Value of more accurate urban weather prediction in the day-ahead energy market

Abstract

Urban heat significantly influences human comfort and leads to additional energy use for space-cooling in built environments. The building sector has a high potential for cost savings and emission reduction by improving decision making in energy markets, and space-cooling of residential and commercial buildings contributed nearly 10 percent of total U.S. electricity consumption in 2021. System operators must make decisions, e.g., purchasing energy in the day-ahead market, under the uncertainty associated with the relationship between ambient temperature and cooling-driven electricity demand. Customers buy energy in the day-ahead market at lower and locked prices, based on their forecasts of the energy use of the next day. As an alternative, the real-time market provides electricity at changing prices, which are higher than day-ahead prices on average. Thus, to minimize costs, system operators must decide how much energy they purchase in advance, in the day-ahead market, knowing that they can access the real-time market for additional needs on the following day. If operators can better predict future weather, they can improve the accuracy of the energy use forecasts, and so make better decisions.

In this paper, we assess the impact of the urban weather modeling on energy cost, using value of information (VoI) analysis. To do that, we couple two models: (a) a probabilistic spatio-temporal model for temperature forecast and (b) a building energy use model for electric load forecast given temperature scenarios. By coupling these two models, we can quantify the propagating uncertainty from the temperature model to load forecasts and assess the impact of a better urban weather modeling on the energy use forecasts and the associated decision-makings. A numerical example indicates that, compared to the case of having no temperature update and depending on the building type, the improved temperature forecasts allow saving 16% to 34% of the additional cost due to uncertainty (i.e., of the difference between the expected energy cost under uncertainty and the cost under perfect prediction on future energy use).