Transfer learning improves solar power forecasts with less data
Transfer Learning with Conformalized Quantile Regression for Solar PV Forecasting Under Load-Shedding-Driven Data Scarcity
Machine Learning
Summary
Predicting solar power output is hard where there isn’t much past data, like in places affected by load shedding. The authors used data from Australia to train a solar power prediction model, then adapted it to new locations with limited data, like Bangladesh. This approach not only made predictions more accurate but also gave better estimates of how confident the predictions were. Their method works well even when only one to three months of new data are available.
What this means in practice
- •For renewable energy operators: Improve solar PV power forecasts in regions with limited historical data by adapting models trained elsewhere.
- •For energy grid managers: Use better-calibrated uncertainty estimates for solar power to plan grid operations under data scarcity conditions.
Authors
Rakib Abdullah, K. M. Tahlil Mahfuz Faruk
Abstract
Solar photovoltaic (PV) forecasting in regions affected by load shedding is challenging because reliable historical observations are scarce. This study proposes a transfer learning framework combined with Conformalized Quantile Regression (CQR) to improve PV power forecasting and provide reliable uncertainty estimates under severe data scarcity. A source-domain PV dataset from Alice Springs, Australia, is used to pretrain a temporal forecasting model, which is then adapted to simulated Bangladesh PV data representing different levels of historical availability. Experimental results show that transfer learning reduces RMSE by up to 23.7% when only one month of target-domain data is available and by 13.7% with three months of data. The proposed Transfer Learning plus CQR framework achieves 94.3% empirical coverage with three months of target data while producing prediction intervals that are 14% narrower than those obtained without transfer learning. These results demonstrate that combining transfer learning with conformal uncertainty quantification can improve both point forecasting accuracy and uncertainty reliability when target-domain PV data are severely limited.