Research On Location Selection,sizing And Optimization Of Smart Substation Based On GIS System And Load Forecasting Model

Substation is an important part of the power system.The safe and stable operation of the substation plays a vital role in the national development,people’s life and economic development.With the continuous growth of load and the continuous development of renewable energy,the location,sizing and intelligence of substation become more and more important in distribution network planning.In the short and long term,the future power system has a high degree of uncertainty,and various factors including technology,cost and macroeconomic situation will affect the development of the power system in the coming decades.In this paper,the following two models are proposed for the problem of substation location,sizing and improving the flexibility of distribution network based on shortterm and long-term load forecasting.Firstly,in order to overcome the subjectivity and limitations of traditional substation location methods and achieve efficiently and visualized substation location,this paper proposes an improved method for substation location and scale issues based on geographic information and semi-supervised learning.This method can use the minimum investment and annual operating cost for substation site selection,and classify and optimize the site selection results according to the capacity and power supply range.Simulation experiments show that this method can optimize the location,capacity and power supply range of substation with minimal investment.Secondly,this paper proposes a smart substation optimization model based on transfer learning and short-term load forecasting to support the deployment of smart grids.Load prediction through transfer learning enables normal operation in the distribution network even for unmonitored secondary substations,that is,when overload is expected,corrective measures can be taken to transfer the load and change the power grid topology,so as to improve the reliability and recovery ability of the power system and the intelligence of the substation.The experiments compared the prediction results of the existing machine learning-based model,the baseline model and the model proposed in this paper.According to the comparison between the prediction results and the real data error,it can be concluded that the error of the method proposed in this paper is smaller and the prediction performance is improved compared with the method proposed before.