Climate Informed Transmission Expansion Planning Using Predictive Analytics

Abstract

Transmission expansion planning (TEP) traditionally optimizes investments in new lines, reconductoring, or grid-enhancing technologies to meet future demand and renewable integration goals under relatively stable historical assumptions. However, climate change introduces non-stationary uncertainties through rising ambient temperatures, more frequent and intense heatwaves, altered wind and solar patterns, and compound extreme events that derate generation and transmission capacities while spiking cooling-driven demand. Predictive analytics leveraging machine learning, high-resolution climate projections (e.g., Thermodynamic Global Warming or TGW datasets), historical weather reanalysis, SCADA/PMU data, and outage records enable climate-informed TEP by generating accurate forecasts of temperature-dependent parameters such as dynamic line ratings (DLR), load profiles, renewable availability, and outage risks. This research paper develops a comprehensive framework that integrates predictive analytics into multi-stage, stochastic or robust TEP models. It employs predictive models (LSTM, graph neural networks, or hybrid physics-informed networks) for multi-horizon DLR forecasting, load under heat stress, and extreme event probabilities, feeding these outputs into capacity expansion optimization that co-optimizes transmission, generation, and storage under climate ensembles. Formulations use mixed-integer linear programming (MILP) with data-driven uncertainty sets or scenario reduction, explicitly incorporating conductor heat balance equations for ambient-adjusted or dynamic ratings.