Researchers from North Carolina State University demonstrate that combining multiple machine learning models can improve day-ahead solar forecasting by up to 13%. Led by Yen-Hsi Chou and Anderson De Queiroz, the team tests seven models using California utility data from 2019 to 2022. They find that weighted averaging improves forecasts by 11% for the Imperial Irrigation District, while a multi-input approach boosts accuracy by 13% for the Los Angeles Department of Water and Power, aiding grid stability.
Day-ahead solar forecasting accuracy
- ▪The multi-input ensemble approach yielded day-ahead solar forecasting improvements of up to 13% for the Los Angeles Department of Water and Power.
- ▪The weighted averaging ensemble approach provided day-ahead solar forecasting improvements of up to 11% for the Imperial Irrigation District.
- ▪Researchers from North Carolina State University demonstrated ensemble techniques that improve day-ahead solar forecasts by up to 13% over the baseline model.
Machine learning algorithms
- ▪The two ensemble approaches tested by the researchers were weighted averaging and a multi-input approach.
- ▪The researchers tested seven forecasting models using weather and power data from 2019 through 2022 from two California utilities.
- ▪North Carolina State University researchers, including Yen-Hsi Chou and Anderson De Queiroz, evaluated statistical models and artificial neural networks, selecting the BiLSTM model as their baseline.
Satellite data integration
- ▪Recent advancements in day-ahead solar forecasting incorporate satellite data, machine learning algorithms, and historical weather patterns to estimate solar generation.
- ▪Incorporating satellite data and advanced algorithms allows energy companies to predict solar energy production up to a day in advance with greater accuracy.
Grid management optimization
- ▪Improved forecasting accuracy allows energy companies to optimize energy trading and reduce reliance on backup power sources.
- ▪Accurate day-ahead solar forecasting helps grid operators balance supply and demand, optimize energy production, and ensure grid stability.
- ▪More accurate solar forecasting helps energy companies minimize waste and save costs by avoiding the overproduction or underproduction of energy.
Renewable energy integration
- ▪Reliable predictions of solar energy generation help energy companies reduce the curtailment of renewable energy and support solar installation growth.
- ▪Reducing uncertainty in solar energy generation supports increased renewable energy penetration and reduces carbon emissions.
Solar intermittency challenges
- ▪Increasing solar penetration makes forecasting supply and demand more challenging for energy planners and grid operators due to inconsistent sunlight availability.
- ▪The intermittent nature of solar power, which depends on weather conditions and the time of day, makes predicting daily energy generation challenging.
Story comments
Loading comments…