Deep Learning-Based Hurricane Resilient Co-planning of Transmission Lines, Battery Energy Storages and Wind Farms

TL;DR

This paper presents a multi-stage, deep learning-enhanced co-planning model for resilient transmission infrastructure that incorporates hurricane risk assessment, renewable integration, and advanced optimization techniques.

Contribution

It introduces a novel deep learning-based load forecasting method combined with hurricane resilience modeling for transmission expansion planning.

Findings

The model effectively reduces hurricane impact on transmission systems.

Deep learning improves load prediction accuracy.

The approach enhances system resilience and renewable integration.

Abstract

In this paper, a multi-stage model for expansion co-planning of transmission lines, Battery Energy Storages (BESs), and Wind Farms (WFs) is presented considering resilience against extreme weather events. In addition to High Voltage Alternating Current (HVAC) lines, Multi-Terminal Voltage Source Converter (MTVSC) based High Voltage Direct Current (HVDC) lines are planned to reduce the impact of high-risk events. To evaluate the system resilience against hurricanes, probable hurricane speed (HS) scenarios are generated using Monte Carlo Simulation (MCS). The Fragility Curve (FC) concept is utilized for calculating the failure probability of lines due to extreme hurricanes. Based on each hurricane damage, the probable scenarios are incorporated in the proposed model. Renewable Portfolio Standard (RPS) policy is modeled to integrate high penetration of WFs. To deal with the wind power and load demand uncertainties, a Chronological Time-Period Clustering (CTPC) algorithm is introduced for extracting representative hours in each planning stage. A deep learning approach based on Bi-directional Long Short-Term Memory (B-LSTM) networks is presented to forecast the yearly peak loads. The Mixed-Integer Linear Programming (MILP) formulation of the proposed model is solved using a Benders Decomposition (BD) algorithm. A modified IEEE RTS test system is used to evaluate the proposed model effectiveness.