10 likes | 99 Views
Learn how to integrate wind energy into portfolios using load sculpting controllers and predictors, alongside implicit and explicit storage solutions to manage variability. Explore the challenges, strategies, and future work in renewable energy integration.
E N D
Effective supply-following/load-sculpting power power wind Storage ≡ Supply prediction pgrid p′grid ∫ = E ∫ = E time time Implicit storage (thermal mass, deferrable work) Explicit storage (better batteries) ` Towards Supply-following Loads: Lessons from Wind Prediction Mike He, Achintya Madduri, Yanpei Chen, Rean Griffith, Ken Lutz, Seth Sanders, Randy Katz UC Berkeley Overview Prediction horizon matters: Use metrics beyond RMSE and MAE • Integrating renewables into energy portfolios is challenging • High variability • High misprediction rate even with a “good” model • The good news • It’s ok to be wrong • The prediction horizon matters (short vs. longer term) • We can deal with the variability via load-sculpting Figure 1: Wind power trace from a large power plant in the Midwest Goal: Integrating renewables into energy portfolios Load sculpting controllers: Combining predictors & implicit storage • Managing HVAC heating/cooling cycles using optimal control • Use the home thermal model from jronsim (Java Residential Occupied Neighborhood Simulator) • Lookahead controller within 5% of an oracle using a perfect wind predictor Perfect prediction is not a silver bullet. We need storage! Conclusions and future work • Integrating renewables like wind into energy portfolios is challenging • High variability • To evaluate predictors we need to look at additional quality metrics beyond RMSE and MAE • Prediction horizon and prediction error distribution matters • We can deal with the variability using simple predictors and implicit storage Acknowledgments • We would like to thank: • the National Renewable Energy Laboratory (NREL) for providing wind data • Albert Goto (UC Berkeley) for providing our compute infrastructure