PNNL

Abstract

The U.S. Energy Information Administration conducts an annual survey (form EIA-923) to collect, among other variables, annual and monthly net generation for more than ten thousand U.S. power plants. Only ~10% of the ~1,500 hydroelectric plants included in this regular data release come with observations of monthly net generation; the remaining 90% of plants are represented with monthly net generation that is imputed using a statistical method. The imputation method neglects local hydrology and reservoir operations, rendering the monthly generation totals unsuitable for several research applications. Here we present an alternative approach to disaggregate each plant’s reported annual generation using historical reservoir releases (~180 plants) or, for the majority of cases where reservoir releases are unavailable, river discharge recorded at a USGS gage downstream of each dam (~1180 plants). We evaluate the approach using 130 plants for which monthly generation observations are available, showing median Kling Gupta Efficiency (KGE) of 0.74, 90% CI [-0.07, 0.93] across plants with observed reservoir release, and KGE of 0.51 [-0.28, 0.79] for cases imputed using a downstream flow gage. The new dataset—named RectifHyd—provides a timely alternative to EIA-923 for U.S. scale, plant-level, hydropower net generation for analyzing within-year hydropower generation behavior and constraining sub-annual hydropower availability in power grid operations and expansion models.