Machine learning-based PV power forecasting methods for electrical grid management and energy trading

Abstract In this chapter, we show how it is possible to predict the day-ahead photovoltaic (PV) generation using machine learning (ML) techniques both at the level of an individual photovoltaic system and at the national level for the Italian case study. We address some critical issues related to the implementation of ML forecasting models/methods. We show two possible applications of the day-ahead PV generation forecast for PV producers/traders and for the Italian Transmission System Operator (TSO). The first deals with the day-ahead scheduling of the PV generation of utility-scale PV farms that PV producers/traders should deliver to the Italian TSO. The second is related to the imbalance between supply and demand at national level, hence to the prediction of the Italian net load made by the Italian TSO to plan the reserves needed to resolve the day-ahead production uncertainty. For both cases, we provide the economic value of solar forecasting under the current imbalance regulatory framework.