Talking Solar Panels: Using one Rooftop Array to Estimate the Performance of Another
Nicholas Engerer (NICTA)
NICTA SML SEMINARDATE: 2012-11-22
TIME: 11:15:00 - 12:00:00
LOCATION: NICTA - 7 London Circuit
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
In the effort to create a solar forecasting algorithm that can produce estimates of power production from rooftop photovoltaic (PV) systems in a given suburb or sub-suburb region, our main source of available data is the power output time series from a selection of these rooftop PV systems themselves. However, when we begin to analyse the available time series data, we observe that the available systems are all of different sizes (more or less panels in an array), orientations (pointed various directions so peak power output arrives at different times in the day) and have their own losses/inefficiencies (e.g. shading, low quality inverter equipment). However if we normalize these power output time series to their theoretical performance under clear skies, we can much more easily translate the performance of one system to another. This was done by taking a clear sky radiation model, using it as an input to a PV model, matching the characteristics of each array to the PV model's database and running a simulation to get a clear sky curve for that system. An overview of this process will be given. Moving forward, there exist opportunities to use machine learning to improve this approach, and a few specific ideas will be presented. Following this, a discussion on how to further improve these ideas and begin to create forecasting algorithms will be encouraged.
BIO:
