This paper investigates forecasting and market commitment strategies for photovoltaic plants coupled with battery energy storage systems (flexible PV plants). A grey-box forecasting framework is developed to predict firm PV generation, residual variable output, and battery state of charge by combining machine learning based PV forecast with a deterministic power plant controller model. A set of key performance indicators is defined to quantify supply reliability and solar-induced system flexibility. These KPIs are used to compare unconstrained PV generation with different firm and near-firm market commitments. The forecasting framework is applied both to correct/optimize baseload commitment (firm strategy) and to predict residual variable components of flexible PV generation, enabling simultaneous firm and residual generation trading (generation splitting strategy). The methodology is tested using measurements from a 662 kWp PV plant in Bolzano, Italy. Results show that reliable 24/365 baseload supply has no impact on system flexibility but, at the current storage costs, remains techno-economically unfeasible. In contrast, forecast-based near-firm commitments provide a more effective trade-off between production costs, cost-optimal firm supply and reduction of solar-induced system flexibility. Generation splitting further improves competitiveness by allowing residual variable output to participate in the day-ahead market, lowering production costs while preserving system benefits.