Supercritical CO 2 Power Cycles Learning Conditionally Calibrated Equations of State for Direct Fired sCO 2 Cycles with Deep Neural Networks