Search papers, labs, and topics across Lattice.
This study addresses the challenge of reliable day-ahead photovoltaic (PV) forecasting at newly deployed sites with limited data by developing an environmental-AI pipeline that integrates measured inverter output with meteorological inputs. The pipeline employs advanced techniques such as timestamp correction, leakage-safe feature construction, and validation-learned stacking of complementary predictors to enhance forecasting accuracy. The results demonstrate a significant reduction in forecast errors, achieving a 32% improvement in daylight normalized RMSE compared to traditional methods, highlighting the effectiveness of physics-aware stacking in data-scarce environments.
Achieving a 32% reduction in PV forecast errors demonstrates that physics-aware stacking can significantly enhance predictions even with limited site data.
Reliable photovoltaic (PV) forecasts are needed for low-carbon energy systems, but newly deployed sites often have short, imperfect records. This makes standard day-ahead forecasting difficult: persistence and physical baselines can be sensitive to calibration and timestamp alignment, while single machine-learning models may capture only one structure in the data and overstate skill under non-temporal validation. We study this problem at a United Kingdom charging-station site, where PV forecast errors affect charging availability, storage scheduling, and downstream control. Using measured inverter output and publicly available meteorological inputs, we develop a deployment-oriented environmental-AI pipeline for day-ahead hourly PV forecasting. The pipeline corrects timestamp conventions, constructs leakage-safe solar-geometry and clearness-index features, adds short-term atmospheric context, and combines complementary predictors through validation-learned stacking. Against smart persistence, a clear-sky baseline that adjusts recent PV output using expected clear-sky irradiance, the best ensemble reduces daylight normalised RMSE by about 32% under random day-blocked evaluation and 9% under the stricter rolling-origin protocol. It also reduces daylight RMSE relative to the strongest individual machine-learning baseline by 6.6% and 6.4%, respectively. The results show that physics-aware stacking can support PV forecasts from limited site data, but its value depends on model class, evaluation protocol, and deployment context.
