Big data with temporal dependence brings unique challenges in effective prediction and data analysis. The complex high-dimensional interactions between observations in such data brings unique challenges which standard off-the-shelf machine learning algorithms cannot handle. Even basic tasks of clustering, visualization and identification of recurring patterns are difficult.
The Quadracci Sustainable Engineering Lab (qSEL) has several research efforts related to the low-carbon energy transition, including pathways to decarbonize building space heating. Recent work has produced large model data sets that have supported recent journal articles. Several maps have been produced using QGIS and data has been made public, but user functionality is limited. While we continue to build on these efforts, we also want to make our results and data available more widely for other researchers and policymakers. The large data sets (10 years of hourly data for more than 72,000 census tracts and six scenarios) and different spatial aggregations (e.g. states and electricity planning/operating regions) present challenges. In this project, the DSI Scholar would first work with qSEL researchers to develop an interactive web interface to display maps of relevant analyses and allow users to produce time series data from the underlying models. Additional research would include further analysis at a regional level – likely New York State – to refine the current model based on additional intraregional and energy source data. The project has the possibility of extending through Summer 2020, subject to fundraising efforts and the success of the Spring 2020 project.
A major obstacle to the decarbonization of the electricity production systems is the multi scale (space and time) variability of wind, solar and hydro energy sources. Much work is being done to understand the high frequency variations in these sources from the perspective of grid integration. However, as with rainfall and other natural systems, these variables can exhibit log-log fractal scaling in space and time, such that the variance of the process increases with temporal duration and with spatial scale. Focusing on high frequency variations thus grossly understates the systemic risk that is associated with these sources. Appropriate national grid design including electricity storage allocation, needs to consider both the periodic annual cycle variations and quasi-periodic inter-annual variability which have larger variance, and the phase lags in these variations across space. The proposed project would explore the development of a multi-level, hierarchical spatio-temporal model for wind or solar using data from the continental USA and its subregions to explore stochastic simulations and multi-scale predictions of the associated risk to inform system design and financial instruments development.
