High penetration of distributed photovoltaic (PV) systems has significantly increased microgrid net load volatility and uncertainty, posing challenges for conventional point forecasting methods that fail to provide sufficient operational risk information. To address this, this study proposes a multi-step net load forecasting framework that explicitly accounts for renewable energy fluctuations and system dynamics. A multi-quantile model generates 90% confidence prediction intervals for 1 and 4 h horizons at 15 min resolution. To mitigate under-coverage caused by cumulative errors, a stepwise conformal calibration strategy is applied to adjust each forecasting step independently, enhancing interval reliability and consistency. Net load volatility scenarios derived from PV ramping intensity are used to analyze uncertainty evolution under low, medium, and high fluctuation conditions. Case studies based on a high-PV microgrid dataset from eastern China demonstrate that calibrated intervals improve coverage, particularly in high-volatility scenarios, and, when integrated into rolling energy management, enhance battery state-of-charge safety margins and reduce peak grid import with minimal additional cost. The approach maintains point forecast accuracy while providing interpretable net load risk bounds, supporting informed scheduling and demand management in high-renewable microgrids.