Computer Science and Information Systems 2024 Volume 21, Issue 2, Pages: 437-452https://doi.org/10.2298/CSIS220509050M Full text (
Automatic voltage stabilization system for substation using deep learning
Moon Jiyong (Department of Business Administration, Dongguk University, Jung-gu, Seoul, Korea), asdwldyd@dongguk.eduSon Minyeong (Department of Medical Biotechnology, Dongguk University, Jung-gu, Seoul, Korea), smya@dongguk.eduOh Byeongchan (Department of Statistics, Dongguk University, Jung-gu, Seoul, Korea), oxox@dongguk.eduJin Jeongpil (Department of Industrial System Engineering, Dongguk University, Jung-gu, Seoul, Korea), chin@dongguk.eduShin Younsoon (Department of Computer Science, Dongguk University, Jung-gu, Seoul, Korea), ysshin@dongguk.edu
The operating voltage in the substation must be maintained at its rated voltage within the specified standard because a voltage outside the specified range may cause a malfunction of the power facility and interfere with the stable power supply. Therefore, the voltage regulation process to maintain the rated voltage of the substation is essential for the stability of the power system. However, the voltage regulation process is currently performed manually by resident staff. Voltage regulation based on human judgment increases the uncertainty of voltage stabilization and makes efficient operation in consideration of the economic feasibility of power facilities difficult. Therefore, this paper proposes an automatic voltage stabilization system that can automatically perform voltage regulation. Instead of predicting the electrical load or overvoltage conditions studied so far, we focus on more direct, scalable input capacity prediction for an automatic voltage stabilization system. First, the proposed system predicts the input capacity required for a given situation through a trained stacked LSTM model. Second, an optimal regulation plan is derived through an optimization process that considers the economic feasibility of power facility operation. Additionally, the development of the user interface makes it possible to visualize the operation of algorithms and effectively communicate the models’ predictions to the user. Experimental results based on real substation data show that the proposed system can effectively automate the voltage regulation process.
Keywords: automatic voltage stabilization system, energy system, input capacity prediction, deep learning, optimal regulation plan
Show references
Albawi, S., Mohammed, T.A., Al-Zawi, S.: Understanding of a convolutional neural network. In: 2017 International Conference on Engineering and Technology (ICET). pp. 1-6 (2017)