Optimal planning and forecasting of active distribution networks using a multi-stage deep learning based technique

Abstract

This paper presents a comprehensive methodology for long-term planning in distribution networks to address the challenges associated with integrating renewable energy resources (RERs) and battery energy storage systems (BESSs). The methodology incorporates active network management (ANM) schemes, coordinated voltage control (CVC), and demand side management (DSM) to optimize the sizing and siting of RERs and BESS. Deep learning-based long-term forecasting techniques are utilized to capture the hourly variations in load demand, solar irradiance, and wind speed throughout the year. The objective is to minimize overall network costs, considering investment and operational costs, while planning over a 5-year horizon with a 3% annual load growth. The model includes tie-lines as a decision variable for network expansion investments, ensuring efficient network flexibility. The linearized mathematical model is developed as a mixed integer linear programming (MILP) problem to enhance scalability. A case study is conducted on the UKGDS 16-bus system to verify the optimization model, and further validation is performed on the IEEE 69-bus system to assess scalability. The results demonstrate the effectiveness of the proposed method in achieving accurate forecasting and efficient sizing and siting decisions for RERs and BESSs, including tie-line investments, providing valuable insights for real-world distribution network scenarios.

Keywords

• Active network management
• Battery energy storage systems
• Deep learning forecasting
• Planning
• Sizing and siting optimization