The PRESTO model generates random power interruptions that reflect seasonal and diurnal patterns in historic power interruptions at the county-level, most of which are of relatively short duration (minutes to hours). The user selects the county of interest and the model loads expected number of annual events by month and average event duration by month for this location, which the user can override if desired. The model then randomly generates outage events over a specified number of simulation “test” years (ranging from 1,000 to 20,000 years).
The PRESTO model is trained on hourly, county-level power interruption data obtained from PowerOutage.US for the years 2017-2021. Based on that data, LBNL developed probabilistic functions to predict the likelihood that an individual customer in a given county would experience an interruption during any given hour of the year, and the duration of that interruption. For further details and a description of key limitations, please refer to the model documentation .
The main output is a set of simulations, each one with several events and durations that represent plausible power interruptions over a year. The user can download a spreadsheet with the timing and duration of all randomly generated interruption events, in addition to reviewing charts and tables that summarize the simulated interruptions. LBNL has also made the PRESTO model available via an application programming interface (API), in order to allow users to perform batch simulations or to integrate PRESTO into a larger simulation workflow. Instructions to access the API can be found here.
The PRESTO model team gratefully acknowledges funding from the Solar Energy Technology Office (SETO) of the Energy Efficiency and Renewable Energy office of the U.S. Department of Energy under Lawrence Berkeley National Laboratory Contract No. DE-AC02-05CH11231