Heterogeneous Vulnerability of Zero-Carbon Power Grids under Climate-Technological Changes

TL;DR

This study evaluates the reliability of a zero-carbon NYS grid under coupled climatic and technological changes using a high-resolution, 22-year, 300-scenario framework. It integrates weather-driven generation, electrified load, dynamic transmission ratings, HVDC upgrades, and storage within a DC-OPF+MCDA workflow to quantify vulnerabilities and the firm, zero-emission capacity (FZEC) needed. The results reveal strong spatiotemporal heterogeneity: winter vulnerabilities are congestion- and heating-driven, while summer vulnerabilities arise from wind drought and grid constraints, implying FZEC needs up to $61\%$ to $105\%$ higher than the CLCPA scoping-plan estimate (up to about $37$ GW when zonal constraints are considered). The findings motivate deploying long-duration storage and green hydrogen to alleviate congestion and co-variability, offering policy-relevant guidance for region-wide decarbonization and generalizable insights for other grids with similar topology.

Abstract

The transition to decarbonized energy systems has become a priority globally to mitigate carbon emissions and, therefore, climate change. However, the vulnerabilities of zero-carbon power grids under climatic and technological changes have not been thoroughly examined. In this study, we focus on modeling the zero-carbon grid using a dataset that captures diverse future climatic-technological scenarios, with New York State as a case study. By accurately representing the topology and operational constraints of the power grid, we identify spatiotemporal heterogeneity in vulnerabilities arising from the interplay of renewable resource availability, high load, and severe transmission line congestion. Our findings reveal a need for 61-105\% more firm, zero-emission capacity to ensure system reliability. Merely increasing wind and solar capacity is ineffective in improving reliability due to transmission congestion and spatiotemporal variations in vulnerabilities. This underscores the importance of considering spatiotemporal dynamics and operational constraints when making decisions regarding additional investments in renewable resources.

Figures (7)

• Figure 1: Representation of the NYS grid - The NYS grid has 11 load zones indexed from A-K. Zones A-E are upstate zones and F-K are downstate zones. In the NYS representation, there are 57 buses shown as black circles in the figure. The 94 transmission lines are denoted by gray lines with red highlights to denote the Total East interface that transfers power from upstate to downstate and is subject to frequent congestion. Dashed lines represent the new HVDC lines that are designed to alleviate the congestion shown in red. One line is from Hydro-Quebec to NYC (D $\rightarrow$ J) and the other is from upstate NY to NYC (E $\rightarrow$ J), and are scheduled to be online in 2026 and 2027, respectively. The average load for each load zone (after electrification over 22 years) is indicated by the background color, where NYC and Long Island have the majority of the load due to high population density.
• Figure 2: Overview of the Framework - (a) Power grid demand, supply, and transmission are influenced by weather variables to maintain joint co-variability. (b) Climatic and technological factors are sampled to capture a wide range of 300 alternative SOWs for the 22-year horizon. Time series data is aggregated for illustration, but high spatial resolution is preserved in the model (see Methods). (c) The 300 scenarios, each with 22 years of hourly spatiotemporal trajectories, are input to the DC-OPF model, which seeks to minimize the total expected load shedding for a year. Simulated system performance is evaluated by three metrics to identify potential vulnerabilities, summarized in Table \ref{['tab: metricdef']}.
• Figure 3: Aggregated Seasonal Load, Wind and Solar Patterns for 22 years - The simulated Load (upper), Wind (middle), and Solar (lower) for every year are shown in gray lines. The red and blue lines in each panel denote the maximum and minimum values for each hour over the 22 years, respectively. The load profiles have winter peaks after electrification, which is aligned with the expectation in NYISOphaseI2019NYtrends2022. Wind power availability tends to be low over summer while solar power has the highest availability in spring with a little reduction in summer as temperature increases (solar panel efficiency decreases as temperature increases).
• Figure 4: Overview of baseline system vulnerability - The figure highlights the spatial and temporal characteristics of NYS grid vulnerabilities over a 22-year horizon. The vulnerabilities are evaluated under three different metrics in panels (a)-(c), which shows the quarter month (a quarter month is approximately one week, but adjusted for the duration of each month) of a year on the horizontal axis, and the index over the 22-year horizon on the vertical axis. Panel (a) shows the total quantity of load shedding in a quarter month, panel (b) the maximum load shedding event that occurs within each quarter month, and panel (c) shows the total number of hours with load shedding in a quarter month. Pixel color in panels (a)-(c) indicates the severity of the vulnerabilities, ranging from minimal vulnerability (blue) and most severe vulnerability (red). Panels (d)-(f) consider the same vulnerability metrics for the summer season within a spatial context. The size of the marker (dot) indicates the average intensity of each evaluation metric, while the color of the lines connecting each load zone denotes the average likelihood of congestion for each transmission interface connecting two load zones, with darker colors indicating a higher congestion rate. Similarly, panels (g)-(i) highlight the average spatial vulnerabilities for the winter season.
• Figure 5: Ranking of the climatic-technological factors for summer and winter - The importance score for each deeply uncertain parameter is derived from the Gradient Boosted Tree for winter and summer. The summer vulnerability is evaluated in terms of load shedding quantity and the hours of load shedding. (Maximum load shedding is not included in the figure because there are only a few data points that exceed the threshold. Therefore, the dataset is strongly biased for the Gradient Boosted Tree to classify.) The temperature increase is the most important driver for both load shedding quantity and load shedding hours. Winter operations are evaluated by all three metrics, showing temperature increase and wind capacity multiplier as the most important drivers of vulnerability.