Fine-Grained Photovoltaic Output Prediction Using a Bayesian Ensemble

Local and distributed power generation is increasingly reliant on renewable power sources, e.g., solar (photovoltaic or PV) and wind energy. The integration of such sources into the power grid is challenging, however, due to their variable and intermittent energy output. To effectively use them on a large scale, it is essential to be able to predict power generation at a finegrained level. We describe a novel Bayesian ensemble methodology involving three diverse predictors. Each predictor estimates mixing coefficients for integrating PV generation output profiles but captures fundamentally different characteristics. Two of them employ classical parameterized (naive Bayes) and non-parametric (nearest neighbor) methods to model the relationship between weather forecasts and PV output. The third predictor captures the sequentiality implicit in PV generation and uses motifs mined from historical data to estimate the most likely mixture weights using a stream prediction methodology. We demonstrate the success and superiority of our methods on real PV data from two locations that exhibit diverse weather conditions. Predictions from our model can be harnessed to optimize scheduling of delay tolerant workloads, e.g., in a data center. Introduction Increasingly, local and distributed power generation e.g., through solar (photovoltaic or PV), wind, fuel cells, etc., is gaining traction. In fact, integration of distributed, renewable power sources into the power grid is an important goal of the smart grid effort. There are several benefits of deploying renewables, e.g., decreased reliance (and thus, demand) on the public electric grid, reduction in carbon emissions, and significantly lower transmission and distribution losses. Finally, there are emerging government mandates on increasing the proportion of energy coming from renewables, e.g., the Senate Bill X1-2 in California, which requires that one-third of the state’s electricity come from renewable sources by 2020. However, renewable power sources such as photovoltaic (PV) arrays and wind are both variable and intermittent in their energy output, which makes integration with the power grid challenging. PV output is affected by temporal factors Copyright c © 2012, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. such as the time of day and day of the year, and environmental factors such as cloud cover, temperature, and air pollution. To effectively use such sources at a large scale, it is essential to be able to predict power generation. As an example, a fine-grained PV prediction model can help improve workload management in data centers. In particular, a data center’s workloads may be “shaped” so as to closely match the expected generation profile, thereby maximizing the use of locally generated electricity. In this paper, we propose a Bayesian ensemble of three heterogeneous models for fine-grained prediction of PV output. Our contributions are: 1. The use of multiple diverse predictors to address finegrained PV prediction; while two of the predictors employ classical parametrized (naive Bayes) and non-parametric (nearest neighbor) methods, we demonstrate the use of a novel predictor based on motif mining from discretized PV profiles. 2. To accommodate variations in weather profiles, a systematic approach to weight profiles using a Bayesian ensemble; thus accommodating both local and global characteristics in PV prediction. 3. Demonstration of our approach on real data from two locations, and exploring its application to data center workload scheduling.