Construction of a Digital Twin Model for Loss Metering in UHVDC Transmission Systems Based on Deep Learning

Abstract

Ultra-high voltage direct current (UHVDC) transmission systems have multiple inputs and multiple outputs, nonlinearity and strong coupling, making it challenging to accurately measure component and system losses using traditional methods based on system mechanisms and simplified modeling of parameters. In this paper, a digital twin modeling method based on deep learning networks is proposed to solve the above problem. The proposed method extracts sequence features of loss data using LSTM based on improved Res-LSTM network modeling, fuses all features effectively by combining residual structure with convolution kernel, and establishes a digital twin model of the UHVDC transmission system to realize strong coupling and nonlinear mapping from multiple inputs to multiple outputs. The experimental results show that the Res-LSTM model is 37.57% lower than the LSTM model for MSE in the total loss calculation mode, and 11.8% lower than the LSTM model for MSE in the sub-loss sequence calculation mode. It is verified that the Res-LSTM model has high accuracy in constructing the digital twin model to measure the total loss of the UHVDC transmission system. The proposed digital twin model can help analyze and provide some useful suggestions for loss reduction of the UHVDC transmission system.

Keywords

• UHVDC
• loss measurement
• digital twin
• deep learning
• Res-LSTM