, Haibin Zhang, Long Chen, Jinjin Ding, Bin Zhang
Corresponding Author
Haibin Zhang
Abstract
The maintenance and operation costs of UHV substations are affected by many complex factors, and the maintenance cost data records are ambiguous and volatile. To solve the problem of unclear maintenance cost record is unknown, firstly, UHV substation maintenance entries are divided and processed using horizontal and vertical data analysis methods, and then deep neural network is used to predict maintenance costs. The deep neural network is then used to predict the overhaul costs. To improve the prediction accuracy of deep neural network, K-fold cross-validation is used to precisely adjust the original data training model, and genetic algorithm is applied to predict the deep neural network. Genetic algorithm is used to adjust and improve the initial values and threshold values of the deep neural network, to establish an improved deep neural network based on genetic algorithm for maintenance and operation cost prediction method based on genetic algorithm. The proposed method can effectively improve the accuracy of the model prediction. The comparison analysis shows that the proposed method can effectively improve the accuracy of the model prediction, and thus provide reference value for the grid to allocate maintenance costs to UHV substations.
Keywords
UHV substation maintenance cost prediction; deep neural network; genetic algorithm; K-fold cross-validation