Power from proposed photovoltaic power systems is commonly estimated by a performance model using hourly averaged weather data, such as TMY data. Use of hourly averaged weather data introduces error in model estimates independent from other sources of error. We isolate and quantify the error in DC energy that results solely from using time-averaged model inputs. We demonstrate that error in estimated energy arises from two separate approximations: 1) the approximation of PV performance as linear in time-averaged inputs, such as irradiance; and 2) the treatment of time-averaged inputs for partially-lit hours including sunrise and sunset. We show that the net error in DC energy from these approximations can be predicted from characteristics of the PV modules and the frequency of clear-sky conditions. For a typical cSi PV module, error in annual energy ranges between -0.3% for locations with primarily clear-sky conditions to +2.0% for locations with highly variable conditions; errors are greater for systems with amorphous silicon modules and less for systems with CdTe modules. Our analysis permitted quantifying the error, identifying the underlying causes, and proposing a model to estimate the error in annual energy resulting from time-averaged weather data. With current modeling practices, the error in annual energy resulting from time averaging is substantially reduced by use of weather data at 15 minute intervals or less.
展开▼
