The continued global transition to renewable energy in tandem with the widespread adoption of machine learning demands innovative solutions to enhance the efficiency and reliability of energy systems. Across multiple studies, this dissertation presents a suite of advanced computational and machine learning techniques for sizing, analyzing, and optimizing solar energy systems and megawatt-scale (MW-scale) short-term energy storage solutions. The first study addresses the critical role of interannual variability in solar resources and PV power generation, demonstrating its impact on technoeconomic analyses and the consequences of neglecting long-term variability. The second study employs linear programming to solve an inherently non-linear problem, and simultaneously determine the optimal sizing and dispatch of solar photovoltaic (PV) systems and energy storage for residential complexes, minimizing costs while ensuring energy reliability by continuously meeting energy demand. Building on this foundation, the third study utilizes artificial neural networks to model losses and efficiency mapping in the largest grid-connected MW-scale vanadium redox flow battery operating in the United States. These models are applied to energy arbitrage and peak shaving scenarios, optimizing operational strategies. Finally, the fourth study applies Q-learning, a reinforcement learning approach, to grapple with a stochastic price