Economic Value of Research to Improve Solar Power Forecasting

The goal of the DOE-funded project “A Public-Private-Academic Partnership to Advance Solar Power Forecasting” was to build a solar power forecasting system that advances the state-of-the-science through cutting edge research and to test it in several high penetration solar utilities and ISOs with geographic and climatological diversity, using appropriate metrics, and disseminate the research results widely to raise the bar on solar power forecasting technology (Haupt et al. 2016). 

An integral part of the project was to develop and assess metrics for forecast improvement as well as examine the economic value of improved forecasts. In this paper we present analysis of the production cost modeling (PCM) undertaken by Xcel Energy for the Public Service Company of Colorado (PSCo) to derive estimates of the value of reductions in solar power forecast errors and scaling of model results to a national level. The objectives of this socio-economic analysis are (1) to develop initial estimate of the value of improved forecasts and (2) assess approaches to measuring forecast improvements (e.g., metrics) that are meaningful to both the weather forecasters and the solar energy and utility end-users. We thus explicitly tied benefit measures to metrics of forecast quality (e.g., “user-relevant verification”). 

One dimension of the use of solar power forecasts is in day-ahead (DA) decision making with respect to unit commitment. Working with project stakeholder partners, it was determined that production cost modeling (PCM) for use in DA decision-making was a viable approach to evaluate economic benefits of improved solar forecasts in the context of the project. Leveraging regulatory analysis by PSCo on potential significant increases in solar generation, scenarios were implemented for different levels of solar forecast error using their proprietary PCM. This analysis assessed differences in total production costs with current solar forecast error, with perfect forecast, and two forecast improvement scenarios. With a 50% reduction in forecast error (approximately equal to forecast error reductions achieved during the research effort) the error cost falls by $819,200 savings or 69% reduction in error costs. This averages into a production cost savings of $2.82 per MWh reduction in error. Regression analysis was employed to assess the relationship between production costs and forecast errors – including other variables from the data set that may also influence production costs to better understand the specific impact of forecast error on costs. 

Using EIA projections of solar power generation and per MWh savings for forecast error reduction from the baseline analysis, we generate order-of-magnitude estimates of the national value of improved (50% error reduction) solar power forecasts. Using a 3% rate of discount, and a 26-year analysis timeline starting in 2015, we develop an estimate of the present value of benefits from the forecast improvements of $455M in production cost savings.

References:

Haupt, S. E., and Coauthors, 2016: The SunCast Solar Power Forecasting System: The Result of the Public-Private-Academic Partnership to Advance Solar Power Forecasting. NCAR Technical Note NCAR/TN-526+STR, 307 pp, doi:10.5065/D6N58JR2. (available at: http://dx.doi.org/10.5065/D6N58JR2)

Annual Energy Outlook 2015: with projections to 2024. U.S. Energy Information Administration. DOE/EIA-0383(2015).