Electricity demand planning forecasts should consider climate non-stationarity to maintain reserve margins during heat waves. (15th November 2017)

• Record Type: Journal Article
• Authors: Burillo, Daniel; Chester, Mikhail V.; Ruddell, Benjamin; Johnson, Nathan
• Is Part Of: Applied energy. Volume 206(2017)
• Page Start: 267
• Page End: 277
• Publication Date: 2017-11-15
• Subjects: Climate non-stationarity -- Electric power infrastructure -- Peak demand -- Risk management -- Structural equation modeling; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2017.08.141
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: Weather-adjustment methods in electricity infrastructure planning are biased. Climate models project temperatures up to 58 °C (136 °F) in the US Desert Southwest. Peak demand does not increase linearly with temperature; s-curve more accurate. Los Angeles could experience hazardous power shortages in record-breaking heat. Risk management strategies identified via reductions in peak load & load variance. Abstract: Climate non-stationarity is a challenge for electric power infrastructure reliability; recordbreaking heat waves significantly affect peak demand[1], lower contingency capacities, and expose cities to risk of blackouts due to component failures and security threats. The United States' electric grid operates safely for a wide range of load, weather, and power quality conditions. Projected increases in ambient air temperatures could, however, create operating conditions that place the grid outside the boundaries of current reliability tolerances. Advancements in long-term forecasting, including projections of rising air temperatures and more severe heat waves, present opportunities to advance risk management methods for long-term infrastructure planning. This is particularly evident in the US Southwest—a relatively hot region expected to experience significant temperature increases affecting electric loads, generation, and delivery systems. Generation capacity is typically built to meet the 90th percentile ( T90 ) hottest peak demand, plus an additional reserve margin of least 15%, but that may not be sufficient to ensure reliable power services if air temperatures are higher than expected. The problem with this T90 planning approach is that it requires a stationary climate to be completely effective. In reality, annual temperature differences can have more than a 15% effect on system performance. Current long-term infrastructure planning and risk management processes are biased climate data choices that can significantly underestimate peak demand, overestimate generation capacity, and result in major power outages during heat waves. This study used downscaled global climate models (GCMs) to evaluate the effects of non-stationarity on air temperature forecasts, and a new high-level statistical approach was developed to consider the subsequent effects on peak demand, power generation, and local reserve margins (LRMs) compared to previous forecasting methods. Air temperature projections in IPCC RCPs 4.5 and 8.5 are that increases up to 6 °C are possible by the end of century, with highs of 58 °C and 56 °C in Phoenix, Arizona and Los Angeles, California respectively. In the hottest scenarios, we estimated that LRMs for the two metro regions would be on average 30% less than at respective T90 s, which in the case of Los Angeles (a net importer) would require 5 GW of additional power to meet electrical demand. We calculated these values by creating a structural equation model (SEM) for peak demand based on the physics of common AC units; physics-based models are necessary to predict demand under unprecedented conditions for which historical data do not exist. The SEM forecasts for peak demand were close to straight-line regression methods as in prior literature from 25–40 °C (104 °F), but diverged lower at higher temperatures. Power plant generation capacity derating factors were also modeled based on the electrical and thermal performance characteristics of different technologies. Lastly, we discussed several strategic options to reduce the risk of LRM shortages; including implementing technology, market incentives, and urban forms that reduce peak load and load variance per capita as well as their tradeoffs with several other stakeholder objectives.